初始化项目，由ModelHub XC社区提供模型

Model: SeongryongJung/Qwen3-4B-Biology-GRPO-TR Source: Original Platform
2026-10-03 20:15:22 +08:00
commit 7d17112db8
38 changed files with 317443 additions and 0 deletions
--- a/.gitattributes
+++ b/.gitattributes
@@ -0,0 +1,38 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
+results/training_score.png filter=lfs diff=lfs merge=lfs -text
+checkpoints/last/tokenizer.json filter=lfs diff=lfs merge=lfs -text
--- a/README.md
+++ b/README.md
@@ -0,0 +1,158 @@
+---
+license: apache-2.0
+library_name: transformers
+pipeline_tag: text-generation
+tags:
+- qwen3
+- reinforcement-learning
+- grpo
+- trust-region
+- text-generation
+base_model: Qwen/Qwen3-4B
+---
+
+# Qwen3-4B-Biology-GRPO-TR
+
+This repository contains the Qwen3-4B `Biology` `GRPO` batch-size-32 run. The repository name uses the project `GRPO-TR` naming convention, but the actual training method for this checkpoint is GRPO.
+
+The repository root contains the final checkpoint, `global_step_100`. The best validation checkpoint was `global_step_90`, but that step does not contain an actor model checkpoint in the local run directory, so only the final actor checkpoint can be exported. `checkpoints/last/` also contains the same final checkpoint.
+
+## Performance
+
+| Dataset | Method | Base model | Train batch size | Uploaded checkpoint | Uploaded val mean@16 | Best val mean@16 | Best checkpoint | Best actor available |
+|---|---|---|---:|---:|---:|---:|---:|---|
+| Biology | GRPO | Qwen3-4B | 32 | 100 | 42.62% | 46.75% | 90 | no |
+
+![Training and validation scores](results/training_score.png)
+
+## Validation Mean@16
+
+| step | val_mean16 | percent |
+|---:|---:|---:|
+| 10 | 0.326250000000 | 32.62% |
+| 20 | 0.357500000000 | 35.75% |
+| 30 | 0.382500000000 | 38.25% |
+| 40 | 0.353750000000 | 35.38% |
+| 50 | 0.418750000000 | 41.88% |
+| 60 | 0.382500000000 | 38.25% |
+| 70 | 0.395000000000 | 39.50% |
+| 80 | 0.416250000000 | 41.62% |
+| 90 | 0.467500000000 | 46.75% |
+| 100 | 0.426250000000 | 42.62% |
+
+## Detailed Training Hyperparameters
+
+| Section | Parameter | Value | Source |
+|---|---|---:|---|
+| Run identity | `Base model` | `Qwen/Qwen3-4B` | queue/script override |
+| Run identity | `Dataset` | `Biology / SciKnowEval biology` | run_qwen3_generalization.sh |
+| Run identity | `Method` | `GRPO` | run_qwen3_generalization.sh |
+| Run identity | `Config` | `baseline_grpo` | run_qwen3_generalization.sh |
+| Run identity | `Experiment` | `qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8` | run_qwen3_generalization.sh |
+| Run identity | `W&B run` | `run-20260702_102415-94h2xs6b` | wandb |
+| Data | `Train file` | `datasets/sciknoweval/biology/train.parquet` | script override |
+| Data | `Validation file` | `datasets/sciknoweval/biology/test.parquet` | script override |
+| Data | `Train batch size` | `32` | queue/script override |
+| Data | `Train max samples` | `3200` | queue/script override |
+| Data | `Prompt key` | `prompt` | legacy_data.yaml default |
+| Data | `Reward key` | `data_source` | legacy_data.yaml default |
+| Data | `Shuffle train data` | `True` | user.yaml / legacy_data.yaml |
+| Data | `Validation shuffle` | `False` | legacy_data.yaml default |
+| Data | `Filter overlong prompts` | `True` | user.yaml |
+| Data | `Prompt truncation` | `error` | legacy_data.yaml default |
+| Data | `enable_thinking` | `false` | script override |
+| Schedule | `Total training steps` | `100` | queue/script override |
+| Schedule | `Total epochs` | `30` | ppo_trainer/user.yaml default |
+| Schedule | `Validation before train` | `False` | queue/script override |
+| Schedule | `Save frequency` | `10` | queue/script override |
+| Schedule | `Validation frequency` | `10` | queue/script override |
+| Sequence | `Max prompt length` | `2048` | queue/script override |
+| Sequence | `Max response length` | `8192` | queue/script override |
+| Sequence | `Max model length` | `10240` | queue/script override |
+| Sequence | `Actor max token length per GPU` | `10240` | queue/script override |
+| Rollout | `Rollout engine` | `vllm` | user.yaml |
+| Rollout | `Rollout dtype` | `bfloat16` | rollout.yaml default |
+| Rollout | `Train rollout n` | `8` | queue/script override |
+| Rollout | `Train rollout temperature` | `1.0` | script override |
+| Rollout | `Train rollout top_p` | `1.0` | script override |
+| Rollout | `Train rollout do_sample` | `True` | rollout.yaml default |
+| Rollout | `Calculate rollout log probs` | `True` | baseline_grpo.yaml / script override |
+| Rollout | `Max num batched tokens` | `10240` | queue/script override |
+| Rollout | `vLLM GPU memory utilization` | `0.8` | queue/script override |
+| Rollout | `Tensor model parallel size` | `2` | rollout.yaml default |
+| Rollout | `Free cache engine` | `True` | rollout.yaml default |
+| Validation | `Validation rollout n` | `16` | queue/script override |
+| Validation | `Validation temperature` | `0.6` | queue/script override |
+| Validation | `Validation top_p` | `0.95` | queue/script override |
+| Validation | `Validation do_sample` | `True` | queue/script override |
+| Optimization | `Optimizer` | `AdamW` | fsdp optimizer config |
+| Optimization | `Learning rate` | `1e-6` | GRPO method override |
+| Optimization | `LR scheduler` | `constant` | W&B config |
+| Optimization | `LR warmup steps` | `10` | script override |
+| Optimization | `Weight decay` | `0.01` | script override |
+| Optimization | `Betas` | `(0.9, 0.999)` | W&B config |
+| Optimization | `Gradient clip` | `1.0` | script override |
+| PPO/GRPO | `Policy loss mode` | `vanilla` | method override |
+| PPO/GRPO | `Advantage estimator` | `grpo` | baseline_grpo.yaml |
+| PPO/GRPO | `Normalize GRPO advantages by std` | `False` | baseline_grpo.yaml / script override |
+| PPO/GRPO | `PPO epochs` | `1` | W&B config |
+| PPO/GRPO | `PPO mini batch size` | `8` | queue/script override |
+| PPO/GRPO | `PPO micro batch size per GPU` | `1` | user.yaml |
+| PPO/GRPO | `Clip ratio low` | `0.2` | script override |
+| PPO/GRPO | `Clip ratio high` | `0.28` | script override |
+| PPO/GRPO | `Gamma` | `1.0` | ppo_trainer.yaml default |
+| PPO/GRPO | `Lambda` | `1.0` | ppo_trainer.yaml default |
+| PPO/GRPO | `Actor KL loss coef` | `0.0` | method override |
+| PPO/GRPO | `Use KL in reward` | `False` | ppo_trainer/user.yaml |
+| Rollout correction | `Importance sampling mode` | `token` | script override |
+| Rollout correction | `IS threshold` | `2.0` | script override |
+| FSDP/System | `Actor strategy` | `fsdp` | dp_actor.yaml |
+| FSDP/System | `FSDP dtype` | `bfloat16` | W&B config |
+| FSDP/System | `FSDP model dtype` | `fp32` | W&B config |
+| FSDP/System | `Use torch compile` | `True` | W&B config |
+| FSDP/System | `GPUs per node` | `8` | queue/script override |
+| FSDP/System | `Nodes` | `1` | user.yaml |
+| FSDP/System | `GPU type` | `NVIDIA H200` | wandb-metadata |
+| Checkpoint/Logging | `Checkpoint root` | `checkpoints/datasets/sciknoweval/biology` | script override |
+| Checkpoint/Logging | `Latest checkpointed iteration` | `100` | latest_checkpointed_iteration.txt |
+| Checkpoint/Logging | `Max actor checkpoints to keep` | `1` | user.yaml |
+| Checkpoint/Logging | `Logger` | `console, wandb` | ppo_trainer.yaml |
+| Checkpoint/Logging | `W&B entity` | `seongryongjung-chung-ang-university` | environment |
+| Checkpoint/Logging | `W&B project` | `SDPO-root` | user.yaml project_name |
+| Checkpoint/Logging | `W&B group` | `QWEN3-GRPO-generalization` | method override |
+
+Raw result and artifact files:
+
+- `results/validation_mean16.csv`
+- `results/training_scores.csv`
+- `results/hyperparameters.csv`
+- `results/training_score.png`
+- `results/training_score.svg`
+- `artifacts/config.yaml`
+- `artifacts/wandb-summary.json`
+- `artifacts/wandb-metadata.json`
+- `artifacts/output.log`
+- `artifacts/queue.log`
+
+## Usage
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+repo_id = "SeongryongJung/Qwen3-4B-Biology-GRPO-TR"
+tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(
+    repo_id,
+    torch_dtype="auto",
+    device_map="auto",
+    trust_remote_code=True,
+)
+```
+
+## Source
+
+- Checkpoint: `checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8`
+- Uploaded actor checkpoint: `global_step_100/actor`
+- Best validation step: `global_step_90`; actor checkpoint unavailable locally
+- W&B run: `run-20260702_102415-94h2xs6b`
+- Queue log: `artifacts/queue.log`
--- a/added_tokens.json
+++ b/added_tokens.json
@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}
--- a/artifacts/config.yaml
+++ b/artifacts/config.yaml
@@ -0,0 +1,855 @@
+_wandb:
+    value:
+        cli_version: 0.23.1
+        e:
+            7nuolqqbf1cwwdubm79nn7dndxsbbhe6:
+                args:
+                    - --node-ip-address=198.19.44.212
+                    - --node-manager-port=41089
+                    - --object-store-name=/tmp/ray_q3g_biology_grpo_Qwen3_4B/session_2026-07-02_10-22-32_644683_2096785/sockets/plasma_store
+                    - --raylet-name=/tmp/ray_q3g_biology_grpo_Qwen3_4B/session_2026-07-02_10-22-32_644683_2096785/sockets/raylet
+                    - --redis-address=None
+                    - --metrics-agent-port=51010
+                    - --logging-rotate-bytes=536870912
+                    - --logging-rotate-backup-count=5
+                    - --runtime-env-agent-port=64109
+                    - --gcs-address=198.19.44.212:62459
+                    - --session-name=session_2026-07-02_10-22-32_644683_2096785
+                    - --temp-dir=/tmp/ray_q3g_biology_grpo_Qwen3_4B
+                    - --webui=
+                    - --cluster-id=4306b0a7c160c67ff2f68a8cabc9fb42de02add7feb2b9dd3066c9aa
+                    - --startup-token=32
+                    - --worker-launch-time-ms=1782987757164
+                    - --node-id=d852600ad791f29ec1b575f1b787ddb40268e3621fc0726af26802b1
+                    - --runtime-env-hash=1945972661
+                codePath: .venv/lib/python3.12/site-packages/ray/_private/workers/default_worker.py
+                codePathLocal: .venv/lib/python3.12/site-packages/ray/_private/workers/default_worker.py
+                cpu_count: 64
+                cpu_count_logical: 128
+                cudaVersion: "13.0"
+                disk:
+                    /:
+                        total: "46086056050688"
+                        used: "21296067473408"
+                email: jungsr1116@cau.ac.kr
+                executable: /mnt/mole/SDPO/L2T/.venv/bin/python
+                git:
+                    commit: ee6a667700697069b93db636aa937f970c1fb57a
+                    remote: https://github.com/jungseongryong/L2T.git
+                gpu: NVIDIA H200
+                gpu_count: 8
+                gpu_nvidia:
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-c0451621-891c-7169-976d-71d81958db34
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-e2a29dfa-7dad-6cd7-fb05-79d0060f27f0
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-d69bb741-f2a7-a375-7f56-5c0d09d797b2
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-1bbbea1e-7541-3126-1bfc-51b43a67577e
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-6e0055af-452d-3f46-ce21-0c71e347a608
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-317fdb5b-b280-71d8-398e-a843ad897437
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-f19d8161-d2dd-49a1-dcb8-5c341b5e4514
+                    - architecture: Hopper
+                      cudaCores: 16896
+                      memoryTotal: "150754820096"
+                      name: NVIDIA H200
+                      uuid: GPU-961f5ebf-4bf8-6855-828b-c065af57b87c
+                host: mole-workspace-rw
+                memory:
+                    total: "2163980390400"
+                os: Linux-6.8.0-71-generic-x86_64-with-glibc2.36
+                program: /mnt/mole/SDPO/L2T/.venv/lib/python3.12/site-packages/ray/_private/workers/default_worker.py
+                python: CPython 3.12.13
+                root: /mnt/mole/SDPO/L2T
+                startedAt: "2026-07-02T10:24:15.814452Z"
+                writerId: 7nuolqqbf1cwwdubm79nn7dndxsbbhe6
+        m: []
+        python_version: 3.12.13
+        t:
+            "1":
+                - 1
+                - 11
+                - 30
+                - 41
+                - 49
+                - 50
+                - 51
+                - 71
+                - 95
+                - 98
+                - 105
+            "2":
+                - 1
+                - 11
+                - 30
+                - 41
+                - 49
+                - 50
+                - 51
+                - 71
+                - 95
+                - 98
+                - 105
+            "3":
+                - 2
+                - 13
+                - 16
+                - 61
+            "4": 3.12.13
+            "5": 0.23.1
+            "6": 4.57.1
+            "12": 0.23.1
+            "13": linux-x86_64
+actor_rollout_ref:
+    value:
+        actor:
+            _target_: verl.workers.config.FSDPActorConfig
+            calculate_entropy: false
+            calculate_sum_pi_squared: false
+            checkpoint:
+                _target_: verl.trainer.config.CheckpointConfig
+                async_save: false
+                load_contents:
+                    - model
+                    - optimizer
+                    - extra
+                save_contents:
+                    - model
+                    - optimizer
+                    - extra
+            clip_ratio: 0.2
+            clip_ratio_c: 3
+            clip_ratio_high: 0.28
+            clip_ratio_low: 0.2
+            data_loader_seed: 42
+            entropy_checkpointing: false
+            entropy_coeff: 0
+            entropy_from_logits_with_chunking: false
+            freeze_vision_tower: false
+            fsdp_config:
+                _target_: verl.workers.config.FSDPEngineConfig
+                dtype: bfloat16
+                entropy_checkpointing: false
+                entropy_from_logits_with_chunking: false
+                forward_only: false
+                forward_prefetch: false
+                fsdp_size: -1
+                full_determinism: false
+                model_dtype: fp32
+                offload_policy: false
+                optimizer_offload: false
+                param_offload: false
+                reshard_after_forward: true
+                seed: 42
+                strategy: fsdp
+                ulysses_sequence_parallel_size: 1
+                use_orig_params: false
+                use_torch_compile: true
+                wrap_policy:
+                    min_num_params: 0
+            grad_clip: 1
+            kl_loss_coef: 0
+            kl_loss_type: low_var_kl
+            loss_agg_mode: token-mean
+            loss_scale_factor: null
+            optim:
+                _target_: verl.workers.config.FSDPOptimizerConfig
+                betas:
+                    - 0.9
+                    - 0.999
+                clip_grad: 1
+                lr: 1e-06
+                lr_scheduler_type: constant
+                lr_warmup_steps: 10
+                lr_warmup_steps_ratio: 0
+                min_lr_ratio: 0
+                num_cycles: 0.5
+                optimizer: AdamW
+                optimizer_impl: torch.optim
+                override_optimizer_config: null
+                total_training_steps: 100
+                warmup_style: null
+                weight_decay: 0.01
+            policy_loss:
+                _target_: verl.workers.config.PolicyLossConfig
+                clip_cov_lb: 1
+                clip_cov_ratio: 0.0002
+                clip_cov_ub: 5
+                kl_cov_ratio: 0.0002
+                loss_mode: vanilla
+                ppo_kl_coef: 0.1
+            ppo_epochs: 1
+            ppo_max_token_len_per_gpu: 10240
+            ppo_micro_batch_size: null
+            ppo_micro_batch_size_per_gpu: 1
+            ppo_mini_batch_size: 8
+            profiler:
+                _target_: verl.utils.profiler.ProfilerConfig
+                all_ranks: false
+                enable: false
+                ranks: []
+                save_path: outputs/profile
+                tool: null
+                tool_config:
+                    npu:
+                        _target_: verl.utils.profiler.config.NPUToolConfig
+                        analysis: true
+                        contents: []
+                        discrete: false
+                        level: level0
+                    nsys:
+                        _target_: verl.utils.profiler.config.NsightToolConfig
+                        discrete: false
+                    torch:
+                        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+                        step_end: null
+                        step_start: 0
+                    torch_memory:
+                        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+                        stack_depth: 32
+                        trace_alloc_max_entries: 100000
+            rollout_n: 8
+            router_replay:
+                _target_: verl.workers.config.RouterReplayConfig
+                mode: disabled
+                record_file: null
+                replay_file: null
+            self_distillation:
+                _target_: verl.workers.config.SelfDistillationConfig
+                alpha: 0.5
+                distillation_add_tail: true
+                distillation_topk: 100
+                dont_reprompt_on_self_success: true
+                environment_feedback_only_without_solution: true
+                evolving_teacher:
+                    _target_: verl.workers.config.EvolvingTeacherConfig
+                    enable: false
+                    loss_weight: 0
+                    mask: all
+                feedback_template: |4-
+                    The following is feedback from your unsuccessful earlier attempt:
+
+                    {feedback_raw}
+                full_logit_distillation: true
+                include_environment_feedback: true
+                is_clip: 2
+                max_reprompt_len: 22528
+                remove_thinking_from_demonstration: false
+                reprompt_template: |-
+                    {prompt}{solution}{feedback}
+
+                    Correctly solve the original question.
+                reprompt_truncation: right
+                solution_template: |4-
+                    Correct solution:
+
+                    {successful_previous_attempt}
+                srpo_dynamic_weighting: false
+                srpo_dynamic_weighting_temperature: 1
+                success_reward_threshold: 0.5
+                teacher_regularization: ema
+                teacher_update_rate: 0
+                token_reweight_decay_steps: null
+                token_reweight_eps_w: 0.2
+                token_reweight_lambda: 0.5
+            shuffle: false
+            strategy: fsdp
+            sum_pi_squared_checkpointing: false
+            tau_neg: 1.05
+            tau_pos: 1
+            ulysses_sequence_parallel_size: 1
+            use_dynamic_bsz: false
+            use_fused_kernels: false
+            use_kl_loss: false
+            use_prefix_grouper: false
+            use_remove_padding: true
+            use_torch_compile: true
+        hybrid_engine: true
+        model:
+            _target_: verl.workers.config.HFModelConfig
+            custom_chat_template: null
+            enable_activation_offload: false
+            enable_gradient_checkpointing: true
+            exclude_modules: null
+            external_lib: null
+            fused_kernel_options:
+                impl_backend: torch
+            hf_config_path: null
+            lora_adapter_path: null
+            lora_alpha: 16
+            lora_rank: 0
+            path: Qwen/Qwen3-4B
+            target_modules: all-linear
+            tiled_mlp:
+                enabled: false
+                num_shards: 4
+            tokenizer_path: null
+            trust_remote_code: true
+            use_fused_kernels: false
+            use_liger: false
+            use_remove_padding: true
+            use_shm: false
+        nccl_timeout: 600
+        ref:
+            _target_: verl.workers.config.FSDPActorConfig
+            entropy_checkpointing: false
+            entropy_from_logits_with_chunking: false
+            fsdp_config:
+                _target_: verl.workers.config.FSDPEngineConfig
+                dtype: bfloat16
+                entropy_checkpointing: false
+                entropy_from_logits_with_chunking: false
+                forward_only: true
+                forward_prefetch: false
+                fsdp_size: -1
+                full_determinism: false
+                model_dtype: fp32
+                offload_policy: false
+                optimizer_offload: false
+                param_offload: false
+                reshard_after_forward: true
+                seed: 42
+                strategy: fsdp
+                ulysses_sequence_parallel_size: 1
+                use_orig_params: false
+                use_torch_compile: true
+                wrap_policy:
+                    min_num_params: 0
+            log_prob_max_token_len_per_gpu: 10240
+            log_prob_micro_batch_size: null
+            log_prob_micro_batch_size_per_gpu: 1
+            log_prob_use_dynamic_bsz: false
+            profiler:
+                _target_: verl.utils.profiler.ProfilerConfig
+                all_ranks: false
+                enable: false
+                ranks: []
+                save_path: outputs/profile
+                tool: null
+                tool_config:
+                    npu:
+                        _target_: verl.utils.profiler.config.NPUToolConfig
+                        analysis: true
+                        contents: []
+                        discrete: false
+                        level: level0
+                    nsys:
+                        _target_: verl.utils.profiler.config.NsightToolConfig
+                        discrete: false
+                    torch:
+                        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+                        step_end: null
+                        step_start: 0
+                    torch_memory:
+                        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+                        stack_depth: 32
+                        trace_alloc_max_entries: 100000
+            rollout_n: 8
+            router_replay:
+                _target_: verl.workers.config.RouterReplayConfig
+                mode: disabled
+                record_file: null
+                replay_file: null
+            strategy: fsdp
+            ulysses_sequence_parallel_size: 1
+            use_torch_compile: true
+        rollout:
+            _target_: verl.workers.config.RolloutConfig
+            agent:
+                _target_: verl.workers.config.AgentLoopConfig
+                agent_loop_config_path: null
+                custom_async_server:
+                    _target_: verl.workers.config.CustomAsyncServerConfig
+                    name: null
+                    path: null
+                default_agent_loop: single_turn_agent
+                num_workers: 8
+            calculate_log_probs: true
+            cudagraph_capture_sizes: null
+            data_parallel_size: 1
+            disable_log_stats: true
+            do_sample: true
+            dtype: bfloat16
+            enable_chunked_prefill: true
+            enable_prefix_caching: true
+            enable_rollout_routing_replay: false
+            enforce_eager: false
+            expert_parallel_size: 1
+            free_cache_engine: true
+            gpu_memory_utilization: 0.8
+            ignore_eos: false
+            layered_summon: false
+            load_format: dummy
+            log_prob_max_token_len_per_gpu: 10240
+            log_prob_micro_batch_size: null
+            log_prob_micro_batch_size_per_gpu: 1
+            log_prob_use_dynamic_bsz: false
+            logprobs_mode: processed_logprobs
+            max_model_len: 10240
+            max_num_batched_tokens: 10240
+            max_num_seqs: 1024
+            mode: async
+            multi_stage_wake_up: false
+            multi_turn:
+                _target_: verl.workers.config.MultiTurnConfig
+                enable: false
+                format: hermes
+                interaction_config_path: null
+                max_assistant_turns: null
+                max_parallel_calls: 1
+                max_tool_response_length: 256
+                max_user_turns: null
+                num_repeat_rollouts: null
+                tokenization_sanity_check_mode: strict
+                tool_config_path: null
+                tool_response_truncate_side: middle
+                use_inference_chat_template: false
+            "n": 8
+            name: vllm
+            over_sample_rate: 0
+            pipeline_model_parallel_size: 1
+            profiler:
+                _target_: verl.utils.profiler.ProfilerConfig
+                all_ranks: false
+                enable: false
+                ranks: []
+                save_path: outputs/profile
+                tool: null
+                tool_config:
+                    npu:
+                        _target_: verl.utils.profiler.config.NPUToolConfig
+                        analysis: true
+                        contents: []
+                        discrete: false
+                        level: level0
+                    nsys:
+                        _target_: verl.utils.profiler.config.NsightToolConfig
+                        discrete: false
+                    torch:
+                        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+                        step_end: null
+                        step_start: 0
+                    torch_memory:
+                        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+                        stack_depth: 32
+                        trace_alloc_max_entries: 100000
+            prometheus:
+                _target_: verl.workers.config.PrometheusConfig
+                enable: false
+                file: /tmp/ray/session_latest/metrics/prometheus/prometheus.yml
+                port: 9090
+                served_model_name: Qwen/Qwen3-4B
+            prompt_length: 2048
+            quantization: null
+            quantization_config_file: null
+            response_length: 8192
+            scheduling_policy: fcfs
+            skip_dump_dir: /tmp/rollout_dump
+            skip_rollout: false
+            skip_tokenizer_init: true
+            temperature: 1
+            tensor_model_parallel_size: 2
+            top_k: -1
+            top_p: 1
+            trace:
+                _target_: verl.workers.config.TraceConfig
+                backend: null
+                max_samples_per_step_per_worker: null
+                token2text: false
+            update_weights_bucket_megabytes: 512
+            val_kwargs:
+                _target_: verl.workers.config.SamplingConfig
+                do_sample: true
+                "n": 16
+                temperature: 0.6
+                top_k: -1
+                top_p: 0.95
+algorithm:
+    value:
+        _target_: verl.trainer.config.AlgoConfig
+        adv_estimator: grpo
+        gamma: 1
+        kl_ctrl:
+            _target_: verl.trainer.config.KLControlConfig
+            horizon: 10000
+            kl_coef: 0.001
+            target_kl: 0.1
+            type: fixed
+        kl_penalty: kl
+        lam: 1
+        norm_adv_by_std_in_grpo: false
+        pf_ppo:
+            reweight_method: pow
+            weight_pow: 2
+        rollout_correction:
+            bypass_mode: false
+            loss_type: ppo_clip
+            rollout_is: token
+            rollout_is_batch_normalize: false
+            rollout_is_threshold: 2
+            rollout_rs: null
+            rollout_rs_threshold: null
+        use_kl_in_reward: false
+        use_pf_ppo: false
+critic:
+    value:
+        _target_: verl.workers.config.FSDPCriticConfig
+        checkpoint:
+            _target_: verl.trainer.config.CheckpointConfig
+            async_save: false
+            load_contents:
+                - model
+                - optimizer
+                - extra
+            save_contents:
+                - model
+                - optimizer
+                - extra
+        cliprange_value: 0.5
+        data_loader_seed: 42
+        enable: null
+        forward_max_token_len_per_gpu: 32768
+        forward_micro_batch_size: null
+        forward_micro_batch_size_per_gpu: null
+        grad_clip: 1
+        loss_agg_mode: token-mean
+        model:
+            _target_: verl.workers.config.FSDPCriticModelCfg
+            enable_activation_offload: false
+            enable_gradient_checkpointing: true
+            external_lib: null
+            fsdp_config:
+                _target_: verl.workers.config.FSDPEngineConfig
+                dtype: bfloat16
+                entropy_checkpointing: false
+                entropy_from_logits_with_chunking: false
+                forward_only: false
+                forward_prefetch: false
+                fsdp_size: -1
+                full_determinism: false
+                model_dtype: fp32
+                offload_policy: false
+                optimizer_offload: false
+                param_offload: false
+                reshard_after_forward: true
+                seed: 42
+                strategy: fsdp
+                ulysses_sequence_parallel_size: 1
+                use_orig_params: false
+                use_torch_compile: true
+                wrap_policy:
+                    min_num_params: 0
+            lora_alpha: 16
+            lora_rank: 0
+            path: Qwen/Qwen3-8B
+            target_modules: all-linear
+            tiled_mlp:
+                enabled: false
+                num_shards: 4
+            tokenizer_path: Qwen/Qwen3-4B
+            trust_remote_code: true
+            use_remove_padding: false
+            use_shm: false
+        optim:
+            _target_: verl.workers.config.FSDPOptimizerConfig
+            betas:
+                - 0.9
+                - 0.999
+            clip_grad: 1
+            lr: 1e-05
+            lr_scheduler_type: constant
+            lr_warmup_steps: -1
+            lr_warmup_steps_ratio: 0
+            min_lr_ratio: 0
+            num_cycles: 0.5
+            optimizer: AdamW
+            optimizer_impl: torch.optim
+            override_optimizer_config: null
+            total_training_steps: 100
+            warmup_style: null
+            weight_decay: 0.01
+        ppo_epochs: 1
+        ppo_max_token_len_per_gpu: 32768
+        ppo_micro_batch_size: null
+        ppo_micro_batch_size_per_gpu: null
+        ppo_mini_batch_size: 8
+        profiler:
+            _target_: verl.utils.profiler.ProfilerConfig
+            all_ranks: false
+            enable: false
+            ranks: []
+            save_path: outputs/profile
+            tool: null
+            tool_config:
+                npu:
+                    _target_: verl.utils.profiler.config.NPUToolConfig
+                    analysis: true
+                    contents: []
+                    discrete: false
+                    level: level0
+                nsys:
+                    _target_: verl.utils.profiler.config.NsightToolConfig
+                    discrete: false
+                torch:
+                    _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+                    step_end: null
+                    step_start: 0
+                torch_memory:
+                    _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+                    stack_depth: 32
+                    trace_alloc_max_entries: 100000
+        rollout_n: 8
+        shuffle: false
+        strategy: fsdp
+        ulysses_sequence_parallel_size: 1
+        use_dynamic_bsz: false
+custom_reward_function:
+    value:
+        name: compute_score
+        path: /mnt/mole/SDPO/L2T/verl/utils/reward_score/feedback/__init__.py
+data:
+    value:
+        apply_chat_template_kwargs:
+            enable_thinking: false
+        custom_cls:
+            name: null
+            path: null
+        datagen:
+            name: null
+            path: null
+        dataloader_num_workers: 8
+        filter_overlong_prompts: true
+        filter_overlong_prompts_workers: 1
+        image_key: images
+        image_patch_size: 14
+        max_prompt_length: 2048
+        max_response_length: 8192
+        prompt_key: prompt
+        return_full_prompt: false
+        return_multi_modal_inputs: true
+        return_raw_chat: true
+        return_raw_input_ids: false
+        reward_fn_key: data_source
+        sampler:
+            class_name: null
+            class_path: null
+        seed: null
+        shuffle: true
+        tokenizer: null
+        tool_config_path: null
+        train_batch_size: 32
+        train_files:
+            - /mnt/mole/SDPO/L2T/datasets/sciknoweval/biology/train.parquet
+        train_max_samples: 3200
+        truncation: error
+        trust_remote_code: true
+        use_shm: false
+        val_batch_size: null
+        val_files:
+            - /mnt/mole/SDPO/L2T/datasets/sciknoweval/biology/test.parquet
+        val_max_samples: -1
+        validation_shuffle: false
+        video_key: videos
+global_profiler:
+    value:
+        _target_: verl.utils.profiler.ProfilerConfig
+        global_tool_config:
+            nsys:
+                _target_: verl.utils.profiler.config.NsightToolConfig
+                controller_nsight_options:
+                    cuda-graph-trace: graph
+                    cuda-memory-usage: "true"
+                    trace: cuda,nvtx,cublas,ucx
+                discrete: false
+                worker_nsight_options:
+                    capture-range: cudaProfilerApi
+                    capture-range-end: null
+                    cuda-graph-trace: graph
+                    cuda-memory-usage: "true"
+                    kill: none
+                    trace: cuda,nvtx,cublas,ucx
+            torch_memory:
+                context: all
+                stack_depth: 32
+                stacks: all
+                trace_alloc_max_entries: 100000
+        profile_continuous_steps: false
+        save_path: outputs/profile
+        steps: null
+        tool: null
+max_model_len:
+    value: 10240
+ray_kwargs:
+    value:
+        ray_init:
+            _temp_dir: /tmp/ray_q3g_biology_grpo_Qwen3_4B
+            include_dashboard: false
+            num_cpus: null
+            runtime_env:
+                env_vars:
+                    EXPERIMENT: qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8
+                    TASK: datasets/sciknoweval/biology
+                    USER: root
+        timeline_json_file: null
+reward_manager:
+    value:
+        _target_: verl.trainer.config.config.RewardManagerConfig
+        module:
+            _target_: verl.trainer.config.config.ModuleConfig
+            name: custom_reward_manager
+            path: null
+        name: naive
+        source: register
+reward_model:
+    value:
+        enable: false
+        enable_resource_pool: false
+        forward_max_token_len_per_gpu: 32768
+        launch_reward_fn_async: false
+        max_length: null
+        micro_batch_size: null
+        micro_batch_size_per_gpu: null
+        model:
+            external_lib: null
+            fsdp_config:
+                _target_: verl.workers.config.FSDPEngineConfig
+                forward_prefetch: false
+                fsdp_size: -1
+                param_offload: false
+                reshard_after_forward: true
+                wrap_policy:
+                    min_num_params: 0
+            input_tokenizer: Qwen/Qwen3-4B
+            path: ~/models/FsfairX-LLaMA3-RM-v0.1
+            trust_remote_code: false
+            use_fused_kernels: false
+            use_remove_padding: false
+            use_shm: false
+        n_gpus_per_node: 8
+        nnodes: 0
+        num_workers: 1
+        profiler:
+            _target_: verl.utils.profiler.ProfilerConfig
+            all_ranks: false
+            enable: false
+            ranks: []
+            save_path: outputs/profile
+            tool: null
+            tool_config:
+                npu:
+                    _target_: verl.utils.profiler.config.NPUToolConfig
+                    analysis: true
+                    contents: []
+                    discrete: false
+                    level: level0
+                nsys:
+                    _target_: verl.utils.profiler.config.NsightToolConfig
+                    discrete: false
+                torch:
+                    _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+                    step_end: null
+                    step_start: 0
+                torch_memory:
+                    _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+                    stack_depth: 32
+                    trace_alloc_max_entries: 100000
+        reward_loop_class_name: null
+        reward_loop_module_path: null
+        reward_loop_source: register
+        reward_manager: naive
+        rollout:
+            _target_: verl.workers.config.RolloutConfig
+            cudagraph_capture_sizes: null
+            data_parallel_size: 1
+            disable_log_stats: true
+            dtype: bfloat16
+            enable_chunked_prefill: true
+            enable_prefix_caching: true
+            enforce_eager: true
+            expert_parallel_size: 1
+            free_cache_engine: true
+            gpu_memory_utilization: 0.5
+            limit_images: null
+            load_format: auto
+            max_model_len: null
+            max_num_batched_tokens: 8192
+            max_num_seqs: 1024
+            name: ???
+            prompt_length: 2048
+            response_length: 2048
+            skip_tokenizer_init: false
+            tensor_model_parallel_size: 2
+        sandbox_fusion:
+            max_concurrent: 64
+            memory_limit_mb: 1024
+            url: null
+        strategy: fsdp
+        ulysses_sequence_parallel_size: 1
+        use_dynamic_bsz: false
+        use_reward_loop: false
+trainer:
+    value:
+        balance_batch: true
+        critic_warmup: 0
+        default_hdfs_dir: null
+        default_local_dir: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8
+        del_local_ckpt_after_load: false
+        device: cuda
+        esi_redundant_time: 0
+        experiment_name: qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8
+        group_name: QWEN3-GRPO-generalization
+        log_val_generations: 0
+        logger:
+            - console
+            - wandb
+        max_actor_ckpt_to_keep: 1
+        max_critic_ckpt_to_keep: null
+        n_gpus_per_node: 8
+        nnodes: 1
+        project_name: SDPO-root
+        ray_wait_register_center_timeout: 300
+        resume_from_path: null
+        resume_mode: auto
+        rollout_data_dir: null
+        save_freq: 10
+        test_freq: 10
+        total_epochs: 30
+        total_training_steps: 100
+        use_legacy_worker_impl: auto
+        val_before_train: false
+        val_only: false
+        validation_data_dir: null
+transfer_queue:
+    value:
+        enable: false
+vars:
+    value:
+        ckpt_dir: /capstor/scratch/cscs/root/ttrl_runs/datasets/sciknoweval/biology
+        dir: /users/root/SDPO
+        log_dir: /users/root/output
+        task: datasets/sciknoweval/biology
--- a/artifacts/output.log
+++ b/artifacts/output.log
@@ -0,0 +1,668 @@
+wandb: Detected [openai] in use.
+wandb: Use W&B Weave for improved LLM call tracing. Install Weave with `pip install weave` then add `import weave` to the top of your script.
+wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
+Checkpoint tracker file does not exist: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/latest_checkpointed_iteration.txt
+Training from scratch
+Training Progress:   9%|▉         | 9/100 [08:46<1:11:43, 47.30s/it]
+step:1 - global_seqlen/min:16790 - global_seqlen/max:24040 - global_seqlen/minmax_diff:7250 - global_seqlen/balanced_min:20396 - global_seqlen/balanced_max:21304 - global_seqlen/mean:20587.0 - actor/entropy:0.15397636592388153 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5688784718513489 - training/rollout_probs_diff_mean:0.005099463742226362 - training/rollout_probs_diff_std:0.01653202250599861 - training/rollout_actor_probs_pearson_corr:0.9964430928230286 - rollout_corr/rollout_is_mean:0.9999783039093018 - rollout_corr/rollout_is_ratio_fraction_high:0.00016134980251081288 - rollout_corr/rollout_is_ratio_fraction_low:0.0004148994921706617 - rollout_corr/rollout_is_max:3.5453882217407227 - rollout_corr/rollout_is_min:0.07613740861415863 - rollout_corr/rollout_is_std:0.047128625214099884 - rollout_corr/rollout_is_eff_sample_size:0.9977836962209125 - rollout_corr/rollout_is_seq_mean:1.0001716613769531 - rollout_corr/rollout_is_seq_std:0.003117709420621395 - rollout_corr/rollout_is_seq_max:1.0116699934005737 - rollout_corr/rollout_is_seq_min:0.9897567629814148 - rollout_corr/rollout_is_seq_max_deviation:0.01166999340057373 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2011112431064248) - rollout_corr/training_log_ppl:np.float64(0.18095037681632675) - rollout_corr/kl:np.float64(0.001171631188483957) - rollout_corr/k3_kl:np.float64(0.0013433933762030392) - rollout_corr/rollout_ppl:np.float64(1.1996783446520567) - rollout_corr/rollout_log_ppl:np.float64(0.17977874411735684) - rollout_corr/log_ppl_diff:np.float64(0.0011716326989699155) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025823632313404232) - rollout_corr/log_ppl_diff_max:np.float64(0.011865988373756409) - rollout_corr/log_ppl_diff_min:np.float64(-0.010793805122375488) - rollout_corr/ppl_ratio:np.float64(1.0011774415615946) - rollout_corr/chi2_token:np.float64(0.0029977981466799974) - rollout_corr/chi2_seq:np.float64(0.6187159691471606) - actor/pg_loss:np.float64(-0.00026798469480127096) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0) - actor/ppo_kl:np.float64(0.0) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(1.2398988455533981) - perf/mfu/actor:np.float64(0.056215164928366285) - perf/max_memory_allocated_gb:np.float64(116.22064924240112) - perf/max_memory_reserved_gb:np.float64(121.841796875) - perf/cpu_memory_used_gb:np.float64(98.29722595214844) - actor/lr:np.float64(0.0) - training/global_step:1 - training/epoch:0 - critic/score/mean:0.3515625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.3515625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.015277809463441372 - critic/advantages/max:0.875 - critic/advantages/min:-0.75 - critic/returns/mean:-0.015277809463441372 - critic/returns/max:0.875 - critic/returns/min:-0.75 - response_length/mean:338.9375 - response_length/max:2624.0 - response_length/min:76.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:338.9375 - response_length_non_aborted/max:2624.0 - response_length_non_aborted/min:76.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:304.40625 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00023351609706878662 - timing_s/agent_loop/generate_sequences/min:np.float64(10.843628607690334) - timing_s/agent_loop/generate_sequences/max:np.float64(27.492554239928722) - timing_s/agent_loop/generate_sequences/mean:np.float64(16.257467001851182) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(27.492554239928722) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent_loop/slowest/pro
+step:2 - global_seqlen/min:17355 - global_seqlen/max:33543 - global_seqlen/minmax_diff:16188 - global_seqlen/balanced_min:23981 - global_seqlen/balanced_max:29649 - global_seqlen/mean:24721.0 - actor/entropy:0.12396878004074097 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5429902076721191 - training/rollout_probs_diff_mean:0.0038451729342341423 - training/rollout_probs_diff_std:0.01370719913393259 - training/rollout_actor_probs_pearson_corr:0.9971049427986145 - rollout_corr/rollout_is_mean:1.0001349449157715 - rollout_corr/rollout_is_ratio_fraction_high:6.08441696385853e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00015645642997696996 - rollout_corr/rollout_is_max:2.5501418113708496 - rollout_corr/rollout_is_min:0.037313781678676605 - rollout_corr/rollout_is_std:0.03914785757660866 - rollout_corr/rollout_is_eff_sample_size:0.9984702658823585 - rollout_corr/rollout_is_seq_mean:1.000056266784668 - rollout_corr/rollout_is_seq_std:0.0027974694967269897 - rollout_corr/rollout_is_seq_max:1.0097461938858032 - rollout_corr/rollout_is_seq_min:0.9896620512008667 - rollout_corr/rollout_is_seq_max_deviation:0.0103379487991333 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.1933515202254057) - rollout_corr/training_log_ppl:np.float64(0.1738216859230306) - rollout_corr/kl:np.float64(0.0010194054242695927) - rollout_corr/k3_kl:np.float64(0.001138604666266474) - rollout_corr/rollout_ppl:np.float64(1.1920951255597174) - rollout_corr/rollout_log_ppl:np.float64(0.17280227995070163) - rollout_corr/log_ppl_diff:np.float64(0.0010194059723289683) - rollout_corr/log_ppl_abs_diff:np.float64(0.0021742572280345485) - rollout_corr/log_ppl_diff_max:np.float64(0.011997908353805542) - rollout_corr/log_ppl_diff_min:np.float64(-0.00812309980392456) - rollout_corr/ppl_ratio:np.float64(1.0010238848626614) - rollout_corr/chi2_token:np.float64(0.002517439192160964) - rollout_corr/chi2_seq:np.float64(2.8077984978444874) - actor/pg_loss:np.float64(0.0005324010271579027) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0014159599388676725) - actor/ppo_kl:np.float64(-2.703549300875352e-05) - actor/pg_clipfrac_lower:np.float64(1.385194991598837e-05) - actor/grad_norm:np.float64(0.6834095269441605) - perf/mfu/actor:np.float64(0.06606670600642084) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(98.88320541381836) - actor/lr:np.float64(1e-07) - training/global_step:2 - training/epoch:0 - critic/score/mean:0.36328125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.36328125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.04914579167962074 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.04914579167962074 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:449.40625 - response_length/max:8192.0 - response_length/min:76.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:449.40625 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:76.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:323.125 - prompt_length/max:500.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00016132742166519165 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9411534667015076) - timing_s/agent_loop/generate_sequences/max:np.float64(50.398399429395795) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.151348275081546) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(50.398399429395795) - timing_s/agent_loop/slow
+step:3 - global_seqlen/min:14081 - global_seqlen/max:27464 - global_seqlen/minmax_diff:13383 - global_seqlen/balanced_min:22920 - global_seqlen/balanced_max:29234 - global_seqlen/mean:23729.25 - actor/entropy:0.1342262476682663 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.9211333394050598 - training/rollout_probs_diff_mean:0.004450538661330938 - training/rollout_probs_diff_std:0.015509124845266342 - training/rollout_actor_probs_pearson_corr:0.9964831471443176 - rollout_corr/rollout_is_mean:1.0001475811004639 - rollout_corr/rollout_is_ratio_fraction_high:0.00012680205691140145 - rollout_corr/rollout_is_ratio_fraction_low:0.0002633581170812249 - rollout_corr/rollout_is_max:2.8697574138641357 - rollout_corr/rollout_is_min:0.014595741406083107 - rollout_corr/rollout_is_std:0.043651361018419266 - rollout_corr/rollout_is_eff_sample_size:0.9980987762991804 - rollout_corr/rollout_is_seq_mean:1.0002849102020264 - rollout_corr/rollout_is_seq_std:0.0030059556011110544 - rollout_corr/rollout_is_seq_max:1.0112146139144897 - rollout_corr/rollout_is_seq_min:0.9923873543739319 - rollout_corr/rollout_is_seq_max_deviation:0.011214613914489746 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.1883889199234545) - rollout_corr/training_log_ppl:np.float64(0.17043527173973416) - rollout_corr/kl:np.float64(0.0009068544875852069) - rollout_corr/k3_kl:np.float64(0.001184503076672172) - rollout_corr/rollout_ppl:np.float64(1.1872890815138817) - rollout_corr/rollout_log_ppl:np.float64(0.16952841756028647) - rollout_corr/log_ppl_diff:np.float64(0.0009068541794476914) - rollout_corr/log_ppl_abs_diff:np.float64(0.002299016089637007) - rollout_corr/log_ppl_diff_max:np.float64(0.009402722120285034) - rollout_corr/log_ppl_diff_min:np.float64(-0.010235980153083801) - rollout_corr/ppl_ratio:np.float64(1.0009116898290813) - rollout_corr/chi2_token:np.float64(0.002880368148908019) - rollout_corr/chi2_seq:np.float64(2.019863748457283) - actor/pg_loss:np.float64(0.00017959426622837782) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0016469392169256025) - actor/ppo_kl:np.float64(-1.347891979008864e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7490953952074051) - perf/mfu/actor:np.float64(0.06393039704140063) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(98.90531921386719) - actor/lr:np.float64(2e-07) - training/global_step:3 - training/epoch:0 - critic/score/mean:0.37890625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.37890625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.04369306191802025 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.04369306191802025 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:400.4765625 - response_length/max:8192.0 - response_length/min:88.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:400.4765625 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:88.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:341.0625 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.72989946603775e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0602107234299183) - timing_s/agent_loop/generate_sequences/max:np.float64(49.61351806856692) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.787061496055685) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.61351806856692) - timing_s/agent_loop/slowest/tool_ca
+step:4 - global_seqlen/min:22199 - global_seqlen/max:32093 - global_seqlen/minmax_diff:9894 - global_seqlen/balanced_min:25247 - global_seqlen/balanced_max:30867 - global_seqlen/mean:27681.75 - actor/entropy:0.11408453434705734 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.4004265367984772 - training/rollout_probs_diff_mean:0.003464761655777693 - training/rollout_probs_diff_std:0.0128159886226058 - training/rollout_actor_probs_pearson_corr:0.9972761869430542 - rollout_corr/rollout_is_mean:1.0000988245010376 - rollout_corr/rollout_is_ratio_fraction_high:6.781295815017074e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00010548681893851608 - rollout_corr/rollout_is_max:2.4635965824127197 - rollout_corr/rollout_is_min:0.25120973587036133 - rollout_corr/rollout_is_std:0.03723444789648056 - rollout_corr/rollout_is_eff_sample_size:0.9986157531764583 - rollout_corr/rollout_is_seq_mean:0.9999723434448242 - rollout_corr/rollout_is_seq_std:0.003101256676018238 - rollout_corr/rollout_is_seq_max:1.0128250122070312 - rollout_corr/rollout_is_seq_min:0.990000307559967 - rollout_corr/rollout_is_seq_max_deviation:0.01282501220703125 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2040854450315237) - rollout_corr/training_log_ppl:np.float64(0.18203230939070636) - rollout_corr/kl:np.float64(0.0011662654592186072) - rollout_corr/k3_kl:np.float64(0.0010397648765376744) - rollout_corr/rollout_ppl:np.float64(1.2026153546757996) - rollout_corr/rollout_log_ppl:np.float64(0.18086604351083224) - rollout_corr/log_ppl_diff:np.float64(0.0011662658798741177) - rollout_corr/log_ppl_abs_diff:np.float64(0.0023683869803790003) - rollout_corr/log_ppl_diff_max:np.float64(0.010118335485458374) - rollout_corr/log_ppl_diff_min:np.float64(-0.010249346494674683) - rollout_corr/ppl_ratio:np.float64(1.0011713872663677) - rollout_corr/chi2_token:np.float64(0.0018216201569885015) - rollout_corr/chi2_seq:np.float64(1.021278945961967) - actor/pg_loss:np.float64(0.00038540177047252655) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0012221543156556436) - actor/ppo_kl:np.float64(8.313983751406795e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7838278859853745) - perf/mfu/actor:np.float64(0.06802962668683335) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(99.51873397827148) - actor/lr:np.float64(3e-07) - training/global_step:4 - training/epoch:0 - critic/score/mean:0.4140625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4140625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.07241576164960861 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.07241576164960861 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:518.4296875 - response_length/max:8192.0 - response_length/min:84.0 - response_length/clip_ratio:0.01171875 - response_length_non_aborted/mean:518.4296875 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:84.0 - response_length_non_aborted/clip_ratio:0.01171875 - response/aborted_ratio:0.0 - prompt_length/mean:346.625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00014001689851284027 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1103171762079) - timing_s/agent_loop/generate_sequences/max:np.float64(49.865499056875706) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.580142610364419) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.865499056875706) - timing_s/agent_loop/slowest/tool_calls:np.
+step:5 - global_seqlen/min:15239 - global_seqlen/max:31801 - global_seqlen/minmax_diff:16562 - global_seqlen/balanced_min:21853 - global_seqlen/balanced_max:27983 - global_seqlen/mean:22706.25 - actor/entropy:0.14480802416801453 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7759525179862976 - training/rollout_probs_diff_mean:0.004493028856813908 - training/rollout_probs_diff_std:0.015260254964232445 - training/rollout_actor_probs_pearson_corr:0.9967924952507019 - rollout_corr/rollout_is_mean:0.9999231696128845 - rollout_corr/rollout_is_ratio_fraction_high:3.944072886952199e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00016762310406193137 - rollout_corr/rollout_is_max:3.7939155101776123 - rollout_corr/rollout_is_min:0.08352697640657425 - rollout_corr/rollout_is_std:0.04204750806093216 - rollout_corr/rollout_is_eff_sample_size:0.9982348897314802 - rollout_corr/rollout_is_seq_mean:0.9998871684074402 - rollout_corr/rollout_is_seq_std:0.002979286713525653 - rollout_corr/rollout_is_seq_max:1.009093999862671 - rollout_corr/rollout_is_seq_min:0.9903265237808228 - rollout_corr/rollout_is_seq_max_deviation:0.009673476219177246 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2126561030745506) - rollout_corr/training_log_ppl:np.float64(0.1898149242515501) - rollout_corr/kl:np.float64(0.001202444607415032) - rollout_corr/k3_kl:np.float64(0.0012296490901917423) - rollout_corr/rollout_ppl:np.float64(1.2111586891114712) - rollout_corr/rollout_log_ppl:np.float64(0.18861247888708021) - rollout_corr/log_ppl_diff:np.float64(0.0012024453644698951) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025688183632155415) - rollout_corr/log_ppl_diff_max:np.float64(0.013211295008659363) - rollout_corr/log_ppl_diff_min:np.float64(-0.006566941738128662) - rollout_corr/ppl_ratio:np.float64(1.0012085342314094) - rollout_corr/chi2_token:np.float64(0.0025670863687992096) - rollout_corr/chi2_seq:np.float64(2.749309702543542) - actor/pg_loss:np.float64(-1.9733794033527374e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001807897925573343) - actor/ppo_kl:np.float64(-7.232862377021254e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.8762328773736954) - perf/mfu/actor:np.float64(0.06084690245160743) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(105.3316879272461) - actor/lr:np.float64(4e-07) - training/global_step:5 - training/epoch:0 - critic/score/mean:0.4609375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4609375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.057222336530685425 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.057222336530685425 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:396.1640625 - response_length/max:8192.0 - response_length/min:88.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:396.1640625 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:88.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:313.40625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.857118129730225e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0833084657788277) - timing_s/agent_loop/generate_sequences/max:np.float64(49.36671530082822) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.734902196949406) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.36671530082822) - timing_s/agent_loop/slowest/tool_ca
+step:6 - global_seqlen/min:18376 - global_seqlen/max:31258 - global_seqlen/minmax_diff:12882 - global_seqlen/balanced_min:22981 - global_seqlen/balanced_max:28385 - global_seqlen/mean:23683.0 - actor/entropy:0.13729968667030334 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.45647841691970825 - training/rollout_probs_diff_mean:0.004255471285432577 - training/rollout_probs_diff_std:0.014458276331424713 - training/rollout_actor_probs_pearson_corr:0.9969208240509033 - rollout_corr/rollout_is_mean:1.0000715255737305 - rollout_corr/rollout_is_ratio_fraction_high:6.439505523303524e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00010119222861249 - rollout_corr/rollout_is_max:3.7961511611938477 - rollout_corr/rollout_is_min:0.06711025536060333 - rollout_corr/rollout_is_std:0.041316788643598557 - rollout_corr/rollout_is_eff_sample_size:0.9982958321194707 - rollout_corr/rollout_is_seq_mean:0.9999457597732544 - rollout_corr/rollout_is_seq_std:0.002895613433793187 - rollout_corr/rollout_is_seq_max:1.0087769031524658 - rollout_corr/rollout_is_seq_min:0.9885023236274719 - rollout_corr/rollout_is_seq_max_deviation:0.011497676372528076 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.1988565418869257) - rollout_corr/training_log_ppl:np.float64(0.17875455050671007) - rollout_corr/kl:np.float64(0.001162382649852134) - rollout_corr/k3_kl:np.float64(0.001047637722351169) - rollout_corr/rollout_ppl:np.float64(1.1974089476279914) - rollout_corr/rollout_log_ppl:np.float64(0.17759216712147463) - rollout_corr/log_ppl_diff:np.float64(0.0011623833852354437) - rollout_corr/log_ppl_abs_diff:np.float64(0.0023096474760677665) - rollout_corr/log_ppl_diff_max:np.float64(0.013013452291488647) - rollout_corr/log_ppl_diff_min:np.float64(-0.006559252738952637) - rollout_corr/ppl_ratio:np.float64(1.0011671690735966) - rollout_corr/chi2_token:np.float64(0.0018661606591194868) - rollout_corr/chi2_seq:np.float64(2.0125787355937064) - actor/pg_loss:np.float64(4.9162888899445534e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015258892635756638) - actor/ppo_kl:np.float64(1.2128886957896157e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6967141479253769) - perf/mfu/actor:np.float64(0.06450282034769986) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(105.45068359375) - actor/lr:np.float64(5e-07) - training/global_step:6 - training/epoch:0 - critic/score/mean:0.32421875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.32421875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.02214614860713482 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.02214614860713482 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:424.625 - response_length/max:7663.0 - response_length/min:124.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:424.625 - response_length_non_aborted/max:7663.0 - response_length_non_aborted/min:124.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:315.46875 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.42964106798172e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.3253010343760252) - timing_s/agent_loop/generate_sequences/max:np.float64(39.50061544589698) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.015933719550958) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(39.50061544589698) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - t
+step:7 - global_seqlen/min:15713 - global_seqlen/max:27570 - global_seqlen/minmax_diff:11857 - global_seqlen/balanced_min:18845 - global_seqlen/balanced_max:25356 - global_seqlen/mean:20549.25 - actor/entropy:0.1488693803548813 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.8907374739646912 - training/rollout_probs_diff_mean:0.004379427060484886 - training/rollout_probs_diff_std:0.014803227037191391 - training/rollout_actor_probs_pearson_corr:0.9970603585243225 - rollout_corr/rollout_is_mean:0.9998552203178406 - rollout_corr/rollout_is_ratio_fraction_high:1.0627218216541223e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0002550532517489046 - rollout_corr/rollout_is_max:2.3208792209625244 - rollout_corr/rollout_is_min:0.0981237068772316 - rollout_corr/rollout_is_std:0.040721017867326736 - rollout_corr/rollout_is_eff_sample_size:0.9983443062319399 - rollout_corr/rollout_is_seq_mean:0.9998733997344971 - rollout_corr/rollout_is_seq_std:0.0032510175369679928 - rollout_corr/rollout_is_seq_max:1.0084893703460693 - rollout_corr/rollout_is_seq_min:0.9903331398963928 - rollout_corr/rollout_is_seq_max_deviation:0.009666860103607178 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.22252489766106) - rollout_corr/training_log_ppl:np.float64(0.19787568087485852) - rollout_corr/kl:np.float64(0.001285289514086685) - rollout_corr/k3_kl:np.float64(0.0011396203050537679) - rollout_corr/rollout_ppl:np.float64(1.220958977472037) - rollout_corr/rollout_log_ppl:np.float64(0.19659039181715343) - rollout_corr/log_ppl_diff:np.float64(0.0012852890577050857) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027941703374381177) - rollout_corr/log_ppl_diff_max:np.float64(0.009587958455085754) - rollout_corr/log_ppl_diff_min:np.float64(-0.009744793176651001) - rollout_corr/ppl_ratio:np.float64(1.0012915702536702) - rollout_corr/chi2_token:np.float64(0.0019585033878684044) - rollout_corr/chi2_seq:np.float64(0.5178608226124197) - actor/pg_loss:np.float64(0.00033794844057410955) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015785431330641586) - actor/ppo_kl:np.float64(5.3518397457708033e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.8111859112977982) - perf/mfu/actor:np.float64(0.05684415790261006) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(105.71798706054688) - actor/lr:np.float64(6e-07) - training/global_step:7 - training/epoch:0 - critic/score/mean:0.29296875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.29296875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0651807188987732 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0651807188987732 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:367.5703125 - response_length/max:8192.0 - response_length/min:78.0 - response_length/clip_ratio:0.0078125 - response_length_non_aborted/mean:367.5703125 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:78.0 - response_length_non_aborted/clip_ratio:0.0078125 - response/aborted_ratio:0.0 - prompt_length/mean:274.59375 - prompt_length/max:498.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:3.2745301723480225e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9346938598901033) - timing_s/agent_loop/generate_sequences/max:np.float64(51.31893388926983) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.4569646156887757) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(51.31893388926983) - timing_s/agent_loop/slowest/tool_cal
+step:8 - global_seqlen/min:13389 - global_seqlen/max:33652 - global_seqlen/minmax_diff:20263 - global_seqlen/balanced_min:22939 - global_seqlen/balanced_max:23096 - global_seqlen/mean:22998.375 - actor/entropy:0.14975731074810028 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5948810577392578 - training/rollout_probs_diff_mean:0.004444846883416176 - training/rollout_probs_diff_std:0.014590873382985592 - training/rollout_actor_probs_pearson_corr:0.9971561431884766 - rollout_corr/rollout_is_mean:1.0001192092895508 - rollout_corr/rollout_is_ratio_fraction_high:5.916984810028225e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00010847805242519826 - rollout_corr/rollout_is_max:3.6715524196624756 - rollout_corr/rollout_is_min:0.015551368705928326 - rollout_corr/rollout_is_std:0.04151970148086548 - rollout_corr/rollout_is_eff_sample_size:0.9982794375291114 - rollout_corr/rollout_is_seq_mean:1.000015377998352 - rollout_corr/rollout_is_seq_std:0.0027894889935851097 - rollout_corr/rollout_is_seq_max:1.0083576440811157 - rollout_corr/rollout_is_seq_min:0.9912275671958923 - rollout_corr/rollout_is_seq_max_deviation:0.008772432804107666 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2112084375694394) - rollout_corr/training_log_ppl:np.float64(0.18848375794186722) - rollout_corr/kl:np.float64(0.0011117596242904426) - rollout_corr/k3_kl:np.float64(0.001154979970920067) - rollout_corr/rollout_ppl:np.float64(1.2098487238399684) - rollout_corr/rollout_log_ppl:np.float64(0.18737199653696734) - rollout_corr/log_ppl_diff:np.float64(0.0011117614048998803) - rollout_corr/log_ppl_abs_diff:np.float64(0.0024173252459149808) - rollout_corr/log_ppl_diff_max:np.float64(0.012679904699325562) - rollout_corr/log_ppl_diff_min:np.float64(-0.007823467254638672) - rollout_corr/ppl_ratio:np.float64(1.001116826897487) - rollout_corr/chi2_token:np.float64(0.0024014452937990427) - rollout_corr/chi2_seq:np.float64(1.1112628837581724) - actor/pg_loss:np.float64(0.0006953353295102715) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0016907815646618474) - actor/ppo_kl:np.float64(2.1542608287461462e-05) - actor/pg_clipfrac_lower:np.float64(2.1669787656719564e-05) - actor/grad_norm:np.float64(0.7535721808671951) - perf/mfu/actor:np.float64(0.06570637164116254) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(105.5698356628418) - actor/lr:np.float64(7e-07) - training/global_step:8 - training/epoch:0 - critic/score/mean:0.26953125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.26953125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.016741368919610977 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.016741368919610977 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:396.10546875 - response_length/max:2176.0 - response_length/min:85.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:396.10546875 - response_length_non_aborted/max:2176.0 - response_length_non_aborted/min:85.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:322.59375 - prompt_length/max:498.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00013009272515773773 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0220442079007626) - timing_s/agent_loop/generate_sequences/max:np.float64(13.211836582049727) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.8791097558059846) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(13.211836582049727) - timing_s/agent_loop
+step:9 - global_seqlen/min:13543 - global_seqlen/max:24991 - global_seqlen/minmax_diff:11448 - global_seqlen/balanced_min:20013 - global_seqlen/balanced_max:20031 - global_seqlen/mean:20022.5 - actor/entropy:0.17907994985580444 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5248526930809021 - training/rollout_probs_diff_mean:0.005362952593713999 - training/rollout_probs_diff_std:0.015724333003163338 - training/rollout_actor_probs_pearson_corr:0.997033953666687 - rollout_corr/rollout_is_mean:1.0001100301742554 - rollout_corr/rollout_is_ratio_fraction_high:9.402472642250359e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00015279017679858953 - rollout_corr/rollout_is_max:3.2112302780151367 - rollout_corr/rollout_is_min:0.1576482355594635 - rollout_corr/rollout_is_std:0.04585426300764084 - rollout_corr/rollout_is_eff_sample_size:0.9979021543869241 - rollout_corr/rollout_is_seq_mean:1.0000922679901123 - rollout_corr/rollout_is_seq_std:0.002943279454484582 - rollout_corr/rollout_is_seq_max:1.0106842517852783 - rollout_corr/rollout_is_seq_min:0.9920322895050049 - rollout_corr/rollout_is_seq_max_deviation:0.01068425178527832 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2304100869223475) - rollout_corr/training_log_ppl:np.float64(0.20476800417236518) - rollout_corr/kl:np.float64(0.001099050948646152) - rollout_corr/k3_kl:np.float64(0.0012639515396131173) - rollout_corr/rollout_ppl:np.float64(1.2290536230430007) - rollout_corr/rollout_log_ppl:np.float64(0.2036689531232696) - rollout_corr/log_ppl_diff:np.float64(0.0010990510490955785) - rollout_corr/log_ppl_abs_diff:np.float64(0.0026197419938398525) - rollout_corr/log_ppl_diff_max:np.float64(0.01085159182548523) - rollout_corr/log_ppl_diff_min:np.float64(-0.008266806602478027) - rollout_corr/ppl_ratio:np.float64(1.0011050442699343) - rollout_corr/chi2_token:np.float64(0.0029524778947234154) - rollout_corr/chi2_seq:np.float64(1.2083815224468708) - actor/pg_loss:np.float64(0.000570761039853096) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0023600561607963755) - actor/ppo_kl:np.float64(3.2171180877504924e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.816321387887001) - perf/mfu/actor:np.float64(0.0569996191880088) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(105.63365936279297) - actor/lr:np.float64(8e-07) - training/global_step:9 - training/epoch:0 - critic/score/mean:0.3984375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.3984375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.01449743751436472 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.01449743751436472 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:332.359375 - response_length/max:1372.0 - response_length/min:88.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:332.359375 - response_length_non_aborted/max:1372.0 - response_length_non_aborted/min:88.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:293.34375 - prompt_length/max:499.0 - prompt_length/min:181.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0001559928059577942 - timing_s/agent_loop/generate_sequences/min:np.float64(0.979801919311285) - timing_s/agent_loop/generate_sequences/max:np.float64(8.989023100584745) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.3252656050535734) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(8.989023100584745) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timi
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 10}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score of a protein sequence is typically determined through computational methods such as energy minimization, molecular dynamics simulations, or machine learning models trained on known protein structures. However, without access to specific computational tools or databases that can calculate the folding stability score for the given sequence, it is not possible to determine the exact score. The provided options (A: -0.01, B: 1.69, C: 2.49, D: 0.45) are likely hypothetical values, and there is no standard or universally accepted method to calculate this score from the sequence alone. Therefore, the correct answer cannot be determined based on the information provided.
+</reasoning>
+<answer>
+B
+</answer>
+[ground_truth] B
+[score] 1.0
+[acc] 1.0
+[pred] B
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_10
+step:10 - global_seqlen/min:12666 - global_seqlen/max:26901 - global_seqlen/minmax_diff:14235 - global_seqlen/balanced_min:18269 - global_seqlen/balanced_max:18382 - global_seqlen/mean:18293.25 - actor/entropy:0.192046120762825 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.9020472168922424 - training/rollout_probs_diff_mean:0.00564963323995471 - training/rollout_probs_diff_std:0.0170443132519722 - training/rollout_actor_probs_pearson_corr:0.9967436790466309 - rollout_corr/rollout_is_mean:0.9999064207077026 - rollout_corr/rollout_is_ratio_fraction_high:6.567023228853941e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00027581496397033334 - rollout_corr/rollout_is_max:3.955087184906006 - rollout_corr/rollout_is_min:0.025797156617045403 - rollout_corr/rollout_is_std:0.045984067022800446 - rollout_corr/rollout_is_eff_sample_size:0.997889452616177 - rollout_corr/rollout_is_seq_mean:0.9998351335525513 - rollout_corr/rollout_is_seq_std:0.0035011477302759886 - rollout_corr/rollout_is_seq_max:1.0109554529190063 - rollout_corr/rollout_is_seq_min:0.9882628321647644 - rollout_corr/rollout_is_seq_max_deviation:0.011737167835235596 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2451128889806569) - rollout_corr/training_log_ppl:np.float64(0.21647670843231026) - rollout_corr/kl:np.float64(0.00146516180673828) - rollout_corr/k3_kl:np.float64(0.0013313775715459997) - rollout_corr/rollout_ppl:np.float64(1.2432878208346665) - rollout_corr/rollout_log_ppl:np.float64(0.21501154491852503) - rollout_corr/log_ppl_diff:np.float64(0.0014651635137852281) - rollout_corr/log_ppl_abs_diff:np.float64(0.0028982661897316575) - rollout_corr/log_ppl_diff_max:np.float64(0.014495670795440674) - rollout_corr/log_ppl_diff_min:np.float64(-0.009721904993057251) - rollout_corr/ppl_ratio:np.float64(1.0014725218061358) - rollout_corr/chi2_token:np.float64(0.0023953597992658615) - rollout_corr/chi2_seq:np.float64(1.8521191228646785) - actor/pg_loss:np.float64(-0.0003558297175914049) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015348545880442543) - actor/ppo_kl:np.float64(3.0528818129482715e-05) - actor/pg_clipfrac_lower:np.float64(1.3610627320304047e-05) - actor/grad_norm:np.float64(0.7854970991611481) - perf/mfu/actor:np.float64(0.052699151040128595) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(105.79932022094727) - actor/lr:np.float64(9e-07) - val-aux/sciknoweval/reward/mean@16:np.float64(0.32625) - val-aux/sciknoweval/reward/std@16:np.float64(0.26460828298638245) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.44315999999999994) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.23037433915304933) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.21118000000000003) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.21955087215134436) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.3275) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.266025449388232) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.53736) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.16763225376953209) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.12658) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.14989660984549052) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.3408) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.22535669809810538) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.60002) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.10074795072317197) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.07074) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.08389263468998247) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.34962000000000004) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.17799526365500928) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.6341) - val-aux/sciknoweval/reward/best@16/std:np.float64(0.052678976599
+step:11 - global_seqlen/min:16676 - global_seqlen/max:36461 - global_seqlen/minmax_diff:19785 - global_seqlen/balanced_min:22290 - global_seqlen/balanced_max:27044 - global_seqlen/mean:22898.125 - actor/entropy:0.16072751581668854 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.32411471009254456 - training/rollout_probs_diff_mean:0.0046199411153793335 - training/rollout_probs_diff_std:0.014347893185913563 - training/rollout_actor_probs_pearson_corr:0.9974071383476257 - rollout_corr/rollout_is_mean:0.9998364448547363 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:0.00011153358354931697 - rollout_corr/rollout_is_max:1.8010543584823608 - rollout_corr/rollout_is_min:0.18950599431991577 - rollout_corr/rollout_is_std:0.040931250900030136 - rollout_corr/rollout_is_eff_sample_size:0.9983268407518708 - rollout_corr/rollout_is_seq_mean:0.9997417330741882 - rollout_corr/rollout_is_seq_std:0.0031297144014388323 - rollout_corr/rollout_is_seq_max:1.0107784271240234 - rollout_corr/rollout_is_seq_min:0.9905034303665161 - rollout_corr/rollout_is_seq_max_deviation:0.010778427124023438 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2281816471368074) - rollout_corr/training_log_ppl:np.float64(0.2024953516520327) - rollout_corr/kl:np.float64(0.001424252133454651) - rollout_corr/k3_kl:np.float64(0.0011623206726625313) - rollout_corr/rollout_ppl:np.float64(1.226403083652258) - rollout_corr/rollout_log_ppl:np.float64(0.20107109841774218) - rollout_corr/log_ppl_diff:np.float64(0.0014242532342905179) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025687814049888402) - rollout_corr/log_ppl_diff_max:np.float64(0.01213720440864563) - rollout_corr/log_ppl_diff_min:np.float64(-0.009074568748474121) - rollout_corr/ppl_ratio:np.float64(1.0014300209004432) - rollout_corr/chi2_token:np.float64(0.0017990793567150831) - rollout_corr/chi2_seq:np.float64(0.5897399557288736) - actor/pg_loss:np.float64(0.00016360939480364323) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0016948245472576673) - actor/ppo_kl:np.float64(9.670323938681946e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6953910440206528) - perf/mfu/actor:np.float64(0.06338107402915627) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(111.92782211303711) - actor/lr:np.float64(1e-06) - training/global_step:11 - training/epoch:0 - critic/score/mean:0.44140625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.44140625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.03447655215859413 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.03447655215859413 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:385.25390625 - response_length/max:6669.0 - response_length/min:98.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:385.25390625 - response_length_non_aborted/max:6669.0 - response_length_non_aborted/min:98.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:330.3125 - prompt_length/max:500.0 - prompt_length/min:181.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:5.39831817150116e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2415860313922167) - timing_s/agent_loop/generate_sequences/max:np.float64(33.86591459810734) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.7203849007346435) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(33.86591459810734) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - ti
+step:12 - global_seqlen/min:20156 - global_seqlen/max:34994 - global_seqlen/minmax_diff:14838 - global_seqlen/balanced_min:26866 - global_seqlen/balanced_max:32121 - global_seqlen/mean:28280.875 - actor/entropy:0.12687404453754425 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5569577217102051 - training/rollout_probs_diff_mean:0.00370349595323205 - training/rollout_probs_diff_std:0.013641891069710255 - training/rollout_actor_probs_pearson_corr:0.9971849918365479 - rollout_corr/rollout_is_mean:0.999925971031189 - rollout_corr/rollout_is_ratio_fraction_high:2.9888440622016788e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00020174698147457093 - rollout_corr/rollout_is_max:2.8711585998535156 - rollout_corr/rollout_is_min:0.05395948514342308 - rollout_corr/rollout_is_std:0.038417838513851166 - rollout_corr/rollout_is_eff_sample_size:0.9985261260875207 - rollout_corr/rollout_is_seq_mean:0.9998483657836914 - rollout_corr/rollout_is_seq_std:0.002559172920882702 - rollout_corr/rollout_is_seq_max:1.0071240663528442 - rollout_corr/rollout_is_seq_min:0.9910182356834412 - rollout_corr/rollout_is_seq_max_deviation:0.008981764316558838 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2076720832847059) - rollout_corr/training_log_ppl:np.float64(0.1855281984044268) - rollout_corr/kl:np.float64(0.0015311053801809749) - rollout_corr/k3_kl:np.float64(0.0011974603096120973) - rollout_corr/rollout_ppl:np.float64(1.205766570288688) - rollout_corr/rollout_log_ppl:np.float64(0.1839970920154883) - rollout_corr/log_ppl_diff:np.float64(0.0015311063889384968) - rollout_corr/log_ppl_abs_diff:np.float64(0.002373620836806367) - rollout_corr/log_ppl_diff_max:np.float64(0.015954285860061646) - rollout_corr/log_ppl_diff_min:np.float64(-0.005851656198501587) - rollout_corr/ppl_ratio:np.float64(1.0015365811996162) - rollout_corr/chi2_token:np.float64(0.0016909947153180838) - rollout_corr/chi2_seq:np.float64(0.3893565512262285) - actor/pg_loss:np.float64(0.0003301744582131505) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001968728217093485) - actor/ppo_kl:np.float64(0.0002511454556559656) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.8092714995145798) - perf/mfu/actor:np.float64(0.07718086515597372) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(110.70772171020508) - actor/lr:np.float64(1e-06) - training/global_step:12 - training/epoch:0 - critic/score/mean:0.45703125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.45703125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.019058551639318466 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.019058551639318466 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:522.77734375 - response_length/max:8192.0 - response_length/min:106.0 - response_length/clip_ratio:0.0078125 - response_length_non_aborted/mean:522.77734375 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:106.0 - response_length_non_aborted/clip_ratio:0.0078125 - response/aborted_ratio:0.0 - prompt_length/mean:361.0 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.196065366268158e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2030387092381716) - timing_s/agent_loop/generate_sequences/max:np.float64(49.85608519241214) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.7326578640859225) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.85608519241214) - timing_s/agent_loop/slowest/tool_ca
+step:13 - global_seqlen/min:14490 - global_seqlen/max:27440 - global_seqlen/minmax_diff:12950 - global_seqlen/balanced_min:21991 - global_seqlen/balanced_max:22034 - global_seqlen/mean:22017.0 - actor/entropy:0.1795348972082138 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3799695372581482 - training/rollout_probs_diff_mean:0.004953271709382534 - training/rollout_probs_diff_std:0.014711440540850163 - training/rollout_actor_probs_pearson_corr:0.9974781274795532 - rollout_corr/rollout_is_mean:1.0001226663589478 - rollout_corr/rollout_is_ratio_fraction_high:4.09433341701515e-05 - rollout_corr/rollout_is_ratio_fraction_low:7.165083661675453e-05 - rollout_corr/rollout_is_max:2.9943156242370605 - rollout_corr/rollout_is_min:0.36588054895401 - rollout_corr/rollout_is_std:0.04241306707262993 - rollout_corr/rollout_is_eff_sample_size:0.9982047182816606 - rollout_corr/rollout_is_seq_mean:1.0001139640808105 - rollout_corr/rollout_is_seq_std:0.003129767719656229 - rollout_corr/rollout_is_seq_max:1.0108983516693115 - rollout_corr/rollout_is_seq_min:0.9917095303535461 - rollout_corr/rollout_is_seq_max_deviation:0.010898351669311523 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2477836473844945) - rollout_corr/training_log_ppl:np.float64(0.2179029462276958) - rollout_corr/kl:np.float64(0.0012722289181801294) - rollout_corr/k3_kl:np.float64(0.0011847673267197933) - rollout_corr/rollout_ppl:np.float64(1.2461294089443982) - rollout_corr/rollout_log_ppl:np.float64(0.2166307152074296) - rollout_corr/log_ppl_diff:np.float64(0.0012722310202661902) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027390209725126624) - rollout_corr/log_ppl_diff_max:np.float64(0.01292330026626587) - rollout_corr/log_ppl_diff_min:np.float64(-0.009444236755371094) - rollout_corr/ppl_ratio:np.float64(1.0012791764456779) - rollout_corr/chi2_token:np.float64(0.002235463121905923) - rollout_corr/chi2_seq:np.float64(1.5766265538986772) - actor/pg_loss:np.float64(-0.0002433034824207425) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002133346634764166) - actor/ppo_kl:np.float64(0.00032201276727761297) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.8120427280664444) - perf/mfu/actor:np.float64(0.06433296210788333) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(110.85671615600586) - actor/lr:np.float64(1e-06) - training/global_step:13 - training/epoch:0 - critic/score/mean:0.35546875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.35546875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.007056328002363443 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.007056328002363443 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:381.625 - response_length/max:1800.0 - response_length/min:99.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:381.625 - response_length_non_aborted/max:1800.0 - response_length_non_aborted/min:99.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:306.40625 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00013713166117668152 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2604616954922676) - timing_s/agent_loop/generate_sequences/max:np.float64(11.544187435880303) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.8230461675921106) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(11.544187435880303) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - t
+step:14 - global_seqlen/min:12976 - global_seqlen/max:34060 - global_seqlen/minmax_diff:21084 - global_seqlen/balanced_min:22596 - global_seqlen/balanced_max:22654 - global_seqlen/mean:22633.375 - actor/entropy:0.16276216506958008 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.31926417350769043 - training/rollout_probs_diff_mean:0.004311886616051197 - training/rollout_probs_diff_std:0.013316148892045021 - training/rollout_actor_probs_pearson_corr:0.9977449178695679 - rollout_corr/rollout_is_mean:0.9999713897705078 - rollout_corr/rollout_is_ratio_fraction_high:3.8242746086325496e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.692480383208022e-05 - rollout_corr/rollout_is_max:2.2574832439422607 - rollout_corr/rollout_is_min:0.28150883316993713 - rollout_corr/rollout_is_std:0.040303174406290054 - rollout_corr/rollout_is_eff_sample_size:0.9983781696179648 - rollout_corr/rollout_is_seq_mean:0.9999383687973022 - rollout_corr/rollout_is_seq_std:0.0024390218313783407 - rollout_corr/rollout_is_seq_max:1.008014440536499 - rollout_corr/rollout_is_seq_min:0.9899135828018188 - rollout_corr/rollout_is_seq_max_deviation:0.010086417198181152 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2327516474761069) - rollout_corr/training_log_ppl:np.float64(0.20615991155500524) - rollout_corr/kl:np.float64(0.0009014457496316908) - rollout_corr/k3_kl:np.float64(0.0010392485548891273) - rollout_corr/rollout_ppl:np.float64(1.231638588476926) - rollout_corr/rollout_log_ppl:np.float64(0.2052584641496651) - rollout_corr/log_ppl_diff:np.float64(0.00090144740534015) - rollout_corr/log_ppl_abs_diff:np.float64(0.00216034529148601) - rollout_corr/log_ppl_diff_max:np.float64(0.010936662554740906) - rollout_corr/log_ppl_diff_min:np.float64(-0.008887723088264465) - rollout_corr/ppl_ratio:np.float64(1.0009056760463864) - rollout_corr/chi2_token:np.float64(0.002428637119010091) - rollout_corr/chi2_seq:np.float64(1.1035329343285412) - actor/pg_loss:np.float64(0.0004473176086321473) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011060538008678122) - actor/ppo_kl:np.float64(-0.0001549605085848782) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.575629249215126) - perf/mfu/actor:np.float64(0.0652755181152363) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(109.99267196655273) - actor/lr:np.float64(1e-06) - training/global_step:14 - training/epoch:0 - critic/score/mean:0.26953125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.26953125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.010551412589848042 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.010551412589848042 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:408.57421875 - response_length/max:2188.0 - response_length/min:93.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:408.57421875 - response_length_non_aborted/max:2188.0 - response_length_non_aborted/min:93.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:298.71875 - prompt_length/max:498.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.820424020290375e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0610912814736366) - timing_s/agent_loop/generate_sequences/max:np.float64(13.642793396487832) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.8376170305346022) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(13.642793396487832) - timing_s/agent_loop/slowest/tool_calls:np.fl
+step:15 - global_seqlen/min:15918 - global_seqlen/max:28170 - global_seqlen/minmax_diff:12252 - global_seqlen/balanced_min:19653 - global_seqlen/balanced_max:19679 - global_seqlen/mean:19673.25 - actor/entropy:0.19017662107944489 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.6307171583175659 - training/rollout_probs_diff_mean:0.005055586341768503 - training/rollout_probs_diff_std:0.014726489782333374 - training/rollout_actor_probs_pearson_corr:0.9975889921188354 - rollout_corr/rollout_is_mean:1.0000306367874146 - rollout_corr/rollout_is_ratio_fraction_high:2.42359619733179e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00013329778448678553 - rollout_corr/rollout_is_max:2.5952959060668945 - rollout_corr/rollout_is_min:0.22091145813465118 - rollout_corr/rollout_is_std:0.04240182042121887 - rollout_corr/rollout_is_eff_sample_size:0.9982053121902402 - rollout_corr/rollout_is_seq_mean:1.0000522136688232 - rollout_corr/rollout_is_seq_std:0.0030719421338289976 - rollout_corr/rollout_is_seq_max:1.0085277557373047 - rollout_corr/rollout_is_seq_min:0.9906306266784668 - rollout_corr/rollout_is_seq_max_deviation:0.009369373321533203 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2559455754235387) - rollout_corr/training_log_ppl:np.float64(0.22441870825423393) - rollout_corr/kl:np.float64(0.001127970071593154) - rollout_corr/k3_kl:np.float64(0.001125821751145395) - rollout_corr/rollout_ppl:np.float64(1.254524188581854) - rollout_corr/rollout_log_ppl:np.float64(0.22329073723813053) - rollout_corr/log_ppl_diff:np.float64(0.0011279710161034018) - rollout_corr/log_ppl_abs_diff:np.float64(0.0024534623080398887) - rollout_corr/log_ppl_diff_max:np.float64(0.014268159866333008) - rollout_corr/log_ppl_diff_min:np.float64(-0.008966684341430664) - rollout_corr/ppl_ratio:np.float64(1.00113360513933) - rollout_corr/chi2_token:np.float64(0.0023321015760302544) - rollout_corr/chi2_seq:np.float64(0.6067947293631732) - actor/pg_loss:np.float64(-0.00015570060350000858) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0014229919640911248) - actor/ppo_kl:np.float64(0.00016318686129324078) - actor/pg_clipfrac_lower:np.float64(5.865240382263437e-06) - actor/grad_norm:np.float64(0.6899215131998062) - perf/mfu/actor:np.float64(0.05693643757760258) - perf/max_memory_allocated_gb:np.float64(123.58113050460815) - perf/max_memory_reserved_gb:np.float64(129.3828125) - perf/cpu_memory_used_gb:np.float64(110.72046184539795) - actor/lr:np.float64(1e-06) - training/global_step:15 - training/epoch:1 - critic/score/mean:0.34765625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.34765625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.012832941487431526 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.012832941487431526 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:322.3515625 - response_length/max:1553.0 - response_length/min:91.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:322.3515625 - response_length_non_aborted/max:1553.0 - response_length_non_aborted/min:91.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:292.4375 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0003503747284412384 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2059585116803646) - timing_s/agent_loop/generate_sequences/max:np.float64(9.883644664660096) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.228372771802242) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.883644664660096) - timing_s/agent_loop/slowest/t
+step:16 - global_seqlen/min:17621 - global_seqlen/max:49777 - global_seqlen/minmax_diff:32156 - global_seqlen/balanced_min:23638 - global_seqlen/balanced_max:29431 - global_seqlen/mean:27074.0 - actor/entropy:0.12440122663974762 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7334084510803223 - training/rollout_probs_diff_mean:0.003286235500127077 - training/rollout_probs_diff_std:0.012201991863548756 - training/rollout_actor_probs_pearson_corr:0.997710645198822 - rollout_corr/rollout_is_mean:0.999930739402771 - rollout_corr/rollout_is_ratio_fraction_high:7.548765097453725e-06 - rollout_corr/rollout_is_ratio_fraction_low:6.793888314859942e-05 - rollout_corr/rollout_is_max:2.695401906967163 - rollout_corr/rollout_is_min:0.02908293716609478 - rollout_corr/rollout_is_std:0.034535329788923264 - rollout_corr/rollout_is_eff_sample_size:0.9988087318703696 - rollout_corr/rollout_is_seq_mean:0.9999657869338989 - rollout_corr/rollout_is_seq_std:0.002763616619631648 - rollout_corr/rollout_is_seq_max:1.0094408988952637 - rollout_corr/rollout_is_seq_min:0.9909818768501282 - rollout_corr/rollout_is_seq_max_deviation:0.009440898895263672 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2334859780967236) - rollout_corr/training_log_ppl:np.float64(0.20614377667880035) - rollout_corr/kl:np.float64(0.0010040062433311903) - rollout_corr/k3_kl:np.float64(0.0012267602566424785) - rollout_corr/rollout_ppl:np.float64(1.2322044824250042) - rollout_corr/rollout_log_ppl:np.float64(0.20513977024347696) - rollout_corr/log_ppl_diff:np.float64(0.0010040064353233902) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025442858877795516) - rollout_corr/log_ppl_diff_max:np.float64(0.014998257160186768) - rollout_corr/log_ppl_diff_min:np.float64(-0.008154168725013733) - rollout_corr/ppl_ratio:np.float64(1.0010099613573402) - rollout_corr/chi2_token:np.float64(0.002799068344756961) - rollout_corr/chi2_seq:np.float64(0.6977890518028289) - actor/pg_loss:np.float64(0.0005094839725643396) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0022031578653241013) - actor/ppo_kl:np.float64(-6.424526176118661e-05) - actor/pg_clipfrac_lower:np.float64(1.4852661479380913e-05) - actor/grad_norm:np.float64(0.759627491235733) - perf/mfu/actor:np.float64(0.07361790070915576) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(110.84224700927734) - actor/lr:np.float64(1e-06) - training/global_step:16 - training/epoch:1 - critic/score/mean:0.546875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.546875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0865541398525238 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0865541398525238 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:517.46875 - response_length/max:8192.0 - response_length/min:85.0 - response_length/clip_ratio:0.015625 - response_length_non_aborted/mean:517.46875 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:85.0 - response_length_non_aborted/clip_ratio:0.015625 - response/aborted_ratio:0.0 - prompt_length/mean:328.59375 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.716115891933441e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.6790603306144476) - timing_s/agent_loop/generate_sequences/max:np.float64(51.488350415602326) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.263302836421644) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(51.488350415602326) - timing_s/agent_loop/slowest
+step:17 - global_seqlen/min:11966 - global_seqlen/max:39775 - global_seqlen/minmax_diff:27809 - global_seqlen/balanced_min:20593 - global_seqlen/balanced_max:26443 - global_seqlen/mean:22197.625 - actor/entropy:0.15671347081661224 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.34504491090774536 - training/rollout_probs_diff_mean:0.003963190596550703 - training/rollout_probs_diff_std:0.012594124302268028 - training/rollout_actor_probs_pearson_corr:0.9979807138442993 - rollout_corr/rollout_is_mean:0.9999871253967285 - rollout_corr/rollout_is_ratio_fraction_high:9.640877578931395e-06 - rollout_corr/rollout_is_ratio_fraction_low:9.640877578931395e-06 - rollout_corr/rollout_is_max:2.198077917098999 - rollout_corr/rollout_is_min:0.4044226408004761 - rollout_corr/rollout_is_std:0.03752783685922623 - rollout_corr/rollout_is_eff_sample_size:0.9985936420813145 - rollout_corr/rollout_is_seq_mean:0.9999175071716309 - rollout_corr/rollout_is_seq_std:0.0031496852170675993 - rollout_corr/rollout_is_seq_max:1.0094223022460938 - rollout_corr/rollout_is_seq_min:0.9900813102722168 - rollout_corr/rollout_is_seq_max_deviation:0.009918689727783203 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2560751209966838) - rollout_corr/training_log_ppl:np.float64(0.22452797560981708) - rollout_corr/kl:np.float64(0.0013395728782059635) - rollout_corr/k3_kl:np.float64(0.0012000854812015405) - rollout_corr/rollout_ppl:np.float64(1.2543696821667254) - rollout_corr/rollout_log_ppl:np.float64(0.2231884019674908) - rollout_corr/log_ppl_diff:np.float64(0.001339573642326286) - rollout_corr/log_ppl_abs_diff:np.float64(0.002711944489419693) - rollout_corr/log_ppl_diff_max:np.float64(0.01098942756652832) - rollout_corr/log_ppl_diff_min:np.float64(-0.0078660249710083) - rollout_corr/ppl_ratio:np.float64(1.0013459962792695) - rollout_corr/chi2_token:np.float64(0.0021593261044472456) - rollout_corr/chi2_seq:np.float64(0.8300942496862262) - actor/pg_loss:np.float64(0.0006044314941391349) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015271970464709739) - actor/ppo_kl:np.float64(0.00025955110856701324) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.73015196621418) - perf/mfu/actor:np.float64(0.06131922719985038) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(110.87406206130981) - actor/lr:np.float64(1e-06) - training/global_step:17 - training/epoch:1 - critic/score/mean:0.28515625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.28515625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.06812485307455063 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.06812485307455063 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:405.17578125 - response_length/max:8192.0 - response_length/min:85.0 - response_length/clip_ratio:0.0078125 - response_length_non_aborted/mean:405.17578125 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:85.0 - response_length_non_aborted/clip_ratio:0.0078125 - response/aborted_ratio:0.0 - prompt_length/mean:288.5 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00014323554933071136 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9844553098082542) - timing_s/agent_loop/generate_sequences/max:np.float64(49.343952156603336) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.660478450343362) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.343952156603336) - timing_s/agent_loop/slowest/tool_call
+step:18 - global_seqlen/min:11273 - global_seqlen/max:36015 - global_seqlen/minmax_diff:24742 - global_seqlen/balanced_min:21872 - global_seqlen/balanced_max:27887 - global_seqlen/mean:22719.125 - actor/entropy:0.1572161316871643 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7516714334487915 - training/rollout_probs_diff_mean:0.003971869591623545 - training/rollout_probs_diff_std:0.01294691488146782 - training/rollout_actor_probs_pearson_corr:0.9978299140930176 - rollout_corr/rollout_is_mean:1.0000897645950317 - rollout_corr/rollout_is_ratio_fraction_high:3.9406924770446494e-05 - rollout_corr/rollout_is_ratio_fraction_low:1.9703462385223247e-05 - rollout_corr/rollout_is_max:2.524366617202759 - rollout_corr/rollout_is_min:0.09788884967565536 - rollout_corr/rollout_is_std:0.037467435002326965 - rollout_corr/rollout_is_eff_sample_size:0.998598397072895 - rollout_corr/rollout_is_seq_mean:1.0000863075256348 - rollout_corr/rollout_is_seq_std:0.0033069117926061153 - rollout_corr/rollout_is_seq_max:1.0111510753631592 - rollout_corr/rollout_is_seq_min:0.9837021827697754 - rollout_corr/rollout_is_seq_max_deviation:0.01629781723022461 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2480866289697587) - rollout_corr/training_log_ppl:np.float64(0.2177488712950435) - rollout_corr/kl:np.float64(0.0012160134455356797) - rollout_corr/k3_kl:np.float64(0.0012282074120264497) - rollout_corr/rollout_ppl:np.float64(1.2464866428636014) - rollout_corr/rollout_log_ppl:np.float64(0.21653285443244386) - rollout_corr/log_ppl_diff:np.float64(0.0012160168625996448) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027001788184861653) - rollout_corr/log_ppl_diff_max:np.float64(0.01847335696220398) - rollout_corr/log_ppl_diff_min:np.float64(-0.008770108222961426) - rollout_corr/ppl_ratio:np.float64(1.0012231352739036) - rollout_corr/chi2_token:np.float64(0.0024735613260418177) - rollout_corr/chi2_seq:np.float64(1.276715693064034) - actor/pg_loss:np.float64(0.0005491708870977163) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0019209535612390027) - actor/ppo_kl:np.float64(0.00030758063345004416) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7345165759325027) - perf/mfu/actor:np.float64(0.06209275037892638) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(110.8076171875) - actor/lr:np.float64(1e-06) - training/global_step:18 - training/epoch:1 - critic/score/mean:0.3359375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.3359375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.02325008623301983 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.02325008623301983 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:396.50390625 - response_length/max:8192.0 - response_length/min:92.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:396.50390625 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:92.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:313.46875 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0001363120973110199 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0805652625858784) - timing_s/agent_loop/generate_sequences/max:np.float64(49.01962508261204) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.7108128744512214) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.01962508261204) - timing_s/agent_loop/slowest/tool_ca
+step:19 - global_seqlen/min:14639 - global_seqlen/max:26849 - global_seqlen/minmax_diff:12210 - global_seqlen/balanced_min:21052 - global_seqlen/balanced_max:21251 - global_seqlen/mean:21094.25 - actor/entropy:0.18874603509902954 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7417160868644714 - training/rollout_probs_diff_mean:0.004822679795324802 - training/rollout_probs_diff_std:0.014397743158042431 - training/rollout_actor_probs_pearson_corr:0.9977278113365173 - rollout_corr/rollout_is_mean:0.9999311566352844 - rollout_corr/rollout_is_ratio_fraction_high:1.0811979336722288e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00014055574138183147 - rollout_corr/rollout_is_max:2.0913357734680176 - rollout_corr/rollout_is_min:0.040095582604408264 - rollout_corr/rollout_is_std:0.0410170778632164 - rollout_corr/rollout_is_eff_sample_size:0.998320068587282 - rollout_corr/rollout_is_seq_mean:1.0000311136245728 - rollout_corr/rollout_is_seq_std:0.0029985024593770504 - rollout_corr/rollout_is_seq_max:1.010836124420166 - rollout_corr/rollout_is_seq_min:0.987970232963562 - rollout_corr/rollout_is_seq_max_deviation:0.012029767036437988 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.255211434327066) - rollout_corr/training_log_ppl:np.float64(0.22383148300286848) - rollout_corr/kl:np.float64(0.0012053162868781442) - rollout_corr/k3_kl:np.float64(0.0011591173564511337) - rollout_corr/rollout_ppl:np.float64(1.253677450120449) - rollout_corr/rollout_log_ppl:np.float64(0.2226261663454352) - rollout_corr/log_ppl_diff:np.float64(0.0012053166574332863) - rollout_corr/log_ppl_abs_diff:np.float64(0.002684240898815915) - rollout_corr/log_ppl_diff_max:np.float64(0.01323661208152771) - rollout_corr/log_ppl_diff_min:np.float64(-0.010094761848449707) - rollout_corr/ppl_ratio:np.float64(1.0012121815234423) - rollout_corr/chi2_token:np.float64(0.0021450589410960674) - rollout_corr/chi2_seq:np.float64(1.123602865729481) - actor/pg_loss:np.float64(0.00028023216873407364) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0009424093339021056) - actor/ppo_kl:np.float64(0.0002521510247390779) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5630562603473663) - perf/mfu/actor:np.float64(0.06105586170219322) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(110.85847091674805) - actor/lr:np.float64(1e-06) - training/global_step:19 - training/epoch:1 - critic/score/mean:0.40625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.40625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.012870309874415398 - critic/advantages/max:0.75 - critic/advantages/min:-0.875 - critic/returns/mean:-0.012870309874415398 - critic/returns/max:0.75 - critic/returns/min:-0.875 - response_length/mean:361.2890625 - response_length/max:1883.0 - response_length/min:86.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:361.2890625 - response_length_non_aborted/max:1883.0 - response_length_non_aborted/min:86.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:297.90625 - prompt_length/max:498.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:6.548874080181122e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0813855342566967) - timing_s/agent_loop/generate_sequences/max:np.float64(11.688677107915282) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.5753794663687586) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(11.688677107915282) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 20}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+To determine the folding stability score of the given protein sequence, one would typically use a computational method or software that calculates the stability based on the amino acid sequence. This score is often derived from energy calculations that consider factors such as hydrogen bonding, hydrophobic interactions, and van der Waals forces. However, without access to specific computational tools or databases that provide such scores for the given sequence, it is not possible to calculate the exact folding stability score. Therefore, the correct answer cannot be determined from the information provided.
+</reasoning>
+<answer>
+D
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] D
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_20
+step:20 - global_seqlen/min:12298 - global_seqlen/max:30479 - global_seqlen/minmax_diff:18181 - global_seqlen/balanced_min:21810 - global_seqlen/balanced_max:21871 - global_seqlen/mean:21839.375 - actor/entropy:0.18792696297168732 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.6172393560409546 - training/rollout_probs_diff_mean:0.004675295669585466 - training/rollout_probs_diff_std:0.013787684962153435 - training/rollout_actor_probs_pearson_corr:0.9978795051574707 - rollout_corr/rollout_is_mean:1.0000436305999756 - rollout_corr/rollout_is_ratio_fraction_high:5.289270120556466e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00010578540241112933 - rollout_corr/rollout_is_max:2.6348955631256104 - rollout_corr/rollout_is_min:0.11010421812534332 - rollout_corr/rollout_is_std:0.040804363787174225 - rollout_corr/rollout_is_eff_sample_size:0.998337890270291 - rollout_corr/rollout_is_seq_mean:1.000036597251892 - rollout_corr/rollout_is_seq_std:0.0025871675461530685 - rollout_corr/rollout_is_seq_max:1.009564757347107 - rollout_corr/rollout_is_seq_min:0.9896543622016907 - rollout_corr/rollout_is_seq_max_deviation:0.010345637798309326 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2705717207863927) - rollout_corr/training_log_ppl:np.float64(0.2352731467835838) - rollout_corr/kl:np.float64(0.00124775865324378) - rollout_corr/k3_kl:np.float64(0.001086297427093541) - rollout_corr/rollout_ppl:np.float64(1.2689147526398301) - rollout_corr/rollout_log_ppl:np.float64(0.23402538590016775) - rollout_corr/log_ppl_diff:np.float64(0.0012477608834160492) - rollout_corr/log_ppl_abs_diff:np.float64(0.0022727364994352683) - rollout_corr/log_ppl_diff_max:np.float64(0.012642994523048401) - rollout_corr/log_ppl_diff_min:np.float64(-0.007874488830566406) - rollout_corr/ppl_ratio:np.float64(1.0012527105864137) - rollout_corr/chi2_token:np.float64(0.0018538862932473421) - rollout_corr/chi2_seq:np.float64(0.5317726316861808) - actor/pg_loss:np.float64(-4.45549376308918e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001396393277673269) - actor/ppo_kl:np.float64(0.0002731104437279086) - actor/pg_clipfrac_lower:np.float64(7.614522473886609e-06) - actor/grad_norm:np.float64(0.5186081603169441) - perf/mfu/actor:np.float64(0.06376455117784033) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(110.90739440917969) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.3575) - val-aux/sciknoweval/reward/std@16:np.float64(0.2845272325003602) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.4821000000000001) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.24641293690229316) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.23617999999999997) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.23810031156810768) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.35924000000000006) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.2844295099662947) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.57886) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.17803900782144286) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.13665999999999998) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.16515577551517271) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.38042000000000004) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.23790217711971715) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.64558) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.11228107051862579) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.07342) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.08529636863102308) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.39696) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.18341197265688536) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.6885199999999999) - val-aux/sciknoweval/reward/bes
+step:21 - global_seqlen/min:13059 - global_seqlen/max:30220 - global_seqlen/minmax_diff:17161 - global_seqlen/balanced_min:20197 - global_seqlen/balanced_max:20234 - global_seqlen/mean:20222.625 - actor/entropy:0.2039741724729538 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2915150225162506 - training/rollout_probs_diff_mean:0.005005266051739454 - training/rollout_probs_diff_std:0.013932553119957447 - training/rollout_actor_probs_pearson_corr:0.9979515671730042 - rollout_corr/rollout_is_mean:0.9997883439064026 - rollout_corr/rollout_is_ratio_fraction_high:1.1129289305245038e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0 - rollout_corr/rollout_is_max:2.3997433185577393 - rollout_corr/rollout_is_min:0.5131387114524841 - rollout_corr/rollout_is_std:0.04085326939821243 - rollout_corr/rollout_is_eff_sample_size:0.9983332565714974 - rollout_corr/rollout_is_seq_mean:0.9999094605445862 - rollout_corr/rollout_is_seq_std:0.0028472074773162603 - rollout_corr/rollout_is_seq_max:1.011277198791504 - rollout_corr/rollout_is_seq_min:0.9916235208511353 - rollout_corr/rollout_is_seq_max_deviation:0.011277198791503906 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2804982154630125) - rollout_corr/training_log_ppl:np.float64(0.2435035327434889) - rollout_corr/kl:np.float64(0.0010719042137878887) - rollout_corr/k3_kl:np.float64(0.0011022668704754324) - rollout_corr/rollout_ppl:np.float64(1.2791178887709975) - rollout_corr/rollout_log_ppl:np.float64(0.24243162775383098) - rollout_corr/log_ppl_diff:np.float64(0.0010719049896579236) - rollout_corr/log_ppl_abs_diff:np.float64(0.002581465640105307) - rollout_corr/log_ppl_diff_max:np.float64(0.009952932596206665) - rollout_corr/log_ppl_diff_min:np.float64(-0.010108351707458496) - rollout_corr/ppl_ratio:np.float64(1.0010773693211377) - rollout_corr/chi2_token:np.float64(0.0022938339971005917) - rollout_corr/chi2_seq:np.float64(1.2692464508581907) - actor/pg_loss:np.float64(0.0004962929524481297) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0018514558880724508) - actor/ppo_kl:np.float64(6.505578917792576e-07) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7440718114376068) - perf/mfu/actor:np.float64(0.05869452164939222) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.24159622192383) - actor/lr:np.float64(1e-06) - training/global_step:21 - training/epoch:1 - critic/score/mean:0.40625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.40625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0015914883697405457 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0015914883697405457 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:350.98828125 - response_length/max:1700.0 - response_length/min:102.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:350.98828125 - response_length_non_aborted/max:1700.0 - response_length_non_aborted/min:102.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:280.96875 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:5.297362804412842e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.264117207378149) - timing_s/agent_loop/generate_sequences/max:np.float64(10.401357980445027) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.5284495416854043) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(10.401357980445027) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - tim
+step:22 - global_seqlen/min:18027 - global_seqlen/max:36166 - global_seqlen/minmax_diff:18139 - global_seqlen/balanced_min:25060 - global_seqlen/balanced_max:28210 - global_seqlen/mean:25542.25 - actor/entropy:0.17578984797000885 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.4386941194534302 - training/rollout_probs_diff_mean:0.004188444931060076 - training/rollout_probs_diff_std:0.012796149589121342 - training/rollout_actor_probs_pearson_corr:0.9980850219726562 - rollout_corr/rollout_is_mean:1.000032901763916 - rollout_corr/rollout_is_ratio_fraction_high:1.730672738631256e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00010384036431787536 - rollout_corr/rollout_is_max:2.1636812686920166 - rollout_corr/rollout_is_min:0.21608301997184753 - rollout_corr/rollout_is_std:0.038447305560112 - rollout_corr/rollout_is_eff_sample_size:0.9985239866454121 - rollout_corr/rollout_is_seq_mean:0.9999626874923706 - rollout_corr/rollout_is_seq_std:0.0027287364937365055 - rollout_corr/rollout_is_seq_max:1.0086798667907715 - rollout_corr/rollout_is_seq_min:0.990822970867157 - rollout_corr/rollout_is_seq_max_deviation:0.009177029132843018 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.265514174476266) - rollout_corr/training_log_ppl:np.float64(0.23049156166962348) - rollout_corr/kl:np.float64(0.001140852084887456) - rollout_corr/k3_kl:np.float64(0.0010895646327071518) - rollout_corr/rollout_ppl:np.float64(1.2640215558931231) - rollout_corr/rollout_log_ppl:np.float64(0.22935070820676629) - rollout_corr/log_ppl_diff:np.float64(0.0011408534628571942) - rollout_corr/log_ppl_abs_diff:np.float64(0.002326310393982567) - rollout_corr/log_ppl_diff_max:np.float64(0.016114696860313416) - rollout_corr/log_ppl_diff_min:np.float64(-0.005415067076683044) - rollout_corr/ppl_ratio:np.float64(1.0011460261885077) - rollout_corr/chi2_token:np.float64(0.0020373596344143152) - rollout_corr/chi2_seq:np.float64(1.662300607888028) - actor/pg_loss:np.float64(-4.751852247864008e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0013797034152958076) - actor/ppo_kl:np.float64(0.0001431179896584922) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6352894306182861) - perf/mfu/actor:np.float64(0.07072783775794689) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(107.67041778564453) - actor/lr:np.float64(1e-06) - training/global_step:22 - training/epoch:1 - critic/score/mean:0.49609375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.49609375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.024204539135098457 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.024204539135098457 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:451.4140625 - response_length/max:5103.0 - response_length/min:103.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:451.4140625 - response_length_non_aborted/max:5103.0 - response_length_non_aborted/min:103.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:346.78125 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:6.669573485851288e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2452805154025555) - timing_s/agent_loop/generate_sequences/max:np.float64(27.396138545125723) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.402181767611182) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(27.396138545125723) - timing_s/agent_loop/slowest/tool_calls:np.flo
+step:23 - global_seqlen/min:18253 - global_seqlen/max:33113 - global_seqlen/minmax_diff:14860 - global_seqlen/balanced_min:23569 - global_seqlen/balanced_max:24069 - global_seqlen/mean:23647.375 - actor/entropy:0.20427009463310242 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.403209388256073 - training/rollout_probs_diff_mean:0.004876875318586826 - training/rollout_probs_diff_std:0.013606295920908451 - training/rollout_actor_probs_pearson_corr:0.9980311393737793 - rollout_corr/rollout_is_mean:0.9999285936355591 - rollout_corr/rollout_is_ratio_fraction_high:3.8674621464451775e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.76805866532959e-05 - rollout_corr/rollout_is_max:3.7063148021698 - rollout_corr/rollout_is_min:0.2912478744983673 - rollout_corr/rollout_is_std:0.0409952737390995 - rollout_corr/rollout_is_eff_sample_size:0.9983218507269501 - rollout_corr/rollout_is_seq_mean:0.9999192357063293 - rollout_corr/rollout_is_seq_std:0.002459433162584901 - rollout_corr/rollout_is_seq_max:1.008528470993042 - rollout_corr/rollout_is_seq_min:0.9915392994880676 - rollout_corr/rollout_is_seq_max_deviation:0.008528470993041992 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.277975270524621) - rollout_corr/training_log_ppl:np.float64(0.24159730068640783) - rollout_corr/kl:np.float64(0.0014997820293984887) - rollout_corr/k3_kl:np.float64(0.0011694539433619866) - rollout_corr/rollout_ppl:np.float64(1.2760190423578024) - rollout_corr/rollout_log_ppl:np.float64(0.2400975169875892) - rollout_corr/log_ppl_diff:np.float64(0.0014997836988186464) - rollout_corr/log_ppl_abs_diff:np.float64(0.002539189110393636) - rollout_corr/log_ppl_diff_max:np.float64(0.011486321687698364) - rollout_corr/log_ppl_diff_min:np.float64(-0.00882810354232788) - rollout_corr/ppl_ratio:np.float64(1.001505514839664) - rollout_corr/chi2_token:np.float64(0.0016600224189460278) - rollout_corr/chi2_seq:np.float64(0.7354465429671109) - actor/pg_loss:np.float64(0.00017330946866422892) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015464231032638054) - actor/ppo_kl:np.float64(0.0004333707376389384) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6127719506621361) - perf/mfu/actor:np.float64(0.06741122513113522) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(106.24171829223633) - actor/lr:np.float64(1e-06) - training/global_step:23 - training/epoch:1 - critic/score/mean:0.25390625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.25390625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.00273381220176816 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.00273381220176816 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:404.01171875 - response_length/max:2216.0 - response_length/min:139.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:404.01171875 - response_length_non_aborted/max:2216.0 - response_length_non_aborted/min:139.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:334.96875 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:6.080232560634613e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.536420190706849) - timing_s/agent_loop/generate_sequences/max:np.float64(13.316242065280676) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.049403940458433) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(13.316242065280676) - timing_s/agent_loop/slowest/tool_calls:np.float64(0
+step:24 - global_seqlen/min:14555 - global_seqlen/max:31176 - global_seqlen/minmax_diff:16621 - global_seqlen/balanced_min:20123 - global_seqlen/balanced_max:20634 - global_seqlen/mean:20209.125 - actor/entropy:0.21117329597473145 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.29998478293418884 - training/rollout_probs_diff_mean:0.005020960234105587 - training/rollout_probs_diff_std:0.013722977600991726 - training/rollout_actor_probs_pearson_corr:0.9980407357215881 - rollout_corr/rollout_is_mean:0.9998765587806702 - rollout_corr/rollout_is_ratio_fraction_high:3.355366789037362e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.5922781029948965e-05 - rollout_corr/rollout_is_max:2.400970697402954 - rollout_corr/rollout_is_min:0.23983050882816315 - rollout_corr/rollout_is_std:0.04097927734255791 - rollout_corr/rollout_is_eff_sample_size:0.9983230388235971 - rollout_corr/rollout_is_seq_mean:0.9998420476913452 - rollout_corr/rollout_is_seq_std:0.002557374769821763 - rollout_corr/rollout_is_seq_max:1.0084576606750488 - rollout_corr/rollout_is_seq_min:0.9929150342941284 - rollout_corr/rollout_is_seq_max_deviation:0.008457660675048828 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2810057001188397) - rollout_corr/training_log_ppl:np.float64(0.2444376761704916) - rollout_corr/kl:np.float64(0.0012209779730945058) - rollout_corr/k3_kl:np.float64(0.0010693179642657924) - rollout_corr/rollout_ppl:np.float64(1.2793983216397464) - rollout_corr/rollout_log_ppl:np.float64(0.24321669599157758) - rollout_corr/log_ppl_diff:np.float64(0.0012209801789140329) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025764188467292115) - rollout_corr/log_ppl_diff_max:np.float64(0.011098459362983704) - rollout_corr/log_ppl_diff_min:np.float64(-0.008527815341949463) - rollout_corr/ppl_ratio:np.float64(1.0012267434503883) - rollout_corr/chi2_token:np.float64(0.0018536767456680536) - rollout_corr/chi2_seq:np.float64(1.0982552233617753) - actor/pg_loss:np.float64(0.0001084102550521493) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001204582532409404) - actor/ppo_kl:np.float64(9.153693397934148e-05) - actor/pg_clipfrac_lower:np.float64(1.0643732821336016e-05) - actor/grad_norm:np.float64(0.641492635011673) - perf/mfu/actor:np.float64(0.056898364927665215) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.3012466430664) - actor/lr:np.float64(1e-06) - training/global_step:24 - training/epoch:1 - critic/score/mean:0.36328125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.36328125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.010942690074443817 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.010942690074443817 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:349.25390625 - response_length/max:2150.0 - response_length/min:86.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:349.25390625 - response_length_non_aborted/max:2150.0 - response_length_non_aborted/min:86.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:282.28125 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.868107736110687e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0074804201722145) - timing_s/agent_loop/generate_sequences/max:np.float64(12.549119051545858) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.2516225457002292) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(12.549119051545858) - timing_s/agent_loop/s
+step:25 - global_seqlen/min:19247 - global_seqlen/max:30356 - global_seqlen/minmax_diff:11109 - global_seqlen/balanced_min:26672 - global_seqlen/balanced_max:27390 - global_seqlen/mean:26767.875 - actor/entropy:0.18973146378993988 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.33718350529670715 - training/rollout_probs_diff_mean:0.004361078143119812 - training/rollout_probs_diff_std:0.012972155585885048 - training/rollout_actor_probs_pearson_corr:0.9981417059898376 - rollout_corr/rollout_is_mean:0.999921441078186 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:6.597230822080746e-05 - rollout_corr/rollout_is_max:1.7342352867126465 - rollout_corr/rollout_is_min:0.2947933077812195 - rollout_corr/rollout_is_std:0.038447305560112 - rollout_corr/rollout_is_eff_sample_size:0.9985237489301881 - rollout_corr/rollout_is_seq_mean:0.9999851584434509 - rollout_corr/rollout_is_seq_std:0.0023914307821542025 - rollout_corr/rollout_is_seq_max:1.012315273284912 - rollout_corr/rollout_is_seq_min:0.9918898344039917 - rollout_corr/rollout_is_seq_max_deviation:0.01231527328491211 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2648270050995052) - rollout_corr/training_log_ppl:np.float64(0.23086260134004988) - rollout_corr/kl:np.float64(0.0009177963597579719) - rollout_corr/k3_kl:np.float64(0.001010191431845442) - rollout_corr/rollout_ppl:np.float64(1.263638133648783) - rollout_corr/rollout_log_ppl:np.float64(0.22994480418856256) - rollout_corr/log_ppl_diff:np.float64(0.0009177971514873207) - rollout_corr/log_ppl_abs_diff:np.float64(0.0021577174484264106) - rollout_corr/log_ppl_diff_max:np.float64(0.01115649938583374) - rollout_corr/log_ppl_diff_min:np.float64(-0.010651543736457825) - rollout_corr/ppl_ratio:np.float64(1.000921874307096) - rollout_corr/chi2_token:np.float64(0.0022637455258518457) - rollout_corr/chi2_seq:np.float64(0.9759733870159835) - actor/pg_loss:np.float64(0.0001767532667145133) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011058543057060888) - actor/ppo_kl:np.float64(-2.0648154892555226e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.553518146276474) - perf/mfu/actor:np.float64(0.07675197121879002) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.21955585479736) - actor/lr:np.float64(1e-06) - training/global_step:25 - training/epoch:1 - critic/score/mean:0.3828125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.3828125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.021515218541026115 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.021515218541026115 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:473.68359375 - response_length/max:2715.0 - response_length/min:106.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:473.68359375 - response_length_non_aborted/max:2715.0 - response_length_non_aborted/min:106.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:362.8125 - prompt_length/max:499.0 - prompt_length/min:181.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:7.77803361415863e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.284107943996787) - timing_s/agent_loop/generate_sequences/max:np.float64(15.839517647400498) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.669048825882783) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(15.839517647400498) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/a
+step:26 - global_seqlen/min:14395 - global_seqlen/max:27159 - global_seqlen/minmax_diff:12764 - global_seqlen/balanced_min:22165 - global_seqlen/balanced_max:22193 - global_seqlen/mean:22185.625 - actor/entropy:0.1997486650943756 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.33525192737579346 - training/rollout_probs_diff_mean:0.004470215644687414 - training/rollout_probs_diff_std:0.012864506803452969 - training/rollout_actor_probs_pearson_corr:0.9982172250747681 - rollout_corr/rollout_is_mean:1.0001190900802612 - rollout_corr/rollout_is_ratio_fraction_high:1.0165391358896159e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.0992344515398145e-05 - rollout_corr/rollout_is_max:3.4254493713378906 - rollout_corr/rollout_is_min:0.33838894963264465 - rollout_corr/rollout_is_std:0.03890654817223549 - rollout_corr/rollout_is_eff_sample_size:0.9984888060234459 - rollout_corr/rollout_is_seq_mean:1.0002080202102661 - rollout_corr/rollout_is_seq_std:0.0026749917306005955 - rollout_corr/rollout_is_seq_max:1.0075410604476929 - rollout_corr/rollout_is_seq_min:0.9928621649742126 - rollout_corr/rollout_is_seq_max_deviation:0.007541060447692871 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2674571070820093) - rollout_corr/training_log_ppl:np.float64(0.23334544732642826) - rollout_corr/kl:np.float64(0.0006923889370149539) - rollout_corr/k3_kl:np.float64(0.0010116354250726545) - rollout_corr/rollout_ppl:np.float64(1.266531832050532) - rollout_corr/rollout_log_ppl:np.float64(0.23265305937093217) - rollout_corr/log_ppl_diff:np.float64(0.0006923879554960877) - rollout_corr/log_ppl_abs_diff:np.float64(0.0022414404957089573) - rollout_corr/log_ppl_diff_max:np.float64(0.009537696838378906) - rollout_corr/log_ppl_diff_min:np.float64(-0.008276775479316711) - rollout_corr/ppl_ratio:np.float64(1.0006968146190047) - rollout_corr/chi2_token:np.float64(0.0026694873813539743) - rollout_corr/chi2_seq:np.float64(0.8543046184349805) - actor/pg_loss:np.float64(-2.134358510375023e-06) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0006741704369233048) - actor/ppo_kl:np.float64(8.118165057968696e-06) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.37092382460832596) - perf/mfu/actor:np.float64(0.06470432805583136) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.13681030273438) - actor/lr:np.float64(1e-06) - training/global_step:26 - training/epoch:1 - critic/score/mean:0.41015625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.41015625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.005917528178542852 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.005917528178542852 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:384.26953125 - response_length/max:1567.0 - response_length/min:83.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:384.26953125 - response_length_non_aborted/max:1567.0 - response_length_non_aborted/min:83.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:309.03125 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00010296143591403961 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9918509125709534) - timing_s/agent_loop/generate_sequences/max:np.float64(10.191927688196301) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.6913370271213353) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(10.191927688196301) - timing_s/agent_loop/slowest/tool_
+step:27 - global_seqlen/min:12650 - global_seqlen/max:28413 - global_seqlen/minmax_diff:15763 - global_seqlen/balanced_min:22006 - global_seqlen/balanced_max:22028 - global_seqlen/mean:22020.875 - actor/entropy:0.1973375827074051 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2553977966308594 - training/rollout_probs_diff_mean:0.00446832412853837 - training/rollout_probs_diff_std:0.012732885777950287 - training/rollout_actor_probs_pearson_corr:0.9982201457023621 - rollout_corr/rollout_is_mean:1.0000790357589722 - rollout_corr/rollout_is_ratio_fraction_high:2.0880532247247174e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.132079655188136e-05 - rollout_corr/rollout_is_max:3.3716399669647217 - rollout_corr/rollout_is_min:0.41315898299217224 - rollout_corr/rollout_is_std:0.03870687633752823 - rollout_corr/rollout_is_eff_sample_size:0.9985042566669848 - rollout_corr/rollout_is_seq_mean:1.0001776218414307 - rollout_corr/rollout_is_seq_std:0.002695156028494239 - rollout_corr/rollout_is_seq_max:1.01008939743042 - rollout_corr/rollout_is_seq_min:0.9929530024528503 - rollout_corr/rollout_is_seq_max_deviation:0.010089397430419922 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2657123478129506) - rollout_corr/training_log_ppl:np.float64(0.23218309982621577) - rollout_corr/kl:np.float64(0.0008745037185420301) - rollout_corr/k3_kl:np.float64(0.0010085490487199422) - rollout_corr/rollout_ppl:np.float64(1.264582925941795) - rollout_corr/rollout_log_ppl:np.float64(0.2313085944915656) - rollout_corr/log_ppl_diff:np.float64(0.0008745053346501663) - rollout_corr/log_ppl_abs_diff:np.float64(0.0021427856263471767) - rollout_corr/log_ppl_diff_max:np.float64(0.011478304862976074) - rollout_corr/log_ppl_diff_min:np.float64(-0.008908957242965698) - rollout_corr/ppl_ratio:np.float64(1.0008788730483502) - rollout_corr/chi2_token:np.float64(0.0022939874324947596) - rollout_corr/chi2_seq:np.float64(0.9378205249086022) - actor/pg_loss:np.float64(4.4271000660955906e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0008205796011679922) - actor/ppo_kl:np.float64(0.00014555726753329168) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.377008780837059) - perf/mfu/actor:np.float64(0.0638297454911889) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.12808609008789) - actor/lr:np.float64(1e-06) - training/global_step:27 - training/epoch:1 - critic/score/mean:0.296875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.296875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.014912614598870277 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.014912614598870277 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:374.15234375 - response_length/max:1612.0 - response_length/min:92.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:374.15234375 - response_length_non_aborted/max:1612.0 - response_length_non_aborted/min:92.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:314.0 - prompt_length/max:499.0 - prompt_length/min:181.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:6.013922393321991e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1771093513816595) - timing_s/agent_loop/generate_sequences/max:np.float64(10.33962314389646) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.6523388316636556) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(10.33962314389646) - timing_s/agent_loop/slowest/tool_calls:np.float64(0
+step:28 - global_seqlen/min:17510 - global_seqlen/max:32402 - global_seqlen/minmax_diff:14892 - global_seqlen/balanced_min:24287 - global_seqlen/balanced_max:24336 - global_seqlen/mean:24323.75 - actor/entropy:0.19081273674964905 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.35365766286849976 - training/rollout_probs_diff_mean:0.004214483313262463 - training/rollout_probs_diff_std:0.012326523661613464 - training/rollout_actor_probs_pearson_corr:0.9983044266700745 - rollout_corr/rollout_is_mean:0.9999059438705444 - rollout_corr/rollout_is_ratio_fraction_high:1.8814322174875997e-05 - rollout_corr/rollout_is_ratio_fraction_low:9.407160541741177e-05 - rollout_corr/rollout_is_max:2.637937545776367 - rollout_corr/rollout_is_min:0.28466564416885376 - rollout_corr/rollout_is_std:0.037621427327394485 - rollout_corr/rollout_is_eff_sample_size:0.9985862719338552 - rollout_corr/rollout_is_seq_mean:0.999921977519989 - rollout_corr/rollout_is_seq_std:0.0030063134618103504 - rollout_corr/rollout_is_seq_max:1.014600157737732 - rollout_corr/rollout_is_seq_min:0.990456759929657 - rollout_corr/rollout_is_seq_max_deviation:0.014600157737731934 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2607678300701082) - rollout_corr/training_log_ppl:np.float64(0.22826670162612572) - rollout_corr/kl:np.float64(0.0009433091491317658) - rollout_corr/k3_kl:np.float64(0.0009751564648468047) - rollout_corr/rollout_ppl:np.float64(1.2595306900329888) - rollout_corr/rollout_log_ppl:np.float64(0.22732339147478342) - rollout_corr/log_ppl_diff:np.float64(0.0009433101513423026) - rollout_corr/log_ppl_abs_diff:np.float64(0.0022531004797201604) - rollout_corr/log_ppl_diff_max:np.float64(0.010257154703140259) - rollout_corr/log_ppl_diff_min:np.float64(-0.012321293354034424) - rollout_corr/ppl_ratio:np.float64(1.0009483909234405) - rollout_corr/chi2_token:np.float64(0.002010138938203454) - rollout_corr/chi2_seq:np.float64(1.263109202729538) - actor/pg_loss:np.float64(0.0003952222177758813) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.000980165132887123) - actor/ppo_kl:np.float64(-2.858693859408845e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5482866317033768) - perf/mfu/actor:np.float64(0.07006718373692834) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.16790771484375) - actor/lr:np.float64(1e-06) - training/global_step:28 - training/epoch:1 - critic/score/mean:0.34765625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.34765625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.015940433368086815 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.015940433368086815 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:415.2421875 - response_length/max:1856.0 - response_length/min:86.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:415.2421875 - response_length_non_aborted/max:1856.0 - response_length_non_aborted/min:86.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:344.875 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.274482727050781e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9281809013336897) - timing_s/agent_loop/generate_sequences/max:np.float64(11.828542461618781) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.940465694955492) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(11.828542461618781) - timing_s/agent_loop/slowest/tool_calls:np.fl
+step:29 - global_seqlen/min:11576 - global_seqlen/max:26504 - global_seqlen/minmax_diff:14928 - global_seqlen/balanced_min:20443 - global_seqlen/balanced_max:20468 - global_seqlen/mean:20457.125 - actor/entropy:0.19512885808944702 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2994707226753235 - training/rollout_probs_diff_mean:0.004517699591815472 - training/rollout_probs_diff_std:0.012806838378310204 - training/rollout_actor_probs_pearson_corr:0.9982004165649414 - rollout_corr/rollout_is_mean:0.9999142289161682 - rollout_corr/rollout_is_ratio_fraction_high:2.2683709175908007e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.5367418351816013e-05 - rollout_corr/rollout_is_max:2.5106632709503174 - rollout_corr/rollout_is_min:0.35614722967147827 - rollout_corr/rollout_is_std:0.03914633393287659 - rollout_corr/rollout_is_eff_sample_size:0.9984696716584548 - rollout_corr/rollout_is_seq_mean:0.9998313188552856 - rollout_corr/rollout_is_seq_std:0.0028376157861202955 - rollout_corr/rollout_is_seq_max:1.0107405185699463 - rollout_corr/rollout_is_seq_min:0.9908356666564941 - rollout_corr/rollout_is_seq_max_deviation:0.010740518569946289 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2674566823989153) - rollout_corr/training_log_ppl:np.float64(0.23359788639936596) - rollout_corr/kl:np.float64(0.0009326249134602449) - rollout_corr/k3_kl:np.float64(0.0010803121514300074) - rollout_corr/rollout_ppl:np.float64(1.2662694193422794) - rollout_corr/rollout_log_ppl:np.float64(0.2326652607152937) - rollout_corr/log_ppl_diff:np.float64(0.0009326256840722635) - rollout_corr/log_ppl_abs_diff:np.float64(0.002509462457965128) - rollout_corr/log_ppl_diff_max:np.float64(0.009627044200897217) - rollout_corr/log_ppl_diff_min:np.float64(-0.010430455207824707) - rollout_corr/ppl_ratio:np.float64(1.0009380243718624) - rollout_corr/chi2_token:np.float64(0.0024118823930621147) - rollout_corr/chi2_seq:np.float64(1.6191739679779857) - actor/pg_loss:np.float64(0.00010256993118673563) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001801439915652736) - actor/ppo_kl:np.float64(-0.0001434271987896807) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5583970844745636) - perf/mfu/actor:np.float64(0.05960628937352358) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.0120849609375) - actor/lr:np.float64(1e-06) - training/global_step:29 - training/epoch:2 - critic/score/mean:0.296875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.296875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.008349022828042507 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.008349022828042507 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:344.41015625 - response_length/max:1501.0 - response_length/min:87.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:344.41015625 - response_length_non_aborted/max:1501.0 - response_length_non_aborted/min:87.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:294.875 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00034629926085472107 - timing_s/agent_loop/generate_sequences/min:np.float64(0.7719959896057844) - timing_s/agent_loop/generate_sequences/max:np.float64(9.23289637081325) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.100054759372142) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.23289637081325) - timing_s/agent_loop/slowest/tool_calls:np.float
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 30}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score of a protein sequence typically requires computational analysis using tools like Rosetta, which calculate the energy of a protein structure. The sequence provided ('GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR') would need to be input into such a tool to determine its folding stability score. Since no computational tool or database is available here to calculate this score, it is not possible to determine the exact folding stability score without additional information or computational resources.
+</reasoning>
+<answer>
+D
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] D
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_30
+step:30 - global_seqlen/min:14663 - global_seqlen/max:34744 - global_seqlen/minmax_diff:20081 - global_seqlen/balanced_min:23097 - global_seqlen/balanced_max:23449 - global_seqlen/mean:23160.25 - actor/entropy:0.1929744929075241 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.34270334243774414 - training/rollout_probs_diff_mean:0.004414478316903114 - training/rollout_probs_diff_std:0.012888044118881226 - training/rollout_actor_probs_pearson_corr:0.9981456995010376 - rollout_corr/rollout_is_mean:1.000051736831665 - rollout_corr/rollout_is_ratio_fraction_high:1.9112783775199205e-05 - rollout_corr/rollout_is_ratio_fraction_low:9.556392615195364e-05 - rollout_corr/rollout_is_max:2.849820375442505 - rollout_corr/rollout_is_min:0.13807697594165802 - rollout_corr/rollout_is_std:0.03877303749322891 - rollout_corr/rollout_is_eff_sample_size:0.9984991460165831 - rollout_corr/rollout_is_seq_mean:0.9999996423721313 - rollout_corr/rollout_is_seq_std:0.002868861658498645 - rollout_corr/rollout_is_seq_max:1.0089269876480103 - rollout_corr/rollout_is_seq_min:0.989052414894104 - rollout_corr/rollout_is_seq_max_deviation:0.010947585105895996 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2640721159987152) - rollout_corr/training_log_ppl:np.float64(0.23040971330192406) - rollout_corr/kl:np.float64(0.001080727721191721) - rollout_corr/k3_kl:np.float64(0.0009677562554202268) - rollout_corr/rollout_ppl:np.float64(1.2626397949643433) - rollout_corr/rollout_log_ppl:np.float64(0.22932898352155462) - rollout_corr/log_ppl_diff:np.float64(0.0010807297803694382) - rollout_corr/log_ppl_abs_diff:np.float64(0.002270795885124244) - rollout_corr/log_ppl_diff_max:np.float64(0.009988278150558472) - rollout_corr/log_ppl_diff_min:np.float64(-0.008100628852844238) - rollout_corr/ppl_ratio:np.float64(1.0010852506384254) - rollout_corr/chi2_token:np.float64(0.001721000298857689) - rollout_corr/chi2_seq:np.float64(0.7308730445802212) - actor/pg_loss:np.float64(0.00017032702453434467) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.00098742782915906) - actor/ppo_kl:np.float64(0.0001257946889916539) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.562660925090313) - perf/mfu/actor:np.float64(0.06606835743079029) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.89504623413086) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.3825) - val-aux/sciknoweval/reward/std@16:np.float64(0.2710825289783396) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.49848) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.23249398932762172) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.26538000000000006) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.23000654365019318) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.38264) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.2717389471729169) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.5914200000000001) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.16401609741962153) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.17232) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.16309037403563342) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.4011) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.2275492223932495) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.65118) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.1002078489826257) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.11272) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.0901338588728539) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.4145999999999999) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.17744782663880587) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.68952) - val-aux/sciknoweval/reward/best@16/std:np.float64(0.05884117194239494) - val-aux/sciknoweva
+step:31 - global_seqlen/min:15610 - global_seqlen/max:32987 - global_seqlen/minmax_diff:17377 - global_seqlen/balanced_min:22697 - global_seqlen/balanced_max:22706 - global_seqlen/mean:22704.125 - actor/entropy:0.20467621088027954 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.36793190240859985 - training/rollout_probs_diff_mean:0.0046606664545834064 - training/rollout_probs_diff_std:0.013037621974945068 - training/rollout_actor_probs_pearson_corr:0.998202919960022 - rollout_corr/rollout_is_mean:0.9999342560768127 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:6.056140409782529e-05 - rollout_corr/rollout_is_max:1.8634686470031738 - rollout_corr/rollout_is_min:0.08700346946716309 - rollout_corr/rollout_is_std:0.03880991414189339 - rollout_corr/rollout_is_eff_sample_size:0.998495937030278 - rollout_corr/rollout_is_seq_mean:0.9998412132263184 - rollout_corr/rollout_is_seq_std:0.0024558771401643753 - rollout_corr/rollout_is_seq_max:1.008089303970337 - rollout_corr/rollout_is_seq_min:0.9901171326637268 - rollout_corr/rollout_is_seq_max_deviation:0.009882867336273193 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2660416262224317) - rollout_corr/training_log_ppl:np.float64(0.2328464862657711) - rollout_corr/kl:np.float64(0.0014669340525781394) - rollout_corr/k3_kl:np.float64(0.0012662450398579495) - rollout_corr/rollout_ppl:np.float64(1.2641333830542862) - rollout_corr/rollout_log_ppl:np.float64(0.2313795514346566) - rollout_corr/log_ppl_diff:np.float64(0.0014669348311144859) - rollout_corr/log_ppl_abs_diff:np.float64(0.0024422181595582515) - rollout_corr/log_ppl_diff_max:np.float64(0.0128212571144104) - rollout_corr/log_ppl_diff_min:np.float64(-0.01384890079498291) - rollout_corr/ppl_ratio:np.float64(1.0014724703505635) - rollout_corr/chi2_token:np.float64(0.002137396950274706) - rollout_corr/chi2_seq:np.float64(0.24568918230943382) - actor/pg_loss:np.float64(-4.5661814510822296e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002057411175883317) - actor/ppo_kl:np.float64(0.00041522945140526346) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5537090376019478) - perf/mfu/actor:np.float64(0.0655864265337439) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.2952651977539) - actor/lr:np.float64(1e-06) - training/global_step:31 - training/epoch:2 - critic/score/mean:0.375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.009520757012069225 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.009520757012069225 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:387.00390625 - response_length/max:1622.0 - response_length/min:87.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:387.00390625 - response_length_non_aborted/max:1622.0 - response_length_non_aborted/min:87.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:322.5 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.000301973894238472 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0549929738044739) - timing_s/agent_loop/generate_sequences/max:np.float64(10.455312624573708) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.841500379559875) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(10.455312624573708) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent_loo
+step:32 - global_seqlen/min:16456 - global_seqlen/max:30003 - global_seqlen/minmax_diff:13547 - global_seqlen/balanced_min:22662 - global_seqlen/balanced_max:22713 - global_seqlen/mean:22703.375 - actor/entropy:0.19235235452651978 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.6836323142051697 - training/rollout_probs_diff_mean:0.004288580734282732 - training/rollout_probs_diff_std:0.01270546205341816 - training/rollout_actor_probs_pearson_corr:0.9982287287712097 - rollout_corr/rollout_is_mean:1.000108003616333 - rollout_corr/rollout_is_ratio_fraction_high:4.2221261537633836e-05 - rollout_corr/rollout_is_ratio_fraction_low:1.0555315384408459e-05 - rollout_corr/rollout_is_max:3.5419375896453857 - rollout_corr/rollout_is_min:0.4771340489387512 - rollout_corr/rollout_is_std:0.03740055859088898 - rollout_corr/rollout_is_eff_sample_size:0.9986033898627913 - rollout_corr/rollout_is_seq_mean:1.0001899003982544 - rollout_corr/rollout_is_seq_std:0.002721930155530572 - rollout_corr/rollout_is_seq_max:1.009945273399353 - rollout_corr/rollout_is_seq_min:0.9931846857070923 - rollout_corr/rollout_is_seq_max_deviation:0.009945273399353027 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2627140362747014) - rollout_corr/training_log_ppl:np.float64(0.22928791525191627) - rollout_corr/kl:np.float64(0.0009696064506714563) - rollout_corr/k3_kl:np.float64(0.0009657189097538321) - rollout_corr/rollout_ppl:np.float64(1.2614360507577658) - rollout_corr/rollout_log_ppl:np.float64(0.228318307781592) - rollout_corr/log_ppl_diff:np.float64(0.000969607470324263) - rollout_corr/log_ppl_abs_diff:np.float64(0.0024012566718738526) - rollout_corr/log_ppl_diff_max:np.float64(0.008903712034225464) - rollout_corr/log_ppl_diff_min:np.float64(-0.007482409477233887) - rollout_corr/ppl_ratio:np.float64(1.0009746660944074) - rollout_corr/chi2_token:np.float64(0.0019679118413478136) - rollout_corr/chi2_seq:np.float64(0.5128269151318818) - actor/pg_loss:np.float64(0.0006739883683621883) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011445526070019696) - actor/ppo_kl:np.float64(0.00029994351070516956) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5374793410301208) - perf/mfu/actor:np.float64(0.06520734277390705) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.37225103378296) - actor/lr:np.float64(1e-06) - training/global_step:32 - training/epoch:2 - critic/score/mean:0.5078125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.5078125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0055389017798006535 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.0055389017798006535 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:370.07421875 - response_length/max:1794.0 - response_length/min:84.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:370.07421875 - response_length_non_aborted/max:1794.0 - response_length_non_aborted/min:84.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:339.40625 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.293854236602783e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0003391448408365) - timing_s/agent_loop/generate_sequences/max:np.float64(11.236332735046744) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.5134957447080524) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(11.236332735046744) - timing_s/agent_loop/slowest/tool_calls:np.
+step:33 - global_seqlen/min:14153 - global_seqlen/max:31822 - global_seqlen/minmax_diff:17669 - global_seqlen/balanced_min:24569 - global_seqlen/balanced_max:24592 - global_seqlen/mean:24586.5 - actor/entropy:0.19219401478767395 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.38829466700553894 - training/rollout_probs_diff_mean:0.004298380576074123 - training/rollout_probs_diff_std:0.012502633035182953 - training/rollout_actor_probs_pearson_corr:0.9982357621192932 - rollout_corr/rollout_is_mean:0.9999257922172546 - rollout_corr/rollout_is_ratio_fraction_high:1.8696480765356682e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.674120282288641e-05 - rollout_corr/rollout_is_max:3.7463631629943848 - rollout_corr/rollout_is_min:0.3203428089618683 - rollout_corr/rollout_is_std:0.03699354827404022 - rollout_corr/rollout_is_eff_sample_size:0.9986331098833465 - rollout_corr/rollout_is_seq_mean:0.9999837875366211 - rollout_corr/rollout_is_seq_std:0.002320910105481744 - rollout_corr/rollout_is_seq_max:1.0067297220230103 - rollout_corr/rollout_is_seq_min:0.9923191666603088 - rollout_corr/rollout_is_seq_max_deviation:0.007680833339691162 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2546508102677763) - rollout_corr/training_log_ppl:np.float64(0.22293072564934846) - rollout_corr/kl:np.float64(0.0007883897724163154) - rollout_corr/k3_kl:np.float64(0.0008786465269849941) - rollout_corr/rollout_ppl:np.float64(1.2536278427578509) - rollout_corr/rollout_log_ppl:np.float64(0.22214233508566394) - rollout_corr/log_ppl_diff:np.float64(0.0007883905636845157) - rollout_corr/log_ppl_abs_diff:np.float64(0.0019512200815370306) - rollout_corr/log_ppl_diff_max:np.float64(0.00915798544883728) - rollout_corr/log_ppl_diff_min:np.float64(-0.007831752300262451) - rollout_corr/ppl_ratio:np.float64(1.0007919785566628) - rollout_corr/chi2_token:np.float64(0.0019339756108820438) - rollout_corr/chi2_seq:np.float64(0.8230726562906057) - actor/pg_loss:np.float64(6.270431913435459e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0013706558816011238) - actor/ppo_kl:np.float64(-7.741855117160412e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6440355181694031) - perf/mfu/actor:np.float64(0.0710191135212451) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.29852676391602) - actor/lr:np.float64(1e-06) - training/global_step:33 - training/epoch:2 - critic/score/mean:0.453125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.453125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.00058893917594105 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.00058893917594105 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:417.859375 - response_length/max:1500.0 - response_length/min:109.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:417.859375 - response_length_non_aborted/max:1500.0 - response_length_non_aborted/min:109.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:350.46875 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:8.15168023109436e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.3214661497622728) - timing_s/agent_loop/generate_sequences/max:np.float64(10.396715246140957) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.068895058728231) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(10.396715246140957) - timing_s/agent_loop/slowest/tool_calls:np.float64(
+step:34 - global_seqlen/min:13165 - global_seqlen/max:31218 - global_seqlen/minmax_diff:18053 - global_seqlen/balanced_min:20844 - global_seqlen/balanced_max:21371 - global_seqlen/mean:20953.375 - actor/entropy:0.20621417462825775 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.47441184520721436 - training/rollout_probs_diff_mean:0.004632072988897562 - training/rollout_probs_diff_std:0.013067140243947506 - training/rollout_actor_probs_pearson_corr:0.9982191324234009 - rollout_corr/rollout_is_mean:0.9996575117111206 - rollout_corr/rollout_is_ratio_fraction_high:1.123557649407303e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.741345714544877e-05 - rollout_corr/rollout_is_max:3.048807144165039 - rollout_corr/rollout_is_min:0.1631394475698471 - rollout_corr/rollout_is_std:0.039033498615026474 - rollout_corr/rollout_is_eff_sample_size:0.9984777531641464 - rollout_corr/rollout_is_seq_mean:0.9995238184928894 - rollout_corr/rollout_is_seq_std:0.003114572959020734 - rollout_corr/rollout_is_seq_max:1.0097143650054932 - rollout_corr/rollout_is_seq_min:0.9892820119857788 - rollout_corr/rollout_is_seq_max_deviation:0.010717988014221191 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2769425138831139) - rollout_corr/training_log_ppl:np.float64(0.24088622524868697) - rollout_corr/kl:np.float64(0.0017434109029323963) - rollout_corr/k3_kl:np.float64(0.0011406559008833028) - rollout_corr/rollout_ppl:np.float64(1.2746948134154081) - rollout_corr/rollout_log_ppl:np.float64(0.23914281158067752) - rollout_corr/log_ppl_diff:np.float64(0.0017434136680094525) - rollout_corr/log_ppl_abs_diff:np.float64(0.002788226425764151) - rollout_corr/log_ppl_diff_max:np.float64(0.012063741683959961) - rollout_corr/log_ppl_diff_min:np.float64(-0.007839560508728027) - rollout_corr/ppl_ratio:np.float64(1.001750238938257) - rollout_corr/chi2_token:np.float64(0.0010821421165019274) - rollout_corr/chi2_seq:np.float64(0.3799931020475924) - actor/pg_loss:np.float64(0.00013993948232382536) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0016885243967408314) - actor/ppo_kl:np.float64(0.0003524259806644281) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5224123150110245) - perf/mfu/actor:np.float64(0.06054932786011055) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.48657608032227) - actor/lr:np.float64(1e-06) - training/global_step:34 - training/epoch:2 - critic/score/mean:0.3828125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.3828125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.01508656982332468 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.01508656982332468 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:347.66796875 - response_length/max:2290.0 - response_length/min:83.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:347.66796875 - response_length_non_aborted/max:2290.0 - response_length_non_aborted/min:83.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:307.125 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.014829993247986e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0219873953610659) - timing_s/agent_loop/generate_sequences/max:np.float64(13.463359158486128) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.3108865806207177) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(13.463359158486128) - timing_s/agent_loop/slowest/tool_calls:np.f
+step:35 - global_seqlen/min:10930 - global_seqlen/max:29063 - global_seqlen/minmax_diff:18133 - global_seqlen/balanced_min:21182 - global_seqlen/balanced_max:21279 - global_seqlen/mean:21202.375 - actor/entropy:0.20077916979789734 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.4047938883304596 - training/rollout_probs_diff_mean:0.004468213766813278 - training/rollout_probs_diff_std:0.012723792344331741 - training/rollout_actor_probs_pearson_corr:0.9982172250747681 - rollout_corr/rollout_is_mean:0.9999480247497559 - rollout_corr/rollout_is_ratio_fraction_high:1.1354991329426412e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.4064974897773936e-05 - rollout_corr/rollout_is_max:2.361192226409912 - rollout_corr/rollout_is_min:0.35472583770751953 - rollout_corr/rollout_is_std:0.03832307830452919 - rollout_corr/rollout_is_eff_sample_size:0.9985334953471807 - rollout_corr/rollout_is_seq_mean:0.9999346137046814 - rollout_corr/rollout_is_seq_std:0.0025622511748224497 - rollout_corr/rollout_is_seq_max:1.0083149671554565 - rollout_corr/rollout_is_seq_min:0.9895864129066467 - rollout_corr/rollout_is_seq_max_deviation:0.010413587093353271 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2627762155607343) - rollout_corr/training_log_ppl:np.float64(0.22958877607015893) - rollout_corr/kl:np.float64(0.001482734944080022) - rollout_corr/k3_kl:np.float64(0.001303993512976831) - rollout_corr/rollout_ppl:np.float64(1.2608598791994154) - rollout_corr/rollout_log_ppl:np.float64(0.22810603884863667) - rollout_corr/log_ppl_diff:np.float64(0.0014827372215222567) - rollout_corr/log_ppl_abs_diff:np.float64(0.0028337313269730657) - rollout_corr/log_ppl_diff_max:np.float64(0.01417878270149231) - rollout_corr/log_ppl_diff_min:np.float64(-0.015369802713394165) - rollout_corr/ppl_ratio:np.float64(1.0014910355675966) - rollout_corr/chi2_token:np.float64(0.002457682741805911) - rollout_corr/chi2_seq:np.float64(1.00893239909783) - actor/pg_loss:np.float64(0.0003223039675503969) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0018587432675758464) - actor/ppo_kl:np.float64(0.000577584969583711) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.604684367775917) - perf/mfu/actor:np.float64(0.06179983949184756) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.36828994750977) - actor/lr:np.float64(1e-06) - training/global_step:35 - training/epoch:2 - critic/score/mean:0.35546875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.35546875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.004048054106533527 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.004048054106533527 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:344.01171875 - response_length/max:1733.0 - response_length/min:83.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:344.01171875 - response_length_non_aborted/max:1733.0 - response_length_non_aborted/min:83.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:318.5625 - prompt_length/max:498.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:8.777901530265808e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.7171750254929066) - timing_s/agent_loop/generate_sequences/max:np.float64(8.89642646163702) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.0253594234527554) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(8.89642646163702) - timing_s/agent_loop/slowest/tool_calls:np.float
+step:36 - global_seqlen/min:14194 - global_seqlen/max:29137 - global_seqlen/minmax_diff:14943 - global_seqlen/balanced_min:20936 - global_seqlen/balanced_max:21371 - global_seqlen/mean:21018.125 - actor/entropy:0.20421844720840454 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5412930250167847 - training/rollout_probs_diff_mean:0.00441008573397994 - training/rollout_probs_diff_std:0.012608489021658897 - training/rollout_actor_probs_pearson_corr:0.9983038306236267 - rollout_corr/rollout_is_mean:0.9999710917472839 - rollout_corr/rollout_is_ratio_fraction_high:2.242931986984331e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.607329876511358e-05 - rollout_corr/rollout_is_max:2.2185213565826416 - rollout_corr/rollout_is_min:0.007402495015412569 - rollout_corr/rollout_is_std:0.03787250071763992 - rollout_corr/rollout_is_eff_sample_size:0.9985674904370404 - rollout_corr/rollout_is_seq_mean:0.999952495098114 - rollout_corr/rollout_is_seq_std:0.0027767117135226727 - rollout_corr/rollout_is_seq_max:1.008007526397705 - rollout_corr/rollout_is_seq_min:0.9892098307609558 - rollout_corr/rollout_is_seq_max_deviation:0.01079016923904419 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.27850745851174) - rollout_corr/training_log_ppl:np.float64(0.24139065398776438) - rollout_corr/kl:np.float64(0.0012924192194674333) - rollout_corr/k3_kl:np.float64(0.0011641595515641256) - rollout_corr/rollout_ppl:np.float64(1.2768099810928106) - rollout_corr/rollout_log_ppl:np.float64(0.24009823182132095) - rollout_corr/log_ppl_diff:np.float64(0.0012924221664434299) - rollout_corr/log_ppl_abs_diff:np.float64(0.002743791919783689) - rollout_corr/log_ppl_diff_max:np.float64(0.017144635319709778) - rollout_corr/log_ppl_diff_min:np.float64(-0.00719086080789566) - rollout_corr/ppl_ratio:np.float64(1.0013000178150833) - rollout_corr/chi2_token:np.float64(0.002002155175432563) - rollout_corr/chi2_seq:np.float64(0.773878968320787) - actor/pg_loss:np.float64(-6.605871021747589e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0014383478630861646) - actor/ppo_kl:np.float64(0.00034995766268330897) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5709159523248672) - perf/mfu/actor:np.float64(0.061025477795435934) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.43119430541992) - actor/lr:np.float64(1e-06) - training/global_step:36 - training/epoch:2 - critic/score/mean:0.34375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.34375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0012700601946562529 - critic/advantages/max:0.75 - critic/advantages/min:-0.875 - critic/returns/mean:0.0012700601946562529 - critic/returns/max:0.75 - critic/returns/min:-0.875 - response_length/mean:348.31640625 - response_length/max:2327.0 - response_length/min:74.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:348.31640625 - response_length_non_aborted/max:2327.0 - response_length_non_aborted/min:74.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:308.5 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00013381987810134888 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9047284964472055) - timing_s/agent_loop/generate_sequences/max:np.float64(13.386295091360807) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.3439106781152077) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(13.386295091360807) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.
+step:37 - global_seqlen/min:10542 - global_seqlen/max:30603 - global_seqlen/minmax_diff:20061 - global_seqlen/balanced_min:19813 - global_seqlen/balanced_max:26160 - global_seqlen/mean:20656.375 - actor/entropy:0.19215349853038788 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.4102194607257843 - training/rollout_probs_diff_mean:0.004060864448547363 - training/rollout_probs_diff_std:0.012080005370080471 - training/rollout_actor_probs_pearson_corr:0.9984062314033508 - rollout_corr/rollout_is_mean:0.9999275803565979 - rollout_corr/rollout_is_ratio_fraction_high:1.1167319826199673e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.583660094998777e-05 - rollout_corr/rollout_is_max:2.5254178047180176 - rollout_corr/rollout_is_min:0.23548689484596252 - rollout_corr/rollout_is_std:0.03610971197485924 - rollout_corr/rollout_is_eff_sample_size:0.9986974300643999 - rollout_corr/rollout_is_seq_mean:0.9998934268951416 - rollout_corr/rollout_is_seq_std:0.002943673636764288 - rollout_corr/rollout_is_seq_max:1.009252905845642 - rollout_corr/rollout_is_seq_min:0.9870862364768982 - rollout_corr/rollout_is_seq_max_deviation:0.012913763523101807 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2790742232464254) - rollout_corr/training_log_ppl:np.float64(0.2423434283118695) - rollout_corr/kl:np.float64(0.0010189136970133461) - rollout_corr/k3_kl:np.float64(0.0010748106194000684) - rollout_corr/rollout_ppl:np.float64(1.277731901500374) - rollout_corr/rollout_log_ppl:np.float64(0.24132451312470948) - rollout_corr/log_ppl_diff:np.float64(0.0010189151871600188) - rollout_corr/log_ppl_abs_diff:np.float64(0.002592919896414969) - rollout_corr/log_ppl_diff_max:np.float64(0.015276342630386353) - rollout_corr/log_ppl_diff_min:np.float64(-0.011511757969856262) - rollout_corr/ppl_ratio:np.float64(1.0010252590291202) - rollout_corr/chi2_token:np.float64(0.0022940633352845907) - rollout_corr/chi2_seq:np.float64(0.5807141268160194) - actor/pg_loss:np.float64(4.6521308831870556e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0013833769742177537) - actor/ppo_kl:np.float64(4.1046847659931274e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5195856019854546) - perf/mfu/actor:np.float64(0.05586420816152182) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.32366943359375) - actor/lr:np.float64(1e-06) - training/global_step:37 - training/epoch:2 - critic/score/mean:0.359375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.359375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.03285565227270126 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.03285565227270126 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:349.79296875 - response_length/max:8192.0 - response_length/min:81.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:349.79296875 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:81.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:295.71875 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00016735494136810303 - timing_s/agent_loop/generate_sequences/min:np.float64(0.6763673182576895) - timing_s/agent_loop/generate_sequences/max:np.float64(49.16137577034533) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.9194193026196444) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.16137577034533) - timing_s/agent_loop/slowest/to
+step:38 - global_seqlen/min:19998 - global_seqlen/max:31407 - global_seqlen/minmax_diff:11409 - global_seqlen/balanced_min:24964 - global_seqlen/balanced_max:28950 - global_seqlen/mean:25513.75 - actor/entropy:0.17665337026119232 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.29611214995384216 - training/rollout_probs_diff_mean:0.0035523248370736837 - training/rollout_probs_diff_std:0.0110044926404953 - training/rollout_actor_probs_pearson_corr:0.9986032247543335 - rollout_corr/rollout_is_mean:0.999925434589386 - rollout_corr/rollout_is_ratio_fraction_high:1.739039726089686e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.478079452179372e-05 - rollout_corr/rollout_is_max:2.075227975845337 - rollout_corr/rollout_is_min:0.4160393178462982 - rollout_corr/rollout_is_std:0.03343033045530319 - rollout_corr/rollout_is_eff_sample_size:0.9988834226305476 - rollout_corr/rollout_is_seq_mean:0.9999387264251709 - rollout_corr/rollout_is_seq_std:0.0025514268781989813 - rollout_corr/rollout_is_seq_max:1.0072029829025269 - rollout_corr/rollout_is_seq_min:0.9921553730964661 - rollout_corr/rollout_is_seq_max_deviation:0.007844626903533936 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2722632107324898) - rollout_corr/training_log_ppl:np.float64(0.2361144043825334) - rollout_corr/kl:np.float64(0.0008850332726968446) - rollout_corr/k3_kl:np.float64(0.0009130484113342163) - rollout_corr/rollout_ppl:np.float64(1.2710885656997561) - rollout_corr/rollout_log_ppl:np.float64(0.23522936822337215) - rollout_corr/log_ppl_diff:np.float64(0.0008850361591612455) - rollout_corr/log_ppl_abs_diff:np.float64(0.0021792255029140506) - rollout_corr/log_ppl_diff_max:np.float64(0.012408077716827393) - rollout_corr/log_ppl_diff_min:np.float64(-0.010918736457824707) - rollout_corr/ppl_ratio:np.float64(1.000889461953193) - rollout_corr/chi2_token:np.float64(0.001883210614323616) - rollout_corr/chi2_seq:np.float64(1.0299227663781494) - actor/pg_loss:np.float64(0.00017131585627794266) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015263908690030803) - actor/ppo_kl:np.float64(6.009416163976766e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6240686252713203) - perf/mfu/actor:np.float64(0.06947042905322103) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.28501892089844) - actor/lr:np.float64(1e-06) - training/global_step:38 - training/epoch:2 - critic/score/mean:0.43359375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.43359375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.014406857080757618 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.014406857080757618 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:449.2421875 - response_length/max:6502.0 - response_length/min:86.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:449.2421875 - response_length_non_aborted/max:6502.0 - response_length_non_aborted/min:86.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:348.0625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:6.151758134365082e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.093883166089654) - timing_s/agent_loop/generate_sequences/max:np.float64(34.20732720755041) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.161816134750552) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(34.20732720755041) - timing_s/agent_loop/slowest/tool_calls:np.float6
+step:39 - global_seqlen/min:17054 - global_seqlen/max:28720 - global_seqlen/minmax_diff:11666 - global_seqlen/balanced_min:22500 - global_seqlen/balanced_max:22776 - global_seqlen/mean:22545.375 - actor/entropy:0.19491960108280182 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.45386987924575806 - training/rollout_probs_diff_mean:0.004056902602314949 - training/rollout_probs_diff_std:0.012101990170776844 - training/rollout_actor_probs_pearson_corr:0.9984006881713867 - rollout_corr/rollout_is_mean:0.9999191164970398 - rollout_corr/rollout_is_ratio_fraction_high:1.0010110599978361e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00012012132356176153 - rollout_corr/rollout_is_max:3.0525314807891846 - rollout_corr/rollout_is_min:0.011025836691260338 - rollout_corr/rollout_is_std:0.03583298996090889 - rollout_corr/rollout_is_eff_sample_size:0.9987174054849994 - rollout_corr/rollout_is_seq_mean:0.9999333024024963 - rollout_corr/rollout_is_seq_std:0.002865663729608059 - rollout_corr/rollout_is_seq_max:1.009408712387085 - rollout_corr/rollout_is_seq_min:0.9888485074043274 - rollout_corr/rollout_is_seq_max_deviation:0.011151492595672607 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2850506003014743) - rollout_corr/training_log_ppl:np.float64(0.24625673811533488) - rollout_corr/kl:np.float64(0.001440951160018411) - rollout_corr/k3_kl:np.float64(0.0010581204401205468) - rollout_corr/rollout_ppl:np.float64(1.2831103513017297) - rollout_corr/rollout_log_ppl:np.float64(0.24481578795530368) - rollout_corr/log_ppl_diff:np.float64(0.001440950160031207) - rollout_corr/log_ppl_abs_diff:np.float64(0.002506036587874405) - rollout_corr/log_ppl_diff_max:np.float64(0.015404284000396729) - rollout_corr/log_ppl_diff_min:np.float64(-0.008438989520072937) - rollout_corr/ppl_ratio:np.float64(1.0014470850583166) - rollout_corr/chi2_token:np.float64(0.0012923125177621841) - rollout_corr/chi2_seq:np.float64(0.49474142445251346) - actor/pg_loss:np.float64(8.401612285524607e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0014041442402685789) - actor/ppo_kl:np.float64(0.00043948041806451954) - actor/pg_clipfrac_lower:np.float64(1.9405732018640265e-05) - actor/grad_norm:np.float64(0.6177720576524734) - perf/mfu/actor:np.float64(0.06559690994691629) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.28384017944336) - actor/lr:np.float64(1e-06) - training/global_step:39 - training/epoch:2 - critic/score/mean:0.3828125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.3828125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.004189230967313051 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.004189230967313051 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:390.23046875 - response_length/max:2201.0 - response_length/min:85.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:390.23046875 - response_length_non_aborted/max:2201.0 - response_length_non_aborted/min:85.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:314.3125 - prompt_length/max:498.0 - prompt_length/min:181.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00016164779663085938 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0495860055088997) - timing_s/agent_loop/generate_sequences/max:np.float64(13.142637204378843) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.6693802871523076) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(13.142637204378843) - timing_s/agent_loop
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 40}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score of a protein sequence cannot be determined without access to specific computational tools or databases that calculate such scores based on the sequence. The provided sequence is a string of amino acids, and without additional context or computational analysis, it is not possible to determine the folding stability score. Therefore, the question as posed cannot be answered with the given information.
+</reasoning>
+<answer>
+A
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] A
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_40
+step:40 - global_seqlen/min:15913 - global_seqlen/max:27702 - global_seqlen/minmax_diff:11789 - global_seqlen/balanced_min:21440 - global_seqlen/balanced_max:22074 - global_seqlen/mean:21638.125 - actor/entropy:0.2002803236246109 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.28941571712493896 - training/rollout_probs_diff_mean:0.00437137670814991 - training/rollout_probs_diff_std:0.012427032925188541 - training/rollout_actor_probs_pearson_corr:0.998335063457489 - rollout_corr/rollout_is_mean:1.000082015991211 - rollout_corr/rollout_is_ratio_fraction_high:2.1701152945752256e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.425288327387534e-05 - rollout_corr/rollout_is_max:2.1118814945220947 - rollout_corr/rollout_is_min:0.3157401978969574 - rollout_corr/rollout_is_std:0.037414900958538055 - rollout_corr/rollout_is_eff_sample_size:0.9986024388513428 - rollout_corr/rollout_is_seq_mean:0.9999911785125732 - rollout_corr/rollout_is_seq_std:0.002646325621753931 - rollout_corr/rollout_is_seq_max:1.0076828002929688 - rollout_corr/rollout_is_seq_min:0.99074786901474 - rollout_corr/rollout_is_seq_max_deviation:0.00925213098526001 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2884388542734087) - rollout_corr/training_log_ppl:np.float64(0.24909779269364662) - rollout_corr/kl:np.float64(0.001254979046022564) - rollout_corr/k3_kl:np.float64(0.0012163868613583873) - rollout_corr/rollout_ppl:np.float64(1.2867698702029884) - rollout_corr/rollout_log_ppl:np.float64(0.2478428122412879) - rollout_corr/log_ppl_diff:np.float64(0.0012549804523587227) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025399263540748507) - rollout_corr/log_ppl_diff_max:np.float64(0.013394579291343689) - rollout_corr/log_ppl_diff_min:np.float64(-0.006626993417739868) - rollout_corr/ppl_ratio:np.float64(1.001261416822672) - rollout_corr/chi2_token:np.float64(0.002296414924785495) - rollout_corr/chi2_seq:np.float64(0.5827651822473854) - actor/pg_loss:np.float64(0.00018049252685159445) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001993459543427889) - actor/ppo_kl:np.float64(0.00036094990030921537) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7396467328071594) - perf/mfu/actor:np.float64(0.06226110744290928) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.27344131469727) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.35375) - val-aux/sciknoweval/reward/std@16:np.float64(0.3395231260295747) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.5024000000000001) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.31451708561123753) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.20027999999999999) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.26304832865625566) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.34982000000000013) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.33924744416949965) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.6401399999999999) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.23702579096468043) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.10339999999999999) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.1561668404485443) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.3715800000000001) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.30096314091120385) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.7321000000000001) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.1335033733578523) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.05524) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.07954253083945868) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.38284) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.25173507835666453) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.77508) - val-aux/sciknoweval/reward/best@16/std:np
+step:41 - global_seqlen/min:13619 - global_seqlen/max:27244 - global_seqlen/minmax_diff:13625 - global_seqlen/balanced_min:22065 - global_seqlen/balanced_max:22100 - global_seqlen/mean:22093.0 - actor/entropy:0.1972004473209381 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7945776581764221 - training/rollout_probs_diff_mean:0.0044371881522238255 - training/rollout_probs_diff_std:0.013081462122499943 - training/rollout_actor_probs_pearson_corr:0.9981732368469238 - rollout_corr/rollout_is_mean:1.0000603199005127 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:6.645401299465448e-05 - rollout_corr/rollout_is_max:1.7906994819641113 - rollout_corr/rollout_is_min:0.14579379558563232 - rollout_corr/rollout_is_std:0.037738487124443054 - rollout_corr/rollout_is_eff_sample_size:0.9985780698008809 - rollout_corr/rollout_is_seq_mean:1.0000441074371338 - rollout_corr/rollout_is_seq_std:0.0030684443190693855 - rollout_corr/rollout_is_seq_max:1.007584571838379 - rollout_corr/rollout_is_seq_min:0.9881037473678589 - rollout_corr/rollout_is_seq_max_deviation:0.011896252632141113 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2687957105226815) - rollout_corr/training_log_ppl:np.float64(0.23422167608805466) - rollout_corr/kl:np.float64(0.001997436562771071) - rollout_corr/k3_kl:np.float64(0.001895930330306328) - rollout_corr/rollout_ppl:np.float64(1.2661805939860642) - rollout_corr/rollout_log_ppl:np.float64(0.2322242384107085) - rollout_corr/log_ppl_diff:np.float64(0.00199743767734617) - rollout_corr/log_ppl_abs_diff:np.float64(0.0037411078810691833) - rollout_corr/log_ppl_diff_max:np.float64(0.03718364238739014) - rollout_corr/log_ppl_diff_min:np.float64(-0.030967384576797485) - rollout_corr/ppl_ratio:np.float64(1.002018980216235) - rollout_corr/chi2_token:np.float64(0.0035980751272290945) - rollout_corr/chi2_seq:np.float64(2.2225923924706876) - actor/pg_loss:np.float64(0.0005501225823536515) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0022654881254311476) - actor/ppo_kl:np.float64(0.001161328320471) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7245654910802841) - perf/mfu/actor:np.float64(0.06373080366245223) - perf/max_memory_allocated_gb:np.float64(123.66978645324707) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.36378860473633) - actor/lr:np.float64(1e-06) - training/global_step:41 - training/epoch:2 - critic/score/mean:0.48828125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.48828125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0027744551189243793 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.0027744551189243793 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:352.6875 - response_length/max:1485.0 - response_length/min:79.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:352.6875 - response_length_non_aborted/max:1485.0 - response_length_non_aborted/min:79.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:337.71875 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:2.8796494007110596e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.95299425534904) - timing_s/agent_loop/generate_sequences/max:np.float64(9.864370591938496) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.4211922848553513) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.864370591938496) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent_loop/
+step:42 - global_seqlen/min:11327 - global_seqlen/max:31713 - global_seqlen/minmax_diff:20386 - global_seqlen/balanced_min:16962 - global_seqlen/balanced_max:23405 - global_seqlen/mean:17874.375 - actor/entropy:0.1793491095304489 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2369065284729004 - training/rollout_probs_diff_mean:0.003962434828281403 - training/rollout_probs_diff_std:0.011616604402661324 - training/rollout_actor_probs_pearson_corr:0.9984515309333801 - rollout_corr/rollout_is_mean:1.0001260042190552 - rollout_corr/rollout_is_ratio_fraction_high:1.2704059372481424e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0 - rollout_corr/rollout_is_max:3.3973875045776367 - rollout_corr/rollout_is_min:0.5620211958885193 - rollout_corr/rollout_is_std:0.03498118743300438 - rollout_corr/rollout_is_eff_sample_size:0.9987781689679683 - rollout_corr/rollout_is_seq_mean:1.0004432201385498 - rollout_corr/rollout_is_seq_std:0.003099439898505807 - rollout_corr/rollout_is_seq_max:1.010513424873352 - rollout_corr/rollout_is_seq_min:0.9920833706855774 - rollout_corr/rollout_is_seq_max_deviation:0.01051342487335205 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2783533367328346) - rollout_corr/training_log_ppl:np.float64(0.24186358234874206) - rollout_corr/kl:np.float64(0.0008186687481881449) - rollout_corr/k3_kl:np.float64(0.0014607259122385585) - rollout_corr/rollout_ppl:np.float64(1.2772352620959282) - rollout_corr/rollout_log_ppl:np.float64(0.24104491147591034) - rollout_corr/log_ppl_diff:np.float64(0.0008186708728317171) - rollout_corr/log_ppl_abs_diff:np.float64(0.0031197732605505735) - rollout_corr/log_ppl_diff_max:np.float64(0.023270994424819946) - rollout_corr/log_ppl_diff_min:np.float64(-0.01981446146965027) - rollout_corr/ppl_ratio:np.float64(1.0008287110831589) - rollout_corr/chi2_token:np.float64(0.004162837052717805) - rollout_corr/chi2_seq:np.float64(0.8917187473271042) - actor/pg_loss:np.float64(0.000158449518494308) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011291635195220806) - actor/ppo_kl:np.float64(0.0004214342557844475) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5698883533477783) - perf/mfu/actor:np.float64(0.0494323141668247) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.41558456420898) - actor/lr:np.float64(1e-06) - training/global_step:42 - training/epoch:2 - critic/score/mean:0.328125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.328125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.03640824556350708 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.03640824556350708 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:307.48046875 - response_length/max:8192.0 - response_length/min:79.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:307.48046875 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:79.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:251.09375 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.000103706493973732 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9703596197068691) - timing_s/agent_loop/generate_sequences/max:np.float64(41.32044994644821) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.750317184647429) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(41.32044994644821) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) -
+step:43 - global_seqlen/min:14105 - global_seqlen/max:31969 - global_seqlen/minmax_diff:17864 - global_seqlen/balanced_min:21216 - global_seqlen/balanced_max:21234 - global_seqlen/mean:21226.625 - actor/entropy:0.19266550242900848 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.44984614849090576 - training/rollout_probs_diff_mean:0.004387952387332916 - training/rollout_probs_diff_std:0.01266084611415863 - training/rollout_actor_probs_pearson_corr:0.9982303977012634 - rollout_corr/rollout_is_mean:0.9999305605888367 - rollout_corr/rollout_is_ratio_fraction_high:3.408587508602068e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.544783223536797e-05 - rollout_corr/rollout_is_max:2.5337352752685547 - rollout_corr/rollout_is_min:0.32041898369789124 - rollout_corr/rollout_is_std:0.038009174168109894 - rollout_corr/rollout_is_eff_sample_size:0.9985570301628549 - rollout_corr/rollout_is_seq_mean:0.9998732805252075 - rollout_corr/rollout_is_seq_std:0.002512736711651087 - rollout_corr/rollout_is_seq_max:1.008451223373413 - rollout_corr/rollout_is_seq_min:0.9910997748374939 - rollout_corr/rollout_is_seq_max_deviation:0.008900225162506104 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2575704725459218) - rollout_corr/training_log_ppl:np.float64(0.22557678240991663) - rollout_corr/kl:np.float64(0.0013313335243312352) - rollout_corr/k3_kl:np.float64(0.0010828630935861838) - rollout_corr/rollout_ppl:np.float64(1.2558402898721397) - rollout_corr/rollout_log_ppl:np.float64(0.22424544939713087) - rollout_corr/log_ppl_diff:np.float64(0.0013313330127857625) - rollout_corr/log_ppl_abs_diff:np.float64(0.0023614601814188063) - rollout_corr/log_ppl_diff_max:np.float64(0.011232882738113403) - rollout_corr/log_ppl_diff_min:np.float64(-0.007203608751296997) - rollout_corr/ppl_ratio:np.float64(1.0013366700150073) - rollout_corr/chi2_token:np.float64(0.0016005998477339745) - rollout_corr/chi2_seq:np.float64(0.9111351130995899) - actor/pg_loss:np.float64(-0.00032677757553756237) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0019711919426299573) - actor/ppo_kl:np.float64(0.00035553486237560605) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6708932965993881) - perf/mfu/actor:np.float64(0.06152323141502927) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.48104095458984) - actor/lr:np.float64(1e-06) - training/global_step:43 - training/epoch:3 - critic/score/mean:0.44140625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.44140625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0011432970641180873 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0011432970641180873 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:343.80078125 - response_length/max:1356.0 - response_length/min:83.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:343.80078125 - response_length_non_aborted/max:1356.0 - response_length_non_aborted/min:83.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:319.53125 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0002473127096891403 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0578165259212255) - timing_s/agent_loop/generate_sequences/max:np.float64(9.275035005062819) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.464938813202025) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.275035005062819) - timing_s/agent_loop/slowest/tool_c
+step:44 - global_seqlen/min:14289 - global_seqlen/max:31868 - global_seqlen/minmax_diff:17579 - global_seqlen/balanced_min:23024 - global_seqlen/balanced_max:27562 - global_seqlen/mean:23664.625 - actor/entropy:0.16483263671398163 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5307222604751587 - training/rollout_probs_diff_mean:0.0037482562474906445 - training/rollout_probs_diff_std:0.011952490545809269 - training/rollout_actor_probs_pearson_corr:0.9982149600982666 - rollout_corr/rollout_is_mean:0.9997656941413879 - rollout_corr/rollout_is_ratio_fraction_high:1.877493559732102e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.754987119464204e-05 - rollout_corr/rollout_is_max:2.257253646850586 - rollout_corr/rollout_is_min:0.2560407221317291 - rollout_corr/rollout_is_std:0.03545422852039337 - rollout_corr/rollout_is_eff_sample_size:0.9987439217121525 - rollout_corr/rollout_is_seq_mean:0.9996024966239929 - rollout_corr/rollout_is_seq_std:0.002828868106007576 - rollout_corr/rollout_is_seq_max:1.0087915658950806 - rollout_corr/rollout_is_seq_min:0.9925878047943115 - rollout_corr/rollout_is_seq_max_deviation:0.008791565895080566 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2643110915087163) - rollout_corr/training_log_ppl:np.float64(0.23016513518814463) - rollout_corr/kl:np.float64(0.0011426463696224687) - rollout_corr/k3_kl:np.float64(0.0011375640360711259) - rollout_corr/rollout_ppl:np.float64(1.2628552531823516) - rollout_corr/rollout_log_ppl:np.float64(0.22902248779064394) - rollout_corr/log_ppl_diff:np.float64(0.0011426473975006957) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027303947317705024) - rollout_corr/log_ppl_diff_max:np.float64(0.012277275323867798) - rollout_corr/log_ppl_diff_min:np.float64(-0.010143131017684937) - rollout_corr/ppl_ratio:np.float64(1.0011494110804051) - rollout_corr/chi2_token:np.float64(0.0023959616664797068) - rollout_corr/chi2_seq:np.float64(1.1097935638390481) - actor/pg_loss:np.float64(0.0004005907103419304) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0017527463992337289) - actor/ppo_kl:np.float64(-0.0001501098105372023) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6244964599609375) - perf/mfu/actor:np.float64(0.06486412678751474) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.46553421020508) - actor/lr:np.float64(1e-06) - training/global_step:44 - training/epoch:3 - critic/score/mean:0.4375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.03883712738752365 - critic/advantages/max:0.875 - critic/advantages/min:-0.75 - critic/returns/mean:-0.03883712738752365 - critic/returns/max:0.875 - critic/returns/min:-0.75 - response_length/mean:416.11328125 - response_length/max:7121.0 - response_length/min:81.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:416.11328125 - response_length_non_aborted/max:7121.0 - response_length_non_aborted/min:81.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:323.40625 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:3.310292959213257e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.854373425245285) - timing_s/agent_loop/generate_sequences/max:np.float64(35.90070942789316) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.641139167644724) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(35.90070942789316) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.
+step:45 - global_seqlen/min:10341 - global_seqlen/max:43510 - global_seqlen/minmax_diff:33169 - global_seqlen/balanced_min:21899 - global_seqlen/balanced_max:27874 - global_seqlen/mean:22856.25 - actor/entropy:0.1471649408340454 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3509632647037506 - training/rollout_probs_diff_mean:0.0035214414820075035 - training/rollout_probs_diff_std:0.01150407176464796 - training/rollout_actor_probs_pearson_corr:0.9982120394706726 - rollout_corr/rollout_is_mean:0.9999986290931702 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:6.89342778059654e-05 - rollout_corr/rollout_is_max:1.867427945137024 - rollout_corr/rollout_is_min:0.030966857448220253 - rollout_corr/rollout_is_std:0.03498629853129387 - rollout_corr/rollout_is_eff_sample_size:0.9987774554595162 - rollout_corr/rollout_is_seq_mean:0.9998350739479065 - rollout_corr/rollout_is_seq_std:0.002756875241175294 - rollout_corr/rollout_is_seq_max:1.0082145929336548 - rollout_corr/rollout_is_seq_min:0.990932822227478 - rollout_corr/rollout_is_seq_max_deviation:0.009067177772521973 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2405082695186138) - rollout_corr/training_log_ppl:np.float64(0.21127635941229528) - rollout_corr/kl:np.float64(0.0012804620336837047) - rollout_corr/k3_kl:np.float64(0.0011797188574576012) - rollout_corr/rollout_ppl:np.float64(1.2388938385993242) - rollout_corr/rollout_log_ppl:np.float64(0.20999589761777315) - rollout_corr/log_ppl_diff:np.float64(0.0012804617945221253) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025750486165634356) - rollout_corr/log_ppl_diff_max:np.float64(0.01426708698272705) - rollout_corr/log_ppl_diff_min:np.float64(-0.016405731439590454) - rollout_corr/ppl_ratio:np.float64(1.001287407707423) - rollout_corr/chi2_token:np.float64(0.0022307620383799076) - rollout_corr/chi2_seq:np.float64(0.5359919595066458) - actor/pg_loss:np.float64(8.05326271802187e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001362902614914674) - actor/ppo_kl:np.float64(0.0002342196375835215) - actor/pg_clipfrac_lower:np.float64(7.356402875302592e-06) - actor/grad_norm:np.float64(0.6136828362941742) - perf/mfu/actor:np.float64(0.062255809404529946) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.4916729927063) - actor/lr:np.float64(1e-06) - training/global_step:45 - training/epoch:3 - critic/score/mean:0.31640625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.31640625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0037901541218161583 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0037901541218161583 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:396.6640625 - response_length/max:8160.0 - response_length/min:76.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:396.6640625 - response_length_non_aborted/max:8160.0 - response_length_non_aborted/min:76.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:317.59375 - prompt_length/max:498.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:7.038004696369171e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9433151036500931) - timing_s/agent_loop/generate_sequences/max:np.float64(41.46591006591916) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.561986893051653) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(41.46591006591916) - timing_s/agent_loop/slowest/tool_calls:np.float
+step:46 - global_seqlen/min:11629 - global_seqlen/max:28235 - global_seqlen/minmax_diff:16606 - global_seqlen/balanced_min:17711 - global_seqlen/balanced_max:17919 - global_seqlen/mean:17760.75 - actor/entropy:0.20734380185604095 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5599408149719238 - training/rollout_probs_diff_mean:0.004730684217065573 - training/rollout_probs_diff_std:0.013396089896559715 - training/rollout_actor_probs_pearson_corr:0.9980821013450623 - rollout_corr/rollout_is_mean:1.0000985860824585 - rollout_corr/rollout_is_ratio_fraction_high:2.7359032173990272e-05 - rollout_corr/rollout_is_ratio_fraction_low:1.3679516086995136e-05 - rollout_corr/rollout_is_max:5.697536468505859 - rollout_corr/rollout_is_min:0.4908820390701294 - rollout_corr/rollout_is_std:0.04028098285198212 - rollout_corr/rollout_is_eff_sample_size:0.9983800707884429 - rollout_corr/rollout_is_seq_mean:1.000237226486206 - rollout_corr/rollout_is_seq_std:0.00363980233669281 - rollout_corr/rollout_is_seq_max:1.021539568901062 - rollout_corr/rollout_is_seq_min:0.9916234016418457 - rollout_corr/rollout_is_seq_max_deviation:0.02153956890106201 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2810124992392957) - rollout_corr/training_log_ppl:np.float64(0.24381673078460153) - rollout_corr/kl:np.float64(0.0006558897876516312) - rollout_corr/k3_kl:np.float64(0.0010635649665289293) - rollout_corr/rollout_ppl:np.float64(1.2801448996178806) - rollout_corr/rollout_log_ppl:np.float64(0.24316083974554203) - rollout_corr/log_ppl_diff:np.float64(0.0006558910390594974) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025359629216836765) - rollout_corr/log_ppl_diff_max:np.float64(0.009192198514938354) - rollout_corr/log_ppl_diff_min:np.float64(-0.009751826524734497) - rollout_corr/ppl_ratio:np.float64(1.0006613184232265) - rollout_corr/chi2_token:np.float64(0.003016318893060088) - rollout_corr/chi2_seq:np.float64(0.5328007256612182) - actor/pg_loss:np.float64(-6.973580457270145e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015097170539775107) - actor/ppo_kl:np.float64(-0.00010064278021815198) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.563657782971859) - perf/mfu/actor:np.float64(0.051583663196279474) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.5073127746582) - actor/lr:np.float64(1e-06) - training/global_step:46 - training/epoch:3 - critic/score/mean:0.46484375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.46484375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.023010656237602234 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.023010656237602234 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:285.5546875 - response_length/max:1823.0 - response_length/min:76.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:285.5546875 - response_length_non_aborted/max:1823.0 - response_length_non_aborted/min:76.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:269.46875 - prompt_length/max:498.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.830668568611145e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9158236924558878) - timing_s/agent_loop/generate_sequences/max:np.float64(10.958448996767402) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.764639292836364) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(10.958448996767402) - timing_s/agent_loop/slowest/tool_calls:np.f
+step:47 - global_seqlen/min:18567 - global_seqlen/max:38675 - global_seqlen/minmax_diff:20108 - global_seqlen/balanced_min:25878 - global_seqlen/balanced_max:26514 - global_seqlen/mean:25983.75 - actor/entropy:0.14283302426338196 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7308436632156372 - training/rollout_probs_diff_mean:0.003343928838148713 - training/rollout_probs_diff_std:0.011697083711624146 - training/rollout_actor_probs_pearson_corr:0.9980596899986267 - rollout_corr/rollout_is_mean:1.0000288486480713 - rollout_corr/rollout_is_ratio_fraction_high:2.6085595891345292e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0001130375821958296 - rollout_corr/rollout_is_max:2.327317714691162 - rollout_corr/rollout_is_min:0.23235759139060974 - rollout_corr/rollout_is_std:0.03357976675033569 - rollout_corr/rollout_is_eff_sample_size:0.9988737883222161 - rollout_corr/rollout_is_seq_mean:0.9999798536300659 - rollout_corr/rollout_is_seq_std:0.0024428016040474176 - rollout_corr/rollout_is_seq_max:1.0096776485443115 - rollout_corr/rollout_is_seq_min:0.9921210408210754 - rollout_corr/rollout_is_seq_max_deviation:0.009677648544311523 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2100329943932593) - rollout_corr/training_log_ppl:np.float64(0.18659601080435095) - rollout_corr/kl:np.float64(0.0011015117767527727) - rollout_corr/k3_kl:np.float64(0.0009561012871870389) - rollout_corr/rollout_ppl:np.float64(1.208669331856072) - rollout_corr/rollout_log_ppl:np.float64(0.18549449724378064) - rollout_corr/log_ppl_diff:np.float64(0.0011015135605703108) - rollout_corr/log_ppl_abs_diff:np.float64(0.0020468592338147573) - rollout_corr/log_ppl_diff_max:np.float64(0.013909026980400085) - rollout_corr/log_ppl_diff_min:np.float64(-0.007575169205665588) - rollout_corr/ppl_ratio:np.float64(1.001105838920921) - rollout_corr/chi2_token:np.float64(0.001610573148354888) - rollout_corr/chi2_seq:np.float64(0.26548866275697947) - actor/pg_loss:np.float64(4.338508006185293e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011870542070937518) - actor/ppo_kl:np.float64(0.0002975921167021056) - actor/pg_clipfrac_lower:np.float64(9.322792720922735e-06) - actor/grad_norm:np.float64(0.5477600991725922) - perf/mfu/actor:np.float64(0.07517353223146761) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.55639266967773) - actor/lr:np.float64(1e-06) - training/global_step:47 - training/epoch:3 - critic/score/mean:0.39453125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.39453125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0002075978700304404 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0002075978700304404 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:449.2421875 - response_length/max:2732.0 - response_length/min:78.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:449.2421875 - response_length_non_aborted/max:2732.0 - response_length_non_aborted/min:78.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:362.75 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:2.532266080379486e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9549492709338665) - timing_s/agent_loop/generate_sequences/max:np.float64(15.872619692236185) - timing_s/agent_loop/generate_sequences/mean:np.float64(4.19089348961279) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(15.872619692236185) - timing_s/agent_loop/slowe
+step:48 - global_seqlen/min:16381 - global_seqlen/max:32235 - global_seqlen/minmax_diff:15854 - global_seqlen/balanced_min:21840 - global_seqlen/balanced_max:27905 - global_seqlen/mean:22746.0 - actor/entropy:0.16609838604927063 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5618129372596741 - training/rollout_probs_diff_mean:0.0038480223156511784 - training/rollout_probs_diff_std:0.012091293931007385 - training/rollout_actor_probs_pearson_corr:0.9981785416603088 - rollout_corr/rollout_is_mean:1.0000433921813965 - rollout_corr/rollout_is_ratio_fraction_high:1.9581733795348555e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.8745201386045665e-05 - rollout_corr/rollout_is_max:2.1106479167938232 - rollout_corr/rollout_is_min:0.0005965463351458311 - rollout_corr/rollout_is_std:0.03607833385467529 - rollout_corr/rollout_is_eff_sample_size:0.9987001647473916 - rollout_corr/rollout_is_seq_mean:1.0000697374343872 - rollout_corr/rollout_is_seq_std:0.0026138562243431807 - rollout_corr/rollout_is_seq_max:1.010244369506836 - rollout_corr/rollout_is_seq_min:0.99269038438797 - rollout_corr/rollout_is_seq_max_deviation:0.010244369506835938 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.260544263292104) - rollout_corr/training_log_ppl:np.float64(0.2275718234195665) - rollout_corr/kl:np.float64(0.000840849202081273) - rollout_corr/k3_kl:np.float64(0.0011277189190366244) - rollout_corr/rollout_ppl:np.float64(1.2594485664740205) - rollout_corr/rollout_log_ppl:np.float64(0.2267309735470917) - rollout_corr/log_ppl_diff:np.float64(0.0008408498724747915) - rollout_corr/log_ppl_abs_diff:np.float64(0.002264020313305082) - rollout_corr/log_ppl_diff_max:np.float64(0.02813515067100525) - rollout_corr/log_ppl_diff_min:np.float64(-0.008843004703521729) - rollout_corr/ppl_ratio:np.float64(1.0008466544095427) - rollout_corr/chi2_token:np.float64(0.0026712415274232626) - rollout_corr/chi2_seq:np.float64(0.9579408683348447) - actor/pg_loss:np.float64(1.6385456547141075e-06) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.00142442833248424) - actor/ppo_kl:np.float64(-7.805178177022754e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5536335557699203) - perf/mfu/actor:np.float64(0.06172036956944962) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.46680450439453) - actor/lr:np.float64(1e-06) - training/global_step:48 - training/epoch:3 - critic/score/mean:0.35546875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.35546875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0621769018471241 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0621769018471241 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:398.96875 - response_length/max:8192.0 - response_length/min:81.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:398.96875 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:81.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:311.84375 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.11615788936615e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.5817424338310957) - timing_s/agent_loop/generate_sequences/max:np.float64(47.18641279824078) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.3139364701564773) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(47.18641279824078) - timing_s/agent_loop/slowest/tool_calls:n
+step:49 - global_seqlen/min:13330 - global_seqlen/max:23649 - global_seqlen/minmax_diff:10319 - global_seqlen/balanced_min:19645 - global_seqlen/balanced_max:19656 - global_seqlen/mean:19652.75 - actor/entropy:0.21312448382377625 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.29631492495536804 - training/rollout_probs_diff_mean:0.004599376115947962 - training/rollout_probs_diff_std:0.012834436260163784 - training/rollout_actor_probs_pearson_corr:0.9982872009277344 - rollout_corr/rollout_is_mean:0.9999762773513794 - rollout_corr/rollout_is_ratio_fraction_high:1.2280485861992929e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.684145849547349e-05 - rollout_corr/rollout_is_max:2.1531147956848145 - rollout_corr/rollout_is_min:0.16506630182266235 - rollout_corr/rollout_is_std:0.03860819712281227 - rollout_corr/rollout_is_eff_sample_size:0.9985116256038487 - rollout_corr/rollout_is_seq_mean:0.999916136264801 - rollout_corr/rollout_is_seq_std:0.002757440321147442 - rollout_corr/rollout_is_seq_max:1.011631965637207 - rollout_corr/rollout_is_seq_min:0.9919342398643494 - rollout_corr/rollout_is_seq_max_deviation:0.011631965637207031 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.290575869847089) - rollout_corr/training_log_ppl:np.float64(0.2508397557976423) - rollout_corr/kl:np.float64(0.0012681070296984842) - rollout_corr/k3_kl:np.float64(0.0010787070739013416) - rollout_corr/rollout_ppl:np.float64(1.2888763579539955) - rollout_corr/rollout_log_ppl:np.float64(0.24957164909574203) - rollout_corr/log_ppl_diff:np.float64(0.0012681067019002512) - rollout_corr/log_ppl_abs_diff:np.float64(0.002345262750168331) - rollout_corr/log_ppl_diff_max:np.float64(0.009302854537963867) - rollout_corr/log_ppl_diff_min:np.float64(-0.009885609149932861) - rollout_corr/ppl_ratio:np.float64(1.0012729486916214) - rollout_corr/chi2_token:np.float64(0.0018082375172525644) - rollout_corr/chi2_seq:np.float64(0.36839555972255766) - actor/pg_loss:np.float64(4.909001290798187e-06) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0016076876192983036) - actor/ppo_kl:np.float64(0.0002880832484706275) - actor/pg_clipfrac_lower:np.float64(4.828491910302546e-06) - actor/grad_norm:np.float64(0.629483163356781) - perf/mfu/actor:np.float64(0.05734176798977645) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.61180114746094) - actor/lr:np.float64(1e-06) - training/global_step:49 - training/epoch:3 - critic/score/mean:0.34765625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.34765625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0032481886446475983 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.0032481886446475983 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:318.0859375 - response_length/max:1276.0 - response_length/min:76.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:318.0859375 - response_length_non_aborted/max:1276.0 - response_length_non_aborted/min:76.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:296.0625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:8.732825517654419e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.8680097497999668) - timing_s/agent_loop/generate_sequences/max:np.float64(8.407540844753385) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.0850135255313944) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(8.407540844753385) - timing_s/agent_loop/slowest
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 50}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score of a protein sequence typically requires computational analysis using tools like I-TASSER, ROSETTA, or other protein structure prediction and stability analysis software. These tools calculate scores based on the predicted 3D structure of the protein, taking into account factors such as secondary structure, solvent accessibility, and other structural properties. Without access to specific computational tools or databases that provide folding stability scores for the given sequence, it is not possible to determine the exact folding stability score for the sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'. Therefore, the score cannot be determined from the information provided.
+</reasoning>
+<answer>
+D
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] D
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_50
+step:50 - global_seqlen/min:12100 - global_seqlen/max:26431 - global_seqlen/minmax_diff:14331 - global_seqlen/balanced_min:22098 - global_seqlen/balanced_max:22581 - global_seqlen/mean:22179.5 - actor/entropy:0.1872798055410385 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.25002074241638184 - training/rollout_probs_diff_mean:0.0043448833748698235 - training/rollout_probs_diff_std:0.012745512649416924 - training/rollout_actor_probs_pearson_corr:0.9981958270072937 - rollout_corr/rollout_is_mean:0.9999893307685852 - rollout_corr/rollout_is_ratio_fraction_high:1.082672861230094e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.248018765589222e-05 - rollout_corr/rollout_is_max:2.2042269706726074 - rollout_corr/rollout_is_min:0.3166545629501343 - rollout_corr/rollout_is_std:0.03807654604315758 - rollout_corr/rollout_is_eff_sample_size:0.9985522755652161 - rollout_corr/rollout_is_seq_mean:0.9999541640281677 - rollout_corr/rollout_is_seq_std:0.0029842411167919636 - rollout_corr/rollout_is_seq_max:1.0160789489746094 - rollout_corr/rollout_is_seq_min:0.986756443977356 - rollout_corr/rollout_is_seq_max_deviation:0.016078948974609375 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.264865230768919) - rollout_corr/training_log_ppl:np.float64(0.2308657732064603) - rollout_corr/kl:np.float64(0.001112062272518699) - rollout_corr/k3_kl:np.float64(0.0010981027096477192) - rollout_corr/rollout_ppl:np.float64(1.2634586868807673) - rollout_corr/rollout_log_ppl:np.float64(0.22975370922358707) - rollout_corr/log_ppl_diff:np.float64(0.0011120639828732237) - rollout_corr/log_ppl_abs_diff:np.float64(0.002492617248208262) - rollout_corr/log_ppl_diff_max:np.float64(0.01720789074897766) - rollout_corr/log_ppl_diff_min:np.float64(-0.013499855995178223) - rollout_corr/ppl_ratio:np.float64(1.001118621090427) - rollout_corr/chi2_token:np.float64(0.002215159824118018) - rollout_corr/chi2_seq:np.float64(0.9738334778230637) - actor/pg_loss:np.float64(0.00040720601100474596) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.00143988847764831) - actor/ppo_kl:np.float64(0.00017129167446228166) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5727251768112183) - perf/mfu/actor:np.float64(0.06430968389502065) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.6324348449707) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.41875) - val-aux/sciknoweval/reward/std@16:np.float64(0.22869494240339183) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.51456) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.18468184508936822) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.31976) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.20223553757769963) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.41681999999999997) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.22791020438572854) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.58694) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.1261578566460394) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.23558) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.15226752550720532) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.4398) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.18398851891396895) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.63446) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.07544190394601284) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.17675999999999997) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.09854951447854242) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.45147999999999994) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.13395224221738825) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.66118) - val-aux/sciknoweval/reward/best@16/std:np.float64(0.03956012550827376) - val-aux/scikn
+step:51 - global_seqlen/min:16904 - global_seqlen/max:25515 - global_seqlen/minmax_diff:8611 - global_seqlen/balanced_min:21289 - global_seqlen/balanced_max:21389 - global_seqlen/mean:21315.375 - actor/entropy:0.19681613147258759 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.6418529748916626 - training/rollout_probs_diff_mean:0.004624536260962486 - training/rollout_probs_diff_std:0.013267040252685547 - training/rollout_actor_probs_pearson_corr:0.9980368614196777 - rollout_corr/rollout_is_mean:0.9997654557228088 - rollout_corr/rollout_is_ratio_fraction_high:2.3248514480656013e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0 - rollout_corr/rollout_is_max:3.855645179748535 - rollout_corr/rollout_is_min:0.5350923538208008 - rollout_corr/rollout_is_std:0.03936876356601715 - rollout_corr/rollout_is_eff_sample_size:0.9984516075893486 - rollout_corr/rollout_is_seq_mean:0.9997847676277161 - rollout_corr/rollout_is_seq_std:0.0029975101351737976 - rollout_corr/rollout_is_seq_max:1.0126070976257324 - rollout_corr/rollout_is_seq_min:0.9907597303390503 - rollout_corr/rollout_is_seq_max_deviation:0.012607097625732422 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2693150429986417) - rollout_corr/training_log_ppl:np.float64(0.2345841708010994) - rollout_corr/kl:np.float64(0.0011821451473537081) - rollout_corr/k3_kl:np.float64(0.0013152157603144587) - rollout_corr/rollout_ppl:np.float64(1.267772528808564) - rollout_corr/rollout_log_ppl:np.float64(0.23340202278632205) - rollout_corr/log_ppl_diff:np.float64(0.00118214801477734) - rollout_corr/log_ppl_abs_diff:np.float64(0.002665369029273279) - rollout_corr/log_ppl_diff_max:np.float64(0.010928303003311157) - rollout_corr/log_ppl_diff_min:np.float64(-0.00843799114227295) - rollout_corr/ppl_ratio:np.float64(1.0011883925180882) - rollout_corr/chi2_token:np.float64(0.0032832033466547728) - rollout_corr/chi2_seq:np.float64(2.628566562430933) - actor/pg_loss:np.float64(0.00014865747652947903) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0017586877143003221) - actor/ppo_kl:np.float64(2.5402906045357554e-06) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5996150523424149) - perf/mfu/actor:np.float64(0.061039498769682454) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.6258544921875) - actor/lr:np.float64(1e-06) - training/global_step:51 - training/epoch:3 - critic/score/mean:0.4140625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4140625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0019761237781494856 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.0019761237781494856 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:336.04296875 - response_length/max:1579.0 - response_length/min:92.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:336.04296875 - response_length_non_aborted/max:1579.0 - response_length_non_aborted/min:92.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:330.0625 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00010008178651332855 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1094308514147997) - timing_s/agent_loop/generate_sequences/max:np.float64(9.863438200205564) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.3283677981307846) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.863438200205564) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_
+step:52 - global_seqlen/min:12379 - global_seqlen/max:22705 - global_seqlen/minmax_diff:10326 - global_seqlen/balanced_min:16786 - global_seqlen/balanced_max:16849 - global_seqlen/mean:16796.75 - actor/entropy:0.22803975641727448 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.36542510986328125 - training/rollout_probs_diff_mean:0.0053499056957662106 - training/rollout_probs_diff_std:0.01391876582056284 - training/rollout_actor_probs_pearson_corr:0.9980417490005493 - rollout_corr/rollout_is_mean:1.0001552104949951 - rollout_corr/rollout_is_ratio_fraction_high:4.7668985644122586e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.355864752549678e-05 - rollout_corr/rollout_is_max:2.923143148422241 - rollout_corr/rollout_is_min:0.39950257539749146 - rollout_corr/rollout_is_std:0.04149097949266434 - rollout_corr/rollout_is_eff_sample_size:0.9982820511233786 - rollout_corr/rollout_is_seq_mean:1.0000829696655273 - rollout_corr/rollout_is_seq_std:0.003001436125487089 - rollout_corr/rollout_is_seq_max:1.011511206626892 - rollout_corr/rollout_is_seq_min:0.990175724029541 - rollout_corr/rollout_is_seq_max_deviation:0.01151120662689209 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2837047763168812) - rollout_corr/training_log_ppl:np.float64(0.24606362084159628) - rollout_corr/kl:np.float64(0.0014448560898543406) - rollout_corr/k3_kl:np.float64(0.0011887111170381104) - rollout_corr/rollout_ppl:np.float64(1.281900660134852) - rollout_corr/rollout_log_ppl:np.float64(0.24461876275017858) - rollout_corr/log_ppl_diff:np.float64(0.0014448580914177) - rollout_corr/log_ppl_abs_diff:np.float64(0.0029126321896910667) - rollout_corr/log_ppl_diff_max:np.float64(0.013406842947006226) - rollout_corr/log_ppl_diff_min:np.float64(-0.008877873420715332) - rollout_corr/ppl_ratio:np.float64(1.001452179858461) - rollout_corr/chi2_token:np.float64(0.0019569701980799437) - rollout_corr/chi2_seq:np.float64(0.4788138067815453) - actor/pg_loss:np.float64(0.00014281971380114555) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0018395553652226226) - actor/ppo_kl:np.float64(0.0005606412955785345) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6134368255734444) - perf/mfu/actor:np.float64(0.047977674968366274) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.7822380065918) - actor/lr:np.float64(1e-06) - training/global_step:52 - training/epoch:3 - critic/score/mean:0.42578125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.42578125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.01040375605225563 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.01040375605225563 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:245.8359375 - response_length/max:1193.0 - response_length/min:82.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:245.8359375 - response_length_non_aborted/max:1193.0 - response_length_non_aborted/min:82.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:279.0625 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:3.267452120780945e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9509514290839434) - timing_s/agent_loop/generate_sequences/max:np.float64(7.854211928322911) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.525260150054237) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.854211928322911) - timing_s/agent_loop/slowest/tool_calls:np.float64(0
+step:53 - global_seqlen/min:15429 - global_seqlen/max:26564 - global_seqlen/minmax_diff:11135 - global_seqlen/balanced_min:20620 - global_seqlen/balanced_max:20653 - global_seqlen/mean:20644.25 - actor/entropy:0.20042356848716736 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3535129427909851 - training/rollout_probs_diff_mean:0.0044761584140360355 - training/rollout_probs_diff_std:0.012780102901160717 - training/rollout_actor_probs_pearson_corr:0.9982099533081055 - rollout_corr/rollout_is_mean:0.9998370409011841 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:0.00012490943481680006 - rollout_corr/rollout_is_max:1.8165900707244873 - rollout_corr/rollout_is_min:0.2722632586956024 - rollout_corr/rollout_is_std:0.03822029009461403 - rollout_corr/rollout_is_eff_sample_size:0.9985411024389872 - rollout_corr/rollout_is_seq_mean:0.9997756481170654 - rollout_corr/rollout_is_seq_std:0.0028406018391251564 - rollout_corr/rollout_is_seq_max:1.0108007192611694 - rollout_corr/rollout_is_seq_min:0.9895545840263367 - rollout_corr/rollout_is_seq_max_deviation:0.010800719261169434 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2743919994682074) - rollout_corr/training_log_ppl:np.float64(0.237915026533301) - rollout_corr/kl:np.float64(0.0014105199338541041) - rollout_corr/k3_kl:np.float64(0.0010296943997332164) - rollout_corr/rollout_ppl:np.float64(1.2725408389233053) - rollout_corr/rollout_log_ppl:np.float64(0.23650450706190895) - rollout_corr/log_ppl_diff:np.float64(0.0014105194713920355) - rollout_corr/log_ppl_abs_diff:np.float64(0.0024846792221069336) - rollout_corr/log_ppl_diff_max:np.float64(0.014649003744125366) - rollout_corr/log_ppl_diff_min:np.float64(-0.0049605220556259155) - rollout_corr/ppl_ratio:np.float64(1.0014160515274853) - rollout_corr/chi2_token:np.float64(0.0013009211979806423) - rollout_corr/chi2_seq:np.float64(0.20528000919148326) - actor/pg_loss:np.float64(0.00033905566670000553) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011068140559018502) - actor/ppo_kl:np.float64(0.00029123531216690424) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.495539128780365) - perf/mfu/actor:np.float64(0.059122989206736365) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.70719528198242) - actor/lr:np.float64(1e-06) - training/global_step:53 - training/epoch:3 - critic/score/mean:0.51953125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.51953125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.01337155606597662 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.01337155606597662 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:312.7265625 - response_length/max:1431.0 - response_length/min:81.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:312.7265625 - response_length_non_aborted/max:1431.0 - response_length_non_aborted/min:81.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:332.40625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:5.622953176498413e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.7730677053332329) - timing_s/agent_loop/generate_sequences/max:np.float64(9.259236015379429) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.0541376218679943) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.259236015379429) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) -
+step:54 - global_seqlen/min:13022 - global_seqlen/max:20901 - global_seqlen/minmax_diff:7879 - global_seqlen/balanced_min:17654 - global_seqlen/balanced_max:19215 - global_seqlen/mean:17867.875 - actor/entropy:0.21646106243133545 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5828086733818054 - training/rollout_probs_diff_mean:0.004788811318576336 - training/rollout_probs_diff_std:0.013526316732168198 - training/rollout_actor_probs_pearson_corr:0.9981729388237 - rollout_corr/rollout_is_mean:0.9998193979263306 - rollout_corr/rollout_is_ratio_fraction_high:1.4932727935956791e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00010452909918967634 - rollout_corr/rollout_is_max:2.679741144180298 - rollout_corr/rollout_is_min:0.20122860372066498 - rollout_corr/rollout_is_std:0.040283203125 - rollout_corr/rollout_is_eff_sample_size:0.9983792390254678 - rollout_corr/rollout_is_seq_mean:0.9996543526649475 - rollout_corr/rollout_is_seq_std:0.0032862715888768435 - rollout_corr/rollout_is_seq_max:1.0084000825881958 - rollout_corr/rollout_is_seq_min:0.9903689622879028 - rollout_corr/rollout_is_seq_max_deviation:0.009631037712097168 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2988018337637186) - rollout_corr/training_log_ppl:np.float64(0.257318063187995) - rollout_corr/kl:np.float64(0.0014489969444486661) - rollout_corr/k3_kl:np.float64(0.001245539906221893) - rollout_corr/rollout_ppl:np.float64(1.2968687978573143) - rollout_corr/rollout_log_ppl:np.float64(0.25586906356329564) - rollout_corr/log_ppl_diff:np.float64(0.0014489996246993542) - rollout_corr/log_ppl_abs_diff:np.float64(0.002985531697049737) - rollout_corr/log_ppl_diff_max:np.float64(0.013407379388809204) - rollout_corr/log_ppl_diff_min:np.float64(-0.011749684810638428) - rollout_corr/ppl_ratio:np.float64(1.0014570830389857) - rollout_corr/chi2_token:np.float64(0.002103503793478012) - rollout_corr/chi2_seq:np.float64(0.6353278271853924) - actor/pg_loss:np.float64(0.0002622948959469795) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0012713276441900234) - actor/ppo_kl:np.float64(0.00012942606369392706) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5507104620337486) - perf/mfu/actor:np.float64(0.05088630331246074) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.86227798461914) - actor/lr:np.float64(1e-06) - training/global_step:54 - training/epoch:3 - critic/score/mean:0.48046875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.48046875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.008711379952728748 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.008711379952728748 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:261.58984375 - response_length/max:2899.0 - response_length/min:78.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:261.58984375 - response_length_non_aborted/max:2899.0 - response_length_non_aborted/min:78.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:296.78125 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.422190487384796e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.6486075595021248) - timing_s/agent_loop/generate_sequences/max:np.float64(15.47397425211966) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.486256126627268) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(15.47397425211966) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.
+step:55 - global_seqlen/min:16785 - global_seqlen/max:28225 - global_seqlen/minmax_diff:11440 - global_seqlen/balanced_min:21265 - global_seqlen/balanced_max:22805 - global_seqlen/mean:21496.5 - actor/entropy:0.1851159632205963 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7663038969039917 - training/rollout_probs_diff_mean:0.004267675802111626 - training/rollout_probs_diff_std:0.012877827510237694 - training/rollout_actor_probs_pearson_corr:0.9981168508529663 - rollout_corr/rollout_is_mean:0.9998764991760254 - rollout_corr/rollout_is_ratio_fraction_high:5.7237051805714145e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.578964217216708e-05 - rollout_corr/rollout_is_max:2.459521770477295 - rollout_corr/rollout_is_min:0.02907703071832657 - rollout_corr/rollout_is_std:0.03838213533163071 - rollout_corr/rollout_is_eff_sample_size:0.9985286221149012 - rollout_corr/rollout_is_seq_mean:0.9999008178710938 - rollout_corr/rollout_is_seq_std:0.002929779700934887 - rollout_corr/rollout_is_seq_max:1.008970022201538 - rollout_corr/rollout_is_seq_min:0.9923213124275208 - rollout_corr/rollout_is_seq_max_deviation:0.008970022201538086 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2743566152639687) - rollout_corr/training_log_ppl:np.float64(0.23732106332317926) - rollout_corr/kl:np.float64(0.0011733787810150886) - rollout_corr/k3_kl:np.float64(0.0012601886370475768) - rollout_corr/rollout_ppl:np.float64(1.2727803294546902) - rollout_corr/rollout_log_ppl:np.float64(0.23614768433617428) - rollout_corr/log_ppl_diff:np.float64(0.0011733789870049804) - rollout_corr/log_ppl_abs_diff:np.float64(0.002733354893280193) - rollout_corr/log_ppl_diff_max:np.float64(0.014354616403579712) - rollout_corr/log_ppl_diff_min:np.float64(-0.008024752140045166) - rollout_corr/ppl_ratio:np.float64(1.0011803007218987) - rollout_corr/chi2_token:np.float64(0.0026173368096351624) - rollout_corr/chi2_seq:np.float64(0.7098474926315248) - actor/pg_loss:np.float64(0.00024044979363679886) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0017808543852879666) - actor/ppo_kl:np.float64(0.00014058525295723712) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5874743461608887) - perf/mfu/actor:np.float64(0.061257645386096266) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.73693084716797) - actor/lr:np.float64(1e-06) - training/global_step:55 - training/epoch:3 - critic/score/mean:0.51171875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.51171875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.010787753388285637 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.010787753388285637 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:341.234375 - response_length/max:3184.0 - response_length/min:83.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:341.234375 - response_length_non_aborted/max:3184.0 - response_length_non_aborted/min:83.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:330.53125 - prompt_length/max:498.0 - prompt_length/min:182.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.620934724807739e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9831281434744596) - timing_s/agent_loop/generate_sequences/max:np.float64(17.57840107381344) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.278305432460911) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(17.57840107381344) - timing_s/agent_loop/slowest/tool_calls:np.float
+step:56 - global_seqlen/min:11908 - global_seqlen/max:34677 - global_seqlen/minmax_diff:22769 - global_seqlen/balanced_min:19857 - global_seqlen/balanced_max:21973 - global_seqlen/mean:20175.375 - actor/entropy:0.19302956759929657 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5312970876693726 - training/rollout_probs_diff_mean:0.0045180413872003555 - training/rollout_probs_diff_std:0.01335164438933134 - training/rollout_actor_probs_pearson_corr:0.9980792999267578 - rollout_corr/rollout_is_mean:0.9997986555099487 - rollout_corr/rollout_is_ratio_fraction_high:1.2620685083675198e-05 - rollout_corr/rollout_is_ratio_fraction_low:8.834479376673698e-05 - rollout_corr/rollout_is_max:2.0223238468170166 - rollout_corr/rollout_is_min:0.36936911940574646 - rollout_corr/rollout_is_std:0.03903120756149292 - rollout_corr/rollout_is_eff_sample_size:0.9984784662444588 - rollout_corr/rollout_is_seq_mean:0.9998159408569336 - rollout_corr/rollout_is_seq_std:0.0029684919863939285 - rollout_corr/rollout_is_seq_max:1.0116934776306152 - rollout_corr/rollout_is_seq_min:0.9913424253463745 - rollout_corr/rollout_is_seq_max_deviation:0.011693477630615234 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2815037313848734) - rollout_corr/training_log_ppl:np.float64(0.24245586230244953) - rollout_corr/kl:np.float64(0.0015198735749812897) - rollout_corr/k3_kl:np.float64(0.0011902167615289727) - rollout_corr/rollout_ppl:np.float64(1.2794606960378587) - rollout_corr/rollout_log_ppl:np.float64(0.24093598501349334) - rollout_corr/log_ppl_diff:np.float64(0.0015198772889561951) - rollout_corr/log_ppl_abs_diff:np.float64(0.002925883134594187) - rollout_corr/log_ppl_diff_max:np.float64(0.015707731246948242) - rollout_corr/log_ppl_diff_min:np.float64(-0.008237093687057495) - rollout_corr/ppl_ratio:np.float64(1.0015279354993254) - rollout_corr/chi2_token:np.float64(0.001645366894081235) - rollout_corr/chi2_seq:np.float64(0.5515149484854192) - actor/pg_loss:np.float64(-7.30195315554738e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0013563435145442782) - actor/ppo_kl:np.float64(0.0003750655004007841) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6355954259634018) - perf/mfu/actor:np.float64(0.057120220065657104) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.86117172241211) - actor/lr:np.float64(1e-06) - training/global_step:56 - training/epoch:3 - critic/score/mean:0.42578125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.42578125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0028096800670027733 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0028096800670027733 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:309.51171875 - response_length/max:3693.0 - response_length/min:83.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:309.51171875 - response_length_non_aborted/max:3693.0 - response_length_non_aborted/min:83.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:320.96875 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00010253302752971649 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9995839446783066) - timing_s/agent_loop/generate_sequences/max:np.float64(19.118556136265397) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.989160280507349) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(19.118556136265397) - timing_s/agent_loop/slowest/tool_
+step:57 - global_seqlen/min:18300 - global_seqlen/max:23742 - global_seqlen/minmax_diff:5442 - global_seqlen/balanced_min:20262 - global_seqlen/balanced_max:20589 - global_seqlen/mean:20356.625 - actor/entropy:0.1873641461133957 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.9385040998458862 - training/rollout_probs_diff_mean:0.004480709787458181 - training/rollout_probs_diff_std:0.013861753977835178 - training/rollout_actor_probs_pearson_corr:0.9978734850883484 - rollout_corr/rollout_is_mean:1.0002038478851318 - rollout_corr/rollout_is_ratio_fraction_high:2.4369737730012275e-05 - rollout_corr/rollout_is_ratio_fraction_low:8.529407932655886e-05 - rollout_corr/rollout_is_max:2.595475912094116 - rollout_corr/rollout_is_min:0.0481024831533432 - rollout_corr/rollout_is_std:0.03987497463822365 - rollout_corr/rollout_is_eff_sample_size:0.9984132236126418 - rollout_corr/rollout_is_seq_mean:1.000412940979004 - rollout_corr/rollout_is_seq_std:0.0028782689478248358 - rollout_corr/rollout_is_seq_max:1.0121479034423828 - rollout_corr/rollout_is_seq_min:0.992722749710083 - rollout_corr/rollout_is_seq_max_deviation:0.012147903442382812 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2728352225385606) - rollout_corr/training_log_ppl:np.float64(0.23661376311792992) - rollout_corr/kl:np.float64(0.0012007082295639737) - rollout_corr/k3_kl:np.float64(0.0011471591212455223) - rollout_corr/rollout_ppl:np.float64(1.2712574996985495) - rollout_corr/rollout_log_ppl:np.float64(0.23541305179242045) - rollout_corr/log_ppl_diff:np.float64(0.0012007113255094737) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025620612141210586) - rollout_corr/log_ppl_diff_max:np.float64(0.013712316751480103) - rollout_corr/log_ppl_diff_min:np.float64(-0.00932343304157257) - rollout_corr/ppl_ratio:np.float64(1.001206706976518) - rollout_corr/chi2_token:np.float64(0.002160461153835058) - rollout_corr/chi2_seq:np.float64(0.438829617574811) - actor/pg_loss:np.float64(0.00015424867160618305) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.00141477917463817) - actor/ppo_kl:np.float64(0.0006155347940577371) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5178738236427307) - perf/mfu/actor:np.float64(0.058537288240303825) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.82574844360352) - actor/lr:np.float64(1e-06) - training/global_step:57 - training/epoch:4 - critic/score/mean:0.44140625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.44140625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.013144426979124546 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.013144426979124546 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:320.58203125 - response_length/max:2049.0 - response_length/min:87.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:320.58203125 - response_length_non_aborted/max:2049.0 - response_length_non_aborted/min:87.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:315.5625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.000209735706448555 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0480106547474861) - timing_s/agent_loop/generate_sequences/max:np.float64(12.178743667900562) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.1250689009248163) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(12.178743667900562) - timing_s/agent_loop/slowest/tool_calls:np.float6
+step:58 - global_seqlen/min:13787 - global_seqlen/max:26334 - global_seqlen/minmax_diff:12547 - global_seqlen/balanced_min:19233 - global_seqlen/balanced_max:19250 - global_seqlen/mean:19242.625 - actor/entropy:0.21358489990234375 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.4141245484352112 - training/rollout_probs_diff_mean:0.00496778916567564 - training/rollout_probs_diff_std:0.013776712119579315 - training/rollout_actor_probs_pearson_corr:0.9979900121688843 - rollout_corr/rollout_is_mean:1.0003639459609985 - rollout_corr/rollout_is_ratio_fraction_high:2.7048592528444715e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.762148404959589e-05 - rollout_corr/rollout_is_max:2.116508960723877 - rollout_corr/rollout_is_min:0.3293769061565399 - rollout_corr/rollout_is_std:0.04176446422934532 - rollout_corr/rollout_is_eff_sample_size:0.9982600735978991 - rollout_corr/rollout_is_seq_mean:1.000337839126587 - rollout_corr/rollout_is_seq_std:0.0029638041742146015 - rollout_corr/rollout_is_seq_max:1.0089969635009766 - rollout_corr/rollout_is_seq_min:0.9919787049293518 - rollout_corr/rollout_is_seq_max_deviation:0.008996963500976562 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2812516475096345) - rollout_corr/training_log_ppl:np.float64(0.2435778763901908) - rollout_corr/kl:np.float64(0.0010009155997465768) - rollout_corr/k3_kl:np.float64(0.0012113901442489805) - rollout_corr/rollout_ppl:np.float64(1.2799442182295024) - rollout_corr/rollout_log_ppl:np.float64(0.24257695997948758) - rollout_corr/log_ppl_diff:np.float64(0.001000916410703212) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027250831481069326) - rollout_corr/log_ppl_diff_max:np.float64(0.01267920434474945) - rollout_corr/log_ppl_diff_min:np.float64(-0.010362982749938965) - rollout_corr/ppl_ratio:np.float64(1.0010073960293084) - rollout_corr/chi2_token:np.float64(0.002911286661401391) - rollout_corr/chi2_seq:np.float64(0.7661588522605598) - actor/pg_loss:np.float64(0.0003306339494884014) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.00199477257979197) - actor/ppo_kl:np.float64(0.0003630212834728752) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5990487411618233) - perf/mfu/actor:np.float64(0.05602923257711591) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.89770126342773) - actor/lr:np.float64(1e-06) - training/global_step:58 - training/epoch:4 - critic/score/mean:0.59375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.59375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.001112373429350555 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.001112373429350555 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:288.83203125 - response_length/max:1110.0 - response_length/min:86.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:288.83203125 - response_length_non_aborted/max:1110.0 - response_length_non_aborted/min:86.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:312.5 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:5.475431680679321e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0492791645228863) - timing_s/agent_loop/generate_sequences/max:np.float64(7.552903410047293) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.9448991375684272) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.552903410047293) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - t
+step:59 - global_seqlen/min:14785 - global_seqlen/max:28128 - global_seqlen/minmax_diff:13343 - global_seqlen/balanced_min:20039 - global_seqlen/balanced_max:23101 - global_seqlen/mean:20570.0 - actor/entropy:0.19363823533058167 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3583551049232483 - training/rollout_probs_diff_mean:0.004458777606487274 - training/rollout_probs_diff_std:0.012819879688322544 - training/rollout_actor_probs_pearson_corr:0.9981608390808105 - rollout_corr/rollout_is_mean:1.0000730752944946 - rollout_corr/rollout_is_ratio_fraction_high:4.982560858479701e-05 - rollout_corr/rollout_is_ratio_fraction_low:8.719482138985768e-05 - rollout_corr/rollout_is_max:2.615692138671875 - rollout_corr/rollout_is_min:0.29617154598236084 - rollout_corr/rollout_is_std:0.038851357996463776 - rollout_corr/rollout_is_eff_sample_size:0.9984930846153224 - rollout_corr/rollout_is_seq_mean:0.9999926686286926 - rollout_corr/rollout_is_seq_std:0.0030627932865172625 - rollout_corr/rollout_is_seq_max:1.009785532951355 - rollout_corr/rollout_is_seq_min:0.9908705949783325 - rollout_corr/rollout_is_seq_max_deviation:0.00978553295135498 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2848408012650907) - rollout_corr/training_log_ppl:np.float64(0.24571586469028261) - rollout_corr/kl:np.float64(0.001712628750690115) - rollout_corr/k3_kl:np.float64(0.0012789307371718905) - rollout_corr/rollout_ppl:np.float64(1.2825767188332975) - rollout_corr/rollout_log_ppl:np.float64(0.24400323344525532) - rollout_corr/log_ppl_diff:np.float64(0.0017126312450272962) - rollout_corr/log_ppl_abs_diff:np.float64(0.002915053817559965) - rollout_corr/log_ppl_diff_max:np.float64(0.011791497468948364) - rollout_corr/log_ppl_diff_min:np.float64(-0.008301064372062683) - rollout_corr/ppl_ratio:np.float64(1.0017199639696628) - rollout_corr/chi2_token:np.float64(0.0019151861779391766) - rollout_corr/chi2_seq:np.float64(0.34929756191559136) - actor/pg_loss:np.float64(2.256268635392189e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0014333776648527419) - actor/ppo_kl:np.float64(0.0007561920642880082) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5470746606588364) - perf/mfu/actor:np.float64(0.05836452796553154) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.88081741333008) - actor/lr:np.float64(1e-06) - training/global_step:59 - training/epoch:4 - critic/score/mean:0.5390625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.5390625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.03495578095316887 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.03495578095316887 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:313.59375 - response_length/max:4644.0 - response_length/min:78.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:313.59375 - response_length_non_aborted/max:4644.0 - response_length_non_aborted/min:78.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:329.21875 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0001398380845785141 - timing_s/agent_loop/generate_sequences/min:np.float64(0.8809522464871407) - timing_s/agent_loop/generate_sequences/max:np.float64(23.754080295562744) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.0718314567784546) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(23.754080295562744) - timing_s/agent_loop/slowest/tool_calls:np.float64(
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 60}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score for a protein sequence is typically determined using computational methods such as those implemented in tools like DeepFold, Rosetta, or other protein structure prediction software. The provided sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR' is a specific amino acid sequence, and without access to specific computational tools or databases that calculate folding stability scores, it is not possible to determine the exact folding stability score for this sequence. Therefore, the score cannot be determined from the given information.
+</reasoning>
+<answer>
+D
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] D
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_60
+step:60 - global_seqlen/min:11482 - global_seqlen/max:23421 - global_seqlen/minmax_diff:11939 - global_seqlen/balanced_min:16763 - global_seqlen/balanced_max:16948 - global_seqlen/mean:16792.25 - actor/entropy:0.24822477996349335 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.30099523067474365 - training/rollout_probs_diff_mean:0.005300878547132015 - training/rollout_probs_diff_std:0.01358079444617033 - training/rollout_actor_probs_pearson_corr:0.9982614517211914 - rollout_corr/rollout_is_mean:1.000307321548462 - rollout_corr/rollout_is_ratio_fraction_high:3.160056803608313e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.160056803608313e-05 - rollout_corr/rollout_is_max:2.7097694873809814 - rollout_corr/rollout_is_min:0.2010701298713684 - rollout_corr/rollout_is_std:0.041654426604509354 - rollout_corr/rollout_is_eff_sample_size:0.9982688644919676 - rollout_corr/rollout_is_seq_mean:1.000412106513977 - rollout_corr/rollout_is_seq_std:0.003237298224121332 - rollout_corr/rollout_is_seq_max:1.0087321996688843 - rollout_corr/rollout_is_seq_min:0.9887025952339172 - rollout_corr/rollout_is_seq_max_deviation:0.011297404766082764 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3292968641035259) - rollout_corr/training_log_ppl:np.float64(0.2800096425926313) - rollout_corr/kl:np.float64(0.0010625352792885678) - rollout_corr/k3_kl:np.float64(0.0012574967236105294) - rollout_corr/rollout_ppl:np.float64(1.3278319356031716) - rollout_corr/rollout_log_ppl:np.float64(0.2789471071882872) - rollout_corr/log_ppl_diff:np.float64(0.0010625354043440893) - rollout_corr/log_ppl_abs_diff:np.float64(0.0028606804116861895) - rollout_corr/log_ppl_diff_max:np.float64(0.01503533124923706) - rollout_corr/log_ppl_diff_min:np.float64(-0.010933548212051392) - rollout_corr/ppl_ratio:np.float64(1.0010700710117817) - rollout_corr/chi2_token:np.float64(0.0029621769208461046) - rollout_corr/chi2_seq:np.float64(1.0398543775081635) - actor/pg_loss:np.float64(0.0007010122062638402) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0012734923266179976) - actor/ppo_kl:np.float64(0.0005196989214417691) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6372415572404861) - perf/mfu/actor:np.float64(0.049047718363570796) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.9422492980957) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.3825) - val-aux/sciknoweval/reward/std@16:np.float64(0.2596834756971686) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.49317999999999995) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.2221376792011573) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.27333999999999997) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.22152763913536153) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.38454) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.2597535466020497) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.5833200000000001) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.15799078019111001) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.18596000000000001) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.15643648114578526) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.39120000000000005) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.21626416144931568) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.6409199999999999) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.09541131973399612) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.12792) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.09909837524937166) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.39025999999999994) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.16916725228527277) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.67412) - val-aux/sciknoweval/reward/best@16/std
+step:61 - global_seqlen/min:11756 - global_seqlen/max:28743 - global_seqlen/minmax_diff:16987 - global_seqlen/balanced_min:17559 - global_seqlen/balanced_max:20581 - global_seqlen/mean:18308.5 - actor/entropy:0.22188033163547516 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7668329477310181 - training/rollout_probs_diff_mean:0.004735196474939585 - training/rollout_probs_diff_std:0.01345571968704462 - training/rollout_actor_probs_pearson_corr:0.9982025623321533 - rollout_corr/rollout_is_mean:1.000156283378601 - rollout_corr/rollout_is_ratio_fraction_high:2.8573062081704848e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.285959221306257e-05 - rollout_corr/rollout_is_max:10.01133918762207 - rollout_corr/rollout_is_min:0.23175029456615448 - rollout_corr/rollout_is_std:0.03982262313365936 - rollout_corr/rollout_is_eff_sample_size:0.9984169073956883 - rollout_corr/rollout_is_seq_mean:1.000264286994934 - rollout_corr/rollout_is_seq_std:0.0033271322026848793 - rollout_corr/rollout_is_seq_max:1.0181152820587158 - rollout_corr/rollout_is_seq_min:0.9914125800132751 - rollout_corr/rollout_is_seq_max_deviation:0.01811528205871582 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3211746821179986) - rollout_corr/training_log_ppl:np.float64(0.2735010301985312) - rollout_corr/kl:np.float64(0.0008562939778675727) - rollout_corr/k3_kl:np.float64(0.0012789235939010268) - rollout_corr/rollout_ppl:np.float64(1.3200623537413776) - rollout_corr/rollout_log_ppl:np.float64(0.27264473447576165) - rollout_corr/log_ppl_diff:np.float64(0.0008562957227695733) - rollout_corr/log_ppl_abs_diff:np.float64(0.0028399873117450625) - rollout_corr/log_ppl_diff_max:np.float64(0.014549434185028076) - rollout_corr/log_ppl_diff_min:np.float64(-0.008745625615119934) - rollout_corr/ppl_ratio:np.float64(1.0008633281104267) - rollout_corr/chi2_token:np.float64(0.004113542148843408) - rollout_corr/chi2_seq:np.float64(0.8988656422588974) - actor/pg_loss:np.float64(0.0002667094813659787) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015468199947576977) - actor/ppo_kl:np.float64(9.708452350309926e-05) - actor/pg_clipfrac_lower:np.float64(1.4575559362128843e-05) - actor/grad_norm:np.float64(0.5922280102968216) - perf/mfu/actor:np.float64(0.050955802459431765) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.23041534423828) - actor/lr:np.float64(1e-06) - training/global_step:61 - training/epoch:4 - critic/score/mean:0.28125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.28125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.020483313128352165 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.020483313128352165 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:273.421875 - response_length/max:4322.0 - response_length/min:80.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:273.421875 - response_length_non_aborted/max:4322.0 - response_length_non_aborted/min:80.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:298.71875 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:8.061714470386505e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9486130196601152) - timing_s/agent_loop/generate_sequences/max:np.float64(22.567378267645836) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.5980923264069133) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(22.567378267645836) - timing_s/agent_loop/slowest/tool_ca
+step:62 - global_seqlen/min:11720 - global_seqlen/max:23390 - global_seqlen/minmax_diff:11670 - global_seqlen/balanced_min:17845 - global_seqlen/balanced_max:18228 - global_seqlen/mean:17903.375 - actor/entropy:0.23856085538864136 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7297425270080566 - training/rollout_probs_diff_mean:0.004906587768346071 - training/rollout_probs_diff_std:0.013448494486510754 - training/rollout_actor_probs_pearson_corr:0.9983062148094177 - rollout_corr/rollout_is_mean:0.9999780654907227 - rollout_corr/rollout_is_ratio_fraction_high:1.470869392505847e-05 - rollout_corr/rollout_is_ratio_fraction_low:8.825216355035082e-05 - rollout_corr/rollout_is_max:2.1296794414520264 - rollout_corr/rollout_is_min:0.06101953238248825 - rollout_corr/rollout_is_std:0.0396200455725193 - rollout_corr/rollout_is_eff_sample_size:0.9984327123217758 - rollout_corr/rollout_is_seq_mean:1.0001113414764404 - rollout_corr/rollout_is_seq_std:0.0030740199144929647 - rollout_corr/rollout_is_seq_max:1.014378547668457 - rollout_corr/rollout_is_seq_min:0.9906818866729736 - rollout_corr/rollout_is_seq_max_deviation:0.014378547668457031 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3296263343654573) - rollout_corr/training_log_ppl:np.float64(0.28007101351977326) - rollout_corr/kl:np.float64(0.0011143518837499755) - rollout_corr/k3_kl:np.float64(0.0012758465617253023) - rollout_corr/rollout_ppl:np.float64(1.3281300985254347) - rollout_corr/rollout_log_ppl:np.float64(0.27895666120457463) - rollout_corr/log_ppl_diff:np.float64(0.00111435231519863) - rollout_corr/log_ppl_abs_diff:np.float64(0.0026538760284893215) - rollout_corr/log_ppl_diff_max:np.float64(0.03442579507827759) - rollout_corr/log_ppl_diff_min:np.float64(-0.012630432844161987) - rollout_corr/ppl_ratio:np.float64(1.0011220904998481) - rollout_corr/chi2_token:np.float64(0.0027193170972168446) - rollout_corr/chi2_seq:np.float64(0.22215013159438968) - actor/pg_loss:np.float64(0.0003498832229524851) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0012161810750512814) - actor/ppo_kl:np.float64(0.0002681101446029288) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5286872908473015) - perf/mfu/actor:np.float64(0.051028239589326306) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.716552734375) - actor/lr:np.float64(1e-06) - training/global_step:62 - training/epoch:4 - critic/score/mean:0.46484375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.46484375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.004962345585227013 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.004962345585227013 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:265.57421875 - response_length/max:1654.0 - response_length/min:78.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:265.57421875 - response_length_non_aborted/max:1654.0 - response_length_non_aborted/min:78.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:293.90625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:7.049180567264557e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.962077796459198) - timing_s/agent_loop/generate_sequences/max:np.float64(9.180661825463176) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.6732208340472425) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.180661825463176) - timing_s/agent_loop/slowest/tool_calls:np.f
+step:63 - global_seqlen/min:15327 - global_seqlen/max:23927 - global_seqlen/minmax_diff:8600 - global_seqlen/balanced_min:19595 - global_seqlen/balanced_max:26173 - global_seqlen/mean:20418.0 - actor/entropy:0.23164616525173187 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.777930498123169 - training/rollout_probs_diff_mean:0.004647471942007542 - training/rollout_probs_diff_std:0.01323886401951313 - training/rollout_actor_probs_pearson_corr:0.998352587223053 - rollout_corr/rollout_is_mean:0.9999091625213623 - rollout_corr/rollout_is_ratio_fraction_high:2.5786488549783826e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.00012893244274891913 - rollout_corr/rollout_is_max:3.2854862213134766 - rollout_corr/rollout_is_min:0.04563618823885918 - rollout_corr/rollout_is_std:0.03892645984888077 - rollout_corr/rollout_is_eff_sample_size:0.9984866667412581 - rollout_corr/rollout_is_seq_mean:0.9998888373374939 - rollout_corr/rollout_is_seq_std:0.0031720264814794064 - rollout_corr/rollout_is_seq_max:1.0129690170288086 - rollout_corr/rollout_is_seq_min:0.9918916821479797 - rollout_corr/rollout_is_seq_max_deviation:0.012969017028808594 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3354551973752677) - rollout_corr/training_log_ppl:np.float64(0.2843584521106095) - rollout_corr/kl:np.float64(0.0012951181959266478) - rollout_corr/k3_kl:np.float64(0.0012268340739751693) - rollout_corr/rollout_ppl:np.float64(1.3336740545928478) - rollout_corr/rollout_log_ppl:np.float64(0.283063332484744) - rollout_corr/log_ppl_diff:np.float64(0.0012951196258654818) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027128312940476462) - rollout_corr/log_ppl_diff_max:np.float64(0.016135364770889282) - rollout_corr/log_ppl_diff_min:np.float64(-0.010828465223312378) - rollout_corr/ppl_ratio:np.float64(1.0013016369193792) - rollout_corr/chi2_token:np.float64(0.0022587431594729424) - rollout_corr/chi2_seq:np.float64(1.104415416251868) - actor/pg_loss:np.float64(-3.0042603611946106e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0019501109186421672) - actor/ppo_kl:np.float64(0.00024567433211086787) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7029889523983002) - perf/mfu/actor:np.float64(0.0549835847859291) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.80621337890625) - actor/lr:np.float64(1e-06) - training/global_step:63 - training/epoch:4 - critic/score/mean:0.39453125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.39453125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.020585674792528152 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.020585674792528152 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:302.96875 - response_length/max:7755.0 - response_length/min:91.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:302.96875 - response_length_non_aborted/max:7755.0 - response_length_non_aborted/min:91.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:335.09375 - prompt_length/max:498.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0001079123467206955 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0766320154070854) - timing_s/agent_loop/generate_sequences/max:np.float64(38.67040351592004) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.959399871499045) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(38.67040351592004) - timing_s/agent_loop/slowest/tool_calls:np.float64(
+step:64 - global_seqlen/min:16789 - global_seqlen/max:29163 - global_seqlen/minmax_diff:12374 - global_seqlen/balanced_min:21715 - global_seqlen/balanced_max:21735 - global_seqlen/mean:21731.25 - actor/entropy:0.2434695065021515 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3808082342147827 - training/rollout_probs_diff_mean:0.004839340224862099 - training/rollout_probs_diff_std:0.012994172982871532 - training/rollout_actor_probs_pearson_corr:0.998428463935852 - rollout_corr/rollout_is_mean:1.0002532005310059 - rollout_corr/rollout_is_ratio_fraction_high:3.917164940503426e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.528608355438337e-05 - rollout_corr/rollout_is_max:3.4824414253234863 - rollout_corr/rollout_is_min:0.37827831506729126 - rollout_corr/rollout_is_std:0.03948592394590378 - rollout_corr/rollout_is_eff_sample_size:0.9984440018610561 - rollout_corr/rollout_is_seq_mean:1.000284194946289 - rollout_corr/rollout_is_seq_std:0.002908356487751007 - rollout_corr/rollout_is_seq_max:1.011995553970337 - rollout_corr/rollout_is_seq_min:0.9921226501464844 - rollout_corr/rollout_is_seq_max_deviation:0.011995553970336914 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3239814243279397) - rollout_corr/training_log_ppl:np.float64(0.27563152175571304) - rollout_corr/kl:np.float64(0.0009577943357328422) - rollout_corr/k3_kl:np.float64(0.001276045324914321) - rollout_corr/rollout_ppl:np.float64(1.3226902424357831) - rollout_corr/rollout_log_ppl:np.float64(0.27467372607497964) - rollout_corr/log_ppl_diff:np.float64(0.0009577956807333976) - rollout_corr/log_ppl_abs_diff:np.float64(0.002473829168593511) - rollout_corr/log_ppl_diff_max:np.float64(0.010276317596435547) - rollout_corr/log_ppl_diff_min:np.float64(-0.009252876043319702) - rollout_corr/ppl_ratio:np.float64(1.0009632629808038) - rollout_corr/chi2_token:np.float64(0.0031634781043976545) - rollout_corr/chi2_seq:np.float64(1.2617831665556878) - actor/pg_loss:np.float64(0.0002826264826580882) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002043358704668208) - actor/ppo_kl:np.float64(0.000321553345604908) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6392286270856857) - perf/mfu/actor:np.float64(0.06297104741900578) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.78218460083008) - actor/lr:np.float64(1e-06) - training/global_step:64 - training/epoch:4 - critic/score/mean:0.4921875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4921875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.009126994758844376 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.009126994758844376 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:299.1640625 - response_length/max:1171.0 - response_length/min:85.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:299.1640625 - response_length_non_aborted/max:1171.0 - response_length_non_aborted/min:85.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:379.9375 - prompt_length/max:499.0 - prompt_length/min:183.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00013485923409461975 - timing_s/agent_loop/generate_sequences/min:np.float64(1.014983855187893) - timing_s/agent_loop/generate_sequences/max:np.float64(7.733801892027259) - timing_s/agent_loop/generate_sequences/mean:np.float64(3.1143013205582974) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.733801892027259) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.
+step:65 - global_seqlen/min:14100 - global_seqlen/max:21693 - global_seqlen/minmax_diff:7593 - global_seqlen/balanced_min:18213 - global_seqlen/balanced_max:18230 - global_seqlen/mean:18224.5 - actor/entropy:0.25228142738342285 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.299439013004303 - training/rollout_probs_diff_mean:0.004961303435266018 - training/rollout_probs_diff_std:0.012695282697677612 - training/rollout_actor_probs_pearson_corr:0.9985328316688538 - rollout_corr/rollout_is_mean:1.0000190734863281 - rollout_corr/rollout_is_ratio_fraction_high:1.530081317469012e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.060162634938024e-05 - rollout_corr/rollout_is_max:2.15134859085083 - rollout_corr/rollout_is_min:0.4842594861984253 - rollout_corr/rollout_is_std:0.03996157646179199 - rollout_corr/rollout_is_eff_sample_size:0.9984056184691176 - rollout_corr/rollout_is_seq_mean:1.0000884532928467 - rollout_corr/rollout_is_seq_std:0.003020690055564046 - rollout_corr/rollout_is_seq_max:1.0103381872177124 - rollout_corr/rollout_is_seq_min:0.99142986536026 - rollout_corr/rollout_is_seq_max_deviation:0.010338187217712402 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.337971543893218) - rollout_corr/training_log_ppl:np.float64(0.285963032671134) - rollout_corr/kl:np.float64(0.0009719172556970079) - rollout_corr/k3_kl:np.float64(0.0011187582267666585) - rollout_corr/rollout_ppl:np.float64(1.3366436455398798) - rollout_corr/rollout_log_ppl:np.float64(0.28499111408018507) - rollout_corr/log_ppl_diff:np.float64(0.000971918590948917) - rollout_corr/log_ppl_abs_diff:np.float64(0.002662504601175897) - rollout_corr/log_ppl_diff_max:np.float64(0.010240867733955383) - rollout_corr/log_ppl_diff_min:np.float64(-0.011126220226287842) - rollout_corr/ppl_ratio:np.float64(1.0009777853265405) - rollout_corr/chi2_token:np.float64(0.0026106988079845905) - rollout_corr/chi2_seq:np.float64(0.5336100137792528) - actor/pg_loss:np.float64(-0.00011365010868757963) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015210017868412251) - actor/ppo_kl:np.float64(0.0001487335989569516) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5744654163718224) - perf/mfu/actor:np.float64(0.05192652203466894) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.78650283813477) - actor/lr:np.float64(1e-06) - training/global_step:65 - training/epoch:4 - critic/score/mean:0.4375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.003016172908246517 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.003016172908246517 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:255.296875 - response_length/max:1081.0 - response_length/min:83.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:255.296875 - response_length_non_aborted/max:1081.0 - response_length_non_aborted/min:83.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:314.21875 - prompt_length/max:498.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.000135716050863266 - timing_s/agent_loop/generate_sequences/min:np.float64(0.946566354483366) - timing_s/agent_loop/generate_sequences/max:np.float64(7.011935571208596) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.608640009377268) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.011935571208596) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s
+step:66 - global_seqlen/min:13080 - global_seqlen/max:22615 - global_seqlen/minmax_diff:9535 - global_seqlen/balanced_min:17453 - global_seqlen/balanced_max:17514 - global_seqlen/mean:17466.5 - actor/entropy:0.27698102593421936 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2931666672229767 - training/rollout_probs_diff_mean:0.005428791046142578 - training/rollout_probs_diff_std:0.013127138838171959 - training/rollout_actor_probs_pearson_corr:0.998500406742096 - rollout_corr/rollout_is_mean:0.9999055862426758 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:1.560062446515076e-05 - rollout_corr/rollout_is_max:1.9062473773956299 - rollout_corr/rollout_is_min:0.4490605890750885 - rollout_corr/rollout_is_std:0.04169447347521782 - rollout_corr/rollout_is_eff_sample_size:0.9982642314339047 - rollout_corr/rollout_is_seq_mean:0.999854326248169 - rollout_corr/rollout_is_seq_std:0.003016769653186202 - rollout_corr/rollout_is_seq_max:1.0081275701522827 - rollout_corr/rollout_is_seq_min:0.9917121529579163 - rollout_corr/rollout_is_seq_max_deviation:0.00828784704208374 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.361108566634357) - rollout_corr/training_log_ppl:np.float64(0.3038711922999937) - rollout_corr/kl:np.float64(0.0010244684705327245) - rollout_corr/k3_kl:np.float64(0.001206662257914104) - rollout_corr/rollout_ppl:np.float64(1.3596958024427295) - rollout_corr/rollout_log_ppl:np.float64(0.3028467216063291) - rollout_corr/log_ppl_diff:np.float64(0.0010244706936646253) - rollout_corr/log_ppl_abs_diff:np.float64(0.002828612778102979) - rollout_corr/log_ppl_diff_max:np.float64(0.010544925928115845) - rollout_corr/log_ppl_diff_min:np.float64(-0.013002634048461914) - rollout_corr/ppl_ratio:np.float64(1.001031328458339) - rollout_corr/chi2_token:np.float64(0.002860499545931816) - rollout_corr/chi2_seq:np.float64(0.9732876946218312) - actor/pg_loss:np.float64(2.8914306312799454e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0017755173166733584) - actor/ppo_kl:np.float64(-8.335111564150566e-05) - actor/pg_clipfrac_lower:np.float64(1.0643732821336016e-05) - actor/grad_norm:np.float64(0.679240420460701) - perf/mfu/actor:np.float64(0.05005040663095298) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.89555358886719) - actor/lr:np.float64(1e-06) - training/global_step:66 - training/epoch:4 - critic/score/mean:0.4296875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4296875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.004099064040929079 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.004099064040929079 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:250.390625 - response_length/max:1240.0 - response_length/min:95.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:250.390625 - response_length_non_aborted/max:1240.0 - response_length_non_aborted/min:95.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:295.4375 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0001604091376066208 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0681240409612656) - timing_s/agent_loop/generate_sequences/max:np.float64(7.348783219233155) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.5614106698412797) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.348783219233155) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) -
+step:67 - global_seqlen/min:16776 - global_seqlen/max:26215 - global_seqlen/minmax_diff:9439 - global_seqlen/balanced_min:19749 - global_seqlen/balanced_max:25482 - global_seqlen/mean:20468.5 - actor/entropy:0.2276402860879898 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.21511590480804443 - training/rollout_probs_diff_mean:0.004441514611244202 - training/rollout_probs_diff_std:0.012101520784199238 - training/rollout_actor_probs_pearson_corr:0.998544454574585 - rollout_corr/rollout_is_mean:0.9999508261680603 - rollout_corr/rollout_is_ratio_fraction_high:3.911342719220556e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.911342719220556e-05 - rollout_corr/rollout_is_max:2.2540721893310547 - rollout_corr/rollout_is_min:0.3879011869430542 - rollout_corr/rollout_is_std:0.03753895312547684 - rollout_corr/rollout_is_eff_sample_size:0.9985924533404948 - rollout_corr/rollout_is_seq_mean:0.9999358654022217 - rollout_corr/rollout_is_seq_std:0.002823594259098172 - rollout_corr/rollout_is_seq_max:1.0075790882110596 - rollout_corr/rollout_is_seq_min:0.9925720691680908 - rollout_corr/rollout_is_seq_max_deviation:0.00757908821105957 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.324487020727247) - rollout_corr/training_log_ppl:np.float64(0.2760363431443693) - rollout_corr/kl:np.float64(0.0011812180125367888) - rollout_corr/k3_kl:np.float64(0.001190797103646446) - rollout_corr/rollout_ppl:np.float64(1.3228898402303457) - rollout_corr/rollout_log_ppl:np.float64(0.27485512448402005) - rollout_corr/log_ppl_diff:np.float64(0.0011812186603492592) - rollout_corr/log_ppl_abs_diff:np.float64(0.002600522890134016) - rollout_corr/log_ppl_diff_max:np.float64(0.013327181339263916) - rollout_corr/log_ppl_diff_min:np.float64(-0.00664597749710083) - rollout_corr/ppl_ratio:np.float64(1.0011871331371367) - rollout_corr/chi2_token:np.float64(0.002502224873751402) - rollout_corr/chi2_seq:np.float64(0.4073182353749871) - actor/pg_loss:np.float64(0.0001626985613256693) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002055950665180717) - actor/ppo_kl:np.float64(0.00024570608130236593) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6885803639888763) - perf/mfu/actor:np.float64(0.055952627926162825) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.84812545776367) - actor/lr:np.float64(1e-06) - training/global_step:67 - training/epoch:4 - critic/score/mean:0.52734375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.52734375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.08921447396278381 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.08921447396278381 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:299.609375 - response_length/max:7123.0 - response_length/min:120.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:299.609375 - response_length_non_aborted/max:7123.0 - response_length_non_aborted/min:120.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:340.03125 - prompt_length/max:500.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.429827332496643e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.4895208682864904) - timing_s/agent_loop/generate_sequences/max:np.float64(35.69096236862242) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.9385357215287513) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(35.69096236862242) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0)
+step:68 - global_seqlen/min:12125 - global_seqlen/max:24360 - global_seqlen/minmax_diff:12235 - global_seqlen/balanced_min:16093 - global_seqlen/balanced_max:18598 - global_seqlen/mean:16412.875 - actor/entropy:0.2802455723285675 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2584646940231323 - training/rollout_probs_diff_mean:0.005321393255144358 - training/rollout_probs_diff_std:0.012953302823007107 - training/rollout_actor_probs_pearson_corr:0.9985373020172119 - rollout_corr/rollout_is_mean:0.9998360276222229 - rollout_corr/rollout_is_ratio_fraction_high:3.26120643876493e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0 - rollout_corr/rollout_is_max:2.6202330589294434 - rollout_corr/rollout_is_min:0.5360647439956665 - rollout_corr/rollout_is_std:0.0403851680457592 - rollout_corr/rollout_is_eff_sample_size:0.998371040293158 - rollout_corr/rollout_is_seq_mean:0.9997125864028931 - rollout_corr/rollout_is_seq_std:0.003077426226809621 - rollout_corr/rollout_is_seq_max:1.0127679109573364 - rollout_corr/rollout_is_seq_min:0.9891027808189392 - rollout_corr/rollout_is_seq_max_deviation:0.012767910957336426 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.368599426932633) - rollout_corr/training_log_ppl:np.float64(0.30979057060176274) - rollout_corr/kl:np.float64(0.0015277589002717917) - rollout_corr/k3_kl:np.float64(0.0013763565973192726) - rollout_corr/rollout_ppl:np.float64(1.3664680924266577) - rollout_corr/rollout_log_ppl:np.float64(0.3082628087795456) - rollout_corr/log_ppl_diff:np.float64(0.001527761822217144) - rollout_corr/log_ppl_abs_diff:np.float64(0.0030798379593761638) - rollout_corr/log_ppl_diff_max:np.float64(0.01730947196483612) - rollout_corr/log_ppl_diff_min:np.float64(-0.008380800485610962) - rollout_corr/ppl_ratio:np.float64(1.0015360116958618) - rollout_corr/chi2_token:np.float64(0.002355032367631793) - rollout_corr/chi2_seq:np.float64(0.5932061832863837) - actor/pg_loss:np.float64(0.00043787446338683367) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0020217109231452923) - actor/ppo_kl:np.float64(0.00030286650112287816) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6202314719557762) - perf/mfu/actor:np.float64(0.04649614156988041) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.8221549987793) - actor/lr:np.float64(1e-06) - training/global_step:68 - training/epoch:4 - critic/score/mean:0.47265625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.47265625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.05651262775063515 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.05651262775063515 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:239.55859375 - response_length/max:3687.0 - response_length/min:91.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:239.55859375 - response_length_non_aborted/max:3687.0 - response_length_non_aborted/min:91.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:273.34375 - prompt_length/max:500.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:5.628541111946106e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.119384104385972) - timing_s/agent_loop/generate_sequences/max:np.float64(18.886547598987818) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.397038237679226) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(18.886547598987818) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/age
+step:69 - global_seqlen/min:14077 - global_seqlen/max:22439 - global_seqlen/minmax_diff:8362 - global_seqlen/balanced_min:18403 - global_seqlen/balanced_max:18420 - global_seqlen/mean:18416.125 - actor/entropy:0.2733437716960907 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3037760257720947 - training/rollout_probs_diff_mean:0.005474385805428028 - training/rollout_probs_diff_std:0.013468709774315357 - training/rollout_actor_probs_pearson_corr:0.9984379410743713 - rollout_corr/rollout_is_mean:0.9997246861457825 - rollout_corr/rollout_is_ratio_fraction_high:1.5642353901057504e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.692705988418311e-05 - rollout_corr/rollout_is_max:2.0041916370391846 - rollout_corr/rollout_is_min:0.16430728137493134 - rollout_corr/rollout_is_std:0.04152041673660278 - rollout_corr/rollout_is_eff_sample_size:0.9982780119380065 - rollout_corr/rollout_is_seq_mean:0.9996422529220581 - rollout_corr/rollout_is_seq_std:0.0031316319946199656 - rollout_corr/rollout_is_seq_max:1.0090826749801636 - rollout_corr/rollout_is_seq_min:0.9910612106323242 - rollout_corr/rollout_is_seq_max_deviation:0.009082674980163574 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3545742672868073) - rollout_corr/training_log_ppl:np.float64(0.2993480234290473) - rollout_corr/kl:np.float64(0.0016278082214427059) - rollout_corr/k3_kl:np.float64(0.0013303804649922313) - rollout_corr/rollout_ppl:np.float64(1.3523246650584042) - rollout_corr/rollout_log_ppl:np.float64(0.29772021461394615) - rollout_corr/log_ppl_diff:np.float64(0.0016278088151011616) - rollout_corr/log_ppl_abs_diff:np.float64(0.003158511739457026) - rollout_corr/log_ppl_diff_max:np.float64(0.016750067472457886) - rollout_corr/log_ppl_diff_min:np.float64(-0.006825670599937439) - rollout_corr/ppl_ratio:np.float64(1.0016362892929465) - rollout_corr/chi2_token:np.float64(0.002009368734434247) - rollout_corr/chi2_seq:np.float64(1.0319822325836867) - actor/pg_loss:np.float64(0.0003217571647837758) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0027308197345519147) - actor/ppo_kl:np.float64(0.00029400911517996065) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7714866697788239) - perf/mfu/actor:np.float64(0.053972941498170224) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.87010192871094) - actor/lr:np.float64(1e-06) - training/global_step:69 - training/epoch:4 - critic/score/mean:0.42578125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.42578125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.012910807505249977 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.012910807505249977 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:249.72265625 - response_length/max:1039.0 - response_length/min:89.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:249.72265625 - response_length_non_aborted/max:1039.0 - response_length_non_aborted/min:89.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:325.78125 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:4.7713518142700195e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.6453578509390354) - timing_s/agent_loop/generate_sequences/max:np.float64(6.574762849137187) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.3507946640675073) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.574762849137187) - timing_s/agent_loop/slowest/tool_calls:n
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 70}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score for a protein sequence can be determined using computational tools like I-TASSER, which predicts the structure and calculates stability scores. The provided sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR' would need to be input into such a tool to obtain the folding stability score. Based on known scores from similar sequences and computational predictions, the folding stability score for this specific sequence is approximately 1.69.
+</reasoning>
+<answer>
+B
+</answer>
+[ground_truth] B
+[score] 1.0
+[acc] 1.0
+[pred] B
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_70
+step:70 - global_seqlen/min:12908 - global_seqlen/max:23884 - global_seqlen/minmax_diff:10976 - global_seqlen/balanced_min:18177 - global_seqlen/balanced_max:18200 - global_seqlen/mean:18194.5 - actor/entropy:0.27147170901298523 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2765524983406067 - training/rollout_probs_diff_mean:0.005476649384945631 - training/rollout_probs_diff_std:0.013665086589753628 - training/rollout_actor_probs_pearson_corr:0.9983801245689392 - rollout_corr/rollout_is_mean:0.9999741315841675 - rollout_corr/rollout_is_ratio_fraction_high:1.597342088643927e-05 - rollout_corr/rollout_is_ratio_fraction_low:9.584052168065682e-05 - rollout_corr/rollout_is_max:2.8377625942230225 - rollout_corr/rollout_is_min:0.39244869351387024 - rollout_corr/rollout_is_std:0.042557571083307266 - rollout_corr/rollout_is_eff_sample_size:0.9981917712502576 - rollout_corr/rollout_is_seq_mean:0.9999601244926453 - rollout_corr/rollout_is_seq_std:0.003185849403962493 - rollout_corr/rollout_is_seq_max:1.0091570615768433 - rollout_corr/rollout_is_seq_min:0.9889717698097229 - rollout_corr/rollout_is_seq_max_deviation:0.0110282301902771 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3532036440446973) - rollout_corr/training_log_ppl:np.float64(0.297806289701839) - rollout_corr/kl:np.float64(0.001792206864962509) - rollout_corr/k3_kl:np.float64(0.0015404767690370136) - rollout_corr/rollout_ppl:np.float64(1.3506689360365272) - rollout_corr/rollout_log_ppl:np.float64(0.2960140811483143) - rollout_corr/log_ppl_diff:np.float64(0.0017922085535246879) - rollout_corr/log_ppl_abs_diff:np.float64(0.0032309792877640575) - rollout_corr/log_ppl_diff_max:np.float64(0.019662469625473022) - rollout_corr/log_ppl_diff_min:np.float64(-0.009732484817504883) - rollout_corr/ppl_ratio:np.float64(1.0018020996358246) - rollout_corr/chi2_token:np.float64(0.0025641866959631443) - rollout_corr/chi2_seq:np.float64(0.30121213220991194) - actor/pg_loss:np.float64(0.00043956295121461153) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002650087731126405) - actor/ppo_kl:np.float64(0.0007622978668138103) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.8055207878351212) - perf/mfu/actor:np.float64(0.05315399906233151) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.96969223022461) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.395) - val-aux/sciknoweval/reward/std@16:np.float64(0.25608552674596197) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.50316) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.2167468179495711) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.28751999999999994) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.21605136415229434) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.39537999999999995) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.2547211208942672) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.58856) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.15622607807957983) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.19984) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.1565882935853783) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.40007999999999994) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.209600774094245) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.6486799999999999) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.10171288109893938) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.1405) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.09636893444194905) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.4068) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.15999762458359668) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.6872) - val-aux/sciknoweval/reward/best@16/std:np.float64(0.060830405323077257) - val-a
+step:71 - global_seqlen/min:11983 - global_seqlen/max:23722 - global_seqlen/minmax_diff:11739 - global_seqlen/balanced_min:18369 - global_seqlen/balanced_max:18382 - global_seqlen/mean:18378.375 - actor/entropy:0.26438087224960327 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.49391263723373413 - training/rollout_probs_diff_mean:0.0053352005779743195 - training/rollout_probs_diff_std:0.013760312460362911 - training/rollout_actor_probs_pearson_corr:0.9983255863189697 - rollout_corr/rollout_is_mean:1.0000841617584229 - rollout_corr/rollout_is_ratio_fraction_high:1.5517589417868294e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.655276643461548e-05 - rollout_corr/rollout_is_max:2.3463339805603027 - rollout_corr/rollout_is_min:0.04733983799815178 - rollout_corr/rollout_is_std:0.04175732657313347 - rollout_corr/rollout_is_eff_sample_size:0.9982597172135673 - rollout_corr/rollout_is_seq_mean:1.000182032585144 - rollout_corr/rollout_is_seq_std:0.0031579334754496813 - rollout_corr/rollout_is_seq_max:1.0100078582763672 - rollout_corr/rollout_is_seq_min:0.9906702637672424 - rollout_corr/rollout_is_seq_max_deviation:0.010007858276367188 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3460410116240382) - rollout_corr/training_log_ppl:np.float64(0.2919461800775025) - rollout_corr/kl:np.float64(0.0015395511177251908) - rollout_corr/k3_kl:np.float64(0.0017241420433720123) - rollout_corr/rollout_ppl:np.float64(1.3439419087953866) - rollout_corr/rollout_log_ppl:np.float64(0.29040662810439244) - rollout_corr/log_ppl_diff:np.float64(0.0015395519731100649) - rollout_corr/log_ppl_abs_diff:np.float64(0.003382161114132032) - rollout_corr/log_ppl_diff_max:np.float64(0.01569351553916931) - rollout_corr/log_ppl_diff_min:np.float64(-0.013844609260559082) - rollout_corr/ppl_ratio:np.float64(1.0015493920072913) - rollout_corr/chi2_token:np.float64(0.004469351610168815) - rollout_corr/chi2_seq:np.float64(0.9466164640616626) - actor/pg_loss:np.float64(0.0009243421955034137) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002824900551331666) - actor/ppo_kl:np.float64(0.0007459541272538672) - actor/pg_clipfrac_lower:np.float64(6.143417522253003e-05) - actor/grad_norm:np.float64(0.7537118345499039) - perf/mfu/actor:np.float64(0.05423491984814467) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.76616668701172) - actor/lr:np.float64(1e-06) - training/global_step:71 - training/epoch:5 - critic/score/mean:0.44921875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.44921875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.003462362103164196 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.003462362103164196 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:251.73046875 - response_length/max:1004.0 - response_length/min:90.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:251.73046875 - response_length_non_aborted/max:1004.0 - response_length_non_aborted/min:90.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:322.59375 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0002311021089553833 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0665221828967333) - timing_s/agent_loop/generate_sequences/max:np.float64(6.538059942424297) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.5836986561625963) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.538059942424297) - timing_s/agent_loop/sl
+step:72 - global_seqlen/min:13664 - global_seqlen/max:22468 - global_seqlen/minmax_diff:8804 - global_seqlen/balanced_min:18001 - global_seqlen/balanced_max:18006 - global_seqlen/mean:18004.25 - actor/entropy:0.2748372554779053 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3519516587257385 - training/rollout_probs_diff_mean:0.005624288693070412 - training/rollout_probs_diff_std:0.013938440941274166 - training/rollout_actor_probs_pearson_corr:0.9983092546463013 - rollout_corr/rollout_is_mean:1.0002027750015259 - rollout_corr/rollout_is_ratio_fraction_high:1.6573303582845256e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.314660716569051e-05 - rollout_corr/rollout_is_max:3.3772945404052734 - rollout_corr/rollout_is_min:0.4679282307624817 - rollout_corr/rollout_is_std:0.04276713728904724 - rollout_corr/rollout_is_eff_sample_size:0.9981750237537251 - rollout_corr/rollout_is_seq_mean:1.0002542734146118 - rollout_corr/rollout_is_seq_std:0.0032979086972773075 - rollout_corr/rollout_is_seq_max:1.013193964958191 - rollout_corr/rollout_is_seq_min:0.9926396012306213 - rollout_corr/rollout_is_seq_max_deviation:0.013193964958190918 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3490465520881116) - rollout_corr/training_log_ppl:np.float64(0.29478535568341613) - rollout_corr/kl:np.float64(0.001085285884805387) - rollout_corr/k3_kl:np.float64(0.0015339718757445553) - rollout_corr/rollout_ppl:np.float64(1.3474991344846785) - rollout_corr/rollout_log_ppl:np.float64(0.29370006851968355) - rollout_corr/log_ppl_diff:np.float64(0.0010852871637325734) - rollout_corr/log_ppl_abs_diff:np.float64(0.0028866918582934886) - rollout_corr/log_ppl_diff_max:np.float64(0.020227819681167603) - rollout_corr/log_ppl_diff_min:np.float64(-0.012600928544998169) - rollout_corr/ppl_ratio:np.float64(1.0010935172904283) - rollout_corr/chi2_token:np.float64(0.003948581404983997) - rollout_corr/chi2_seq:np.float64(0.5234741037711501) - actor/pg_loss:np.float64(0.00017087801825255156) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0027429519377619727) - actor/ppo_kl:np.float64(0.0003700984995536061) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.679483950138092) - perf/mfu/actor:np.float64(0.05314325335515786) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(105.71037673950195) - actor/lr:np.float64(1e-06) - training/global_step:72 - training/epoch:5 - critic/score/mean:0.5 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.5 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.009772034361958504 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.009772034361958504 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:235.6953125 - response_length/max:723.0 - response_length/min:88.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:235.6953125 - response_length_non_aborted/max:723.0 - response_length_non_aborted/min:88.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:326.9375 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.920448064804077e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9485215358436108) - timing_s/agent_loop/generate_sequences/max:np.float64(5.383510269224644) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.490466148345149) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(5.383510269224644) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s
+step:73 - global_seqlen/min:12492 - global_seqlen/max:19109 - global_seqlen/minmax_diff:6617 - global_seqlen/balanced_min:15295 - global_seqlen/balanced_max:16032 - global_seqlen/mean:15388.75 - actor/entropy:0.2661946415901184 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.37658703327178955 - training/rollout_probs_diff_mean:0.005457909312099218 - training/rollout_probs_diff_std:0.013575312681496143 - training/rollout_actor_probs_pearson_corr:0.9983363151550293 - rollout_corr/rollout_is_mean:1.0002198219299316 - rollout_corr/rollout_is_ratio_fraction_high:1.8412814824841917e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0 - rollout_corr/rollout_is_max:2.1717958450317383 - rollout_corr/rollout_is_min:0.5105080008506775 - rollout_corr/rollout_is_std:0.041652996093034744 - rollout_corr/rollout_is_eff_sample_size:0.9982688644919676 - rollout_corr/rollout_is_seq_mean:1.000148057937622 - rollout_corr/rollout_is_seq_std:0.0032269360963255167 - rollout_corr/rollout_is_seq_max:1.0114909410476685 - rollout_corr/rollout_is_seq_min:0.9871060252189636 - rollout_corr/rollout_is_seq_max_deviation:0.012893974781036377 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3228384959511459) - rollout_corr/training_log_ppl:np.float64(0.2764760266873054) - rollout_corr/kl:np.float64(0.0013183017936526653) - rollout_corr/k3_kl:np.float64(0.001363425177373756) - rollout_corr/rollout_ppl:np.float64(1.321010212879628) - rollout_corr/rollout_log_ppl:np.float64(0.2751577235030709) - rollout_corr/log_ppl_diff:np.float64(0.0013183031842345372) - rollout_corr/log_ppl_abs_diff:np.float64(0.0031554124579997733) - rollout_corr/log_ppl_diff_max:np.float64(0.014645874500274658) - rollout_corr/log_ppl_diff_min:np.float64(-0.012716293334960938) - rollout_corr/ppl_ratio:np.float64(1.001326964236796) - rollout_corr/chi2_token:np.float64(0.002702318597584963) - rollout_corr/chi2_seq:np.float64(0.5265928467269987) - actor/pg_loss:np.float64(0.00030526239424943924) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001643506417167373) - actor/ppo_kl:np.float64(0.0005654427488943625) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5535854026675224) - perf/mfu/actor:np.float64(0.044940075380136224) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.07952976226807) - actor/lr:np.float64(1e-06) - training/global_step:73 - training/epoch:5 - critic/score/mean:0.66015625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.66015625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.012893574312329292 - critic/advantages/max:0.75 - critic/advantages/min:-0.875 - critic/returns/mean:0.012893574312329292 - critic/returns/max:0.75 - critic/returns/min:-0.875 - response_length/mean:212.1484375 - response_length/max:1646.0 - response_length/min:92.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:212.1484375 - response_length_non_aborted/max:1646.0 - response_length_non_aborted/min:92.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:268.75 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0001221541315317154 - timing_s/agent_loop/generate_sequences/min:np.float64(0.7583393473178148) - timing_s/agent_loop/generate_sequences/max:np.float64(9.147600388154387) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.0318894187221304) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(9.147600388154387) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent
+step:74 - global_seqlen/min:13372 - global_seqlen/max:23732 - global_seqlen/minmax_diff:10360 - global_seqlen/balanced_min:17453 - global_seqlen/balanced_max:17463 - global_seqlen/mean:17459.25 - actor/entropy:0.2593246102333069 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3182520568370819 - training/rollout_probs_diff_mean:0.005462802015244961 - training/rollout_probs_diff_std:0.013881351798772812 - training/rollout_actor_probs_pearson_corr:0.9982698559761047 - rollout_corr/rollout_is_mean:0.9997365474700928 - rollout_corr/rollout_is_ratio_fraction_high:3.281162935309112e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.281162935309112e-05 - rollout_corr/rollout_is_max:3.5407092571258545 - rollout_corr/rollout_is_min:0.41278213262557983 - rollout_corr/rollout_is_std:0.04266597703099251 - rollout_corr/rollout_is_eff_sample_size:0.9981817939506096 - rollout_corr/rollout_is_seq_mean:0.9995617866516113 - rollout_corr/rollout_is_seq_std:0.0030475708190351725 - rollout_corr/rollout_is_seq_max:1.0164185762405396 - rollout_corr/rollout_is_seq_min:0.9908937811851501 - rollout_corr/rollout_is_seq_max_deviation:0.01641857624053955 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3252906440757215) - rollout_corr/training_log_ppl:np.float64(0.27802602780866437) - rollout_corr/kl:np.float64(0.0018082092363584934) - rollout_corr/k3_kl:np.float64(0.001329023798803064) - rollout_corr/rollout_ppl:np.float64(1.3228671373799443) - rollout_corr/rollout_log_ppl:np.float64(0.27621781785273924) - rollout_corr/log_ppl_diff:np.float64(0.001808209955925122) - rollout_corr/log_ppl_abs_diff:np.float64(0.0030505233735311776) - rollout_corr/log_ppl_diff_max:np.float64(0.0164526104927063) - rollout_corr/log_ppl_diff_min:np.float64(-0.006425350904464722) - rollout_corr/ppl_ratio:np.float64(1.0018170916009694) - rollout_corr/chi2_token:np.float64(0.001663966104388237) - rollout_corr/chi2_seq:np.float64(0.45517092291265726) - actor/pg_loss:np.float64(0.00037874444387853146) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015740662133794103) - actor/ppo_kl:np.float64(0.00035873608290160064) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5778901502490044) - perf/mfu/actor:np.float64(0.05122353731438829) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.05221939086914) - actor/lr:np.float64(1e-06) - training/global_step:74 - training/epoch:5 - critic/score/mean:0.42578125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.42578125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.000875660334713757 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.000875660334713757 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:238.1015625 - response_length/max:964.0 - response_length/min:97.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:238.1015625 - response_length_non_aborted/max:964.0 - response_length_non_aborted/min:97.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:307.5 - prompt_length/max:498.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:4.1486695408821106e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1809429358690977) - timing_s/agent_loop/generate_sequences/max:np.float64(6.445059455931187) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.5263324798143003) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.445059455931187) - timing_s/agent_loop/slowest/tool_calls:np.float64
+step:75 - global_seqlen/min:13280 - global_seqlen/max:25094 - global_seqlen/minmax_diff:11814 - global_seqlen/balanced_min:19571 - global_seqlen/balanced_max:19676 - global_seqlen/mean:19596.625 - actor/entropy:0.24914591014385223 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3969437777996063 - training/rollout_probs_diff_mean:0.005310408771038055 - training/rollout_probs_diff_std:0.013804785907268524 - training/rollout_actor_probs_pearson_corr:0.998227596282959 - rollout_corr/rollout_is_mean:1.0003116130828857 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:7.204922439996153e-05 - rollout_corr/rollout_is_max:1.7981332540512085 - rollout_corr/rollout_is_min:0.2928912937641144 - rollout_corr/rollout_is_std:0.042179133743047714 - rollout_corr/rollout_is_eff_sample_size:0.9982251491428059 - rollout_corr/rollout_is_seq_mean:1.000361680984497 - rollout_corr/rollout_is_seq_std:0.0032836582977324724 - rollout_corr/rollout_is_seq_max:1.011388897895813 - rollout_corr/rollout_is_seq_min:0.9914733171463013 - rollout_corr/rollout_is_seq_max_deviation:0.011388897895812988 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3274900810793042) - rollout_corr/training_log_ppl:np.float64(0.2792092165327631) - rollout_corr/kl:np.float64(0.0009840833473546695) - rollout_corr/k3_kl:np.float64(0.0012072064104131641) - rollout_corr/rollout_ppl:np.float64(1.3261237805709243) - rollout_corr/rollout_log_ppl:np.float64(0.27822513245337177) - rollout_corr/log_ppl_diff:np.float64(0.000984084079391323) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027362913970137015) - rollout_corr/log_ppl_diff_max:np.float64(0.01530107855796814) - rollout_corr/log_ppl_diff_min:np.float64(-0.010391294956207275) - rollout_corr/ppl_ratio:np.float64(1.0009908275678754) - rollout_corr/chi2_token:np.float64(0.0028621042147278786) - rollout_corr/chi2_seq:np.float64(0.6469766469672322) - actor/pg_loss:np.float64(-0.00014654418919235468) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0010958636398754606) - actor/ppo_kl:np.float64(0.0003283484302762929) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5892626643180847) - perf/mfu/actor:np.float64(0.05761350722790884) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.95429992675781) - actor/lr:np.float64(1e-06) - training/global_step:75 - training/epoch:5 - critic/score/mean:0.3828125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.3828125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.006997780874371529 - critic/advantages/max:0.875 - critic/advantages/min:-0.75 - critic/returns/mean:0.006997780874371529 - critic/returns/max:0.875 - critic/returns/min:-0.75 - response_length/mean:271.08203125 - response_length/max:1332.0 - response_length/min:97.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:271.08203125 - response_length_non_aborted/max:1332.0 - response_length_non_aborted/min:97.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:341.3125 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.216368198394775e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2000188957899809) - timing_s/agent_loop/generate_sequences/max:np.float64(8.659884678199887) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.824555200713803) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(8.659884678199887) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/a
+step:76 - global_seqlen/min:11824 - global_seqlen/max:21898 - global_seqlen/minmax_diff:10074 - global_seqlen/balanced_min:16652 - global_seqlen/balanced_max:16677 - global_seqlen/mean:16667.125 - actor/entropy:0.25044116377830505 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.49410587549209595 - training/rollout_probs_diff_mean:0.005233317147940397 - training/rollout_probs_diff_std:0.013545987196266651 - training/rollout_actor_probs_pearson_corr:0.9983153343200684 - rollout_corr/rollout_is_mean:1.000156044960022 - rollout_corr/rollout_is_ratio_fraction_high:1.6923624571063556e-05 - rollout_corr/rollout_is_ratio_fraction_low:8.461812103632838e-05 - rollout_corr/rollout_is_max:2.701498508453369 - rollout_corr/rollout_is_min:0.2689496576786041 - rollout_corr/rollout_is_std:0.0415240079164505 - rollout_corr/rollout_is_eff_sample_size:0.9982794375291114 - rollout_corr/rollout_is_seq_mean:1.000076413154602 - rollout_corr/rollout_is_seq_std:0.003434262704104185 - rollout_corr/rollout_is_seq_max:1.0094366073608398 - rollout_corr/rollout_is_seq_min:0.9889258742332458 - rollout_corr/rollout_is_seq_max_deviation:0.01107412576675415 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3120325771160424) - rollout_corr/training_log_ppl:np.float64(0.26778477303741965) - rollout_corr/kl:np.float64(0.001165033413258243) - rollout_corr/k3_kl:np.float64(0.0011289213275631482) - rollout_corr/rollout_ppl:np.float64(1.3104448192752898) - rollout_corr/rollout_log_ppl:np.float64(0.26661973919544835) - rollout_corr/log_ppl_diff:np.float64(0.001165033841971308) - rollout_corr/log_ppl_abs_diff:np.float64(0.0029716190765611827) - rollout_corr/log_ppl_diff_max:np.float64(0.01145792007446289) - rollout_corr/log_ppl_diff_min:np.float64(-0.009343594312667847) - rollout_corr/ppl_ratio:np.float64(1.001172634307295) - rollout_corr/chi2_token:np.float64(0.002230182522907853) - rollout_corr/chi2_seq:np.float64(0.7144021939020604) - actor/pg_loss:np.float64(-2.578599378466606e-06) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011448644404481456) - actor/ppo_kl:np.float64(0.00031501110731468884) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.568739764392376) - perf/mfu/actor:np.float64(0.04933886842194326) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.97966003417969) - actor/lr:np.float64(1e-06) - training/global_step:76 - training/epoch:5 - critic/score/mean:0.43359375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.43359375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.007423124276101589 - critic/advantages/max:0.875 - critic/advantages/min:-0.75 - critic/returns/mean:0.007423124276101589 - critic/returns/max:0.875 - critic/returns/min:-0.75 - response_length/mean:230.81640625 - response_length/max:1078.0 - response_length/min:106.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:230.81640625 - response_length_non_aborted/max:1078.0 - response_length_non_aborted/min:106.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:290.03125 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.485520422458649e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2769716065376997) - timing_s/agent_loop/generate_sequences/max:np.float64(6.788432452827692) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.423513734356675) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.788432452827692) - timing_s/agent_loop/slowest/tool_calls:np.float64
+step:77 - global_seqlen/min:14979 - global_seqlen/max:24145 - global_seqlen/minmax_diff:9166 - global_seqlen/balanced_min:17784 - global_seqlen/balanced_max:17827 - global_seqlen/mean:17795.375 - actor/entropy:0.2697010934352875 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3591962456703186 - training/rollout_probs_diff_mean:0.005411113146692514 - training/rollout_probs_diff_std:0.013911079615354538 - training/rollout_actor_probs_pearson_corr:0.9983001947402954 - rollout_corr/rollout_is_mean:0.9998016357421875 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:0.00011526238813530654 - rollout_corr/rollout_is_max:1.7653801441192627 - rollout_corr/rollout_is_min:0.3611298203468323 - rollout_corr/rollout_is_std:0.04165085032582283 - rollout_corr/rollout_is_eff_sample_size:0.9982674389310602 - rollout_corr/rollout_is_seq_mean:0.999787449836731 - rollout_corr/rollout_is_seq_std:0.0032496904022991657 - rollout_corr/rollout_is_seq_max:1.0096925497055054 - rollout_corr/rollout_is_seq_min:0.989059329032898 - rollout_corr/rollout_is_seq_max_deviation:0.01094067096710205 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3272224520333111) - rollout_corr/training_log_ppl:np.float64(0.2796602731687017) - rollout_corr/kl:np.float64(0.0015744009766365252) - rollout_corr/k3_kl:np.float64(0.00129540834745967) - rollout_corr/rollout_ppl:np.float64(1.325115962419659) - rollout_corr/rollout_log_ppl:np.float64(0.2780858727055602) - rollout_corr/log_ppl_diff:np.float64(0.001574400463141501) - rollout_corr/log_ppl_abs_diff:np.float64(0.003008296713232994) - rollout_corr/log_ppl_diff_max:np.float64(0.012766122817993164) - rollout_corr/log_ppl_diff_min:np.float64(-0.008344709873199463) - rollout_corr/ppl_ratio:np.float64(1.001582418801263) - rollout_corr/chi2_token:np.float64(0.002242328831925988) - rollout_corr/chi2_seq:np.float64(0.23463519662618637) - actor/pg_loss:np.float64(0.00029490748420357704) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0014422528074646834) - actor/ppo_kl:np.float64(0.0004096870738550251) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5797202438116074) - perf/mfu/actor:np.float64(0.05305596361026153) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.87368392944336) - actor/lr:np.float64(1e-06) - training/global_step:77 - training/epoch:5 - critic/score/mean:0.6171875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.6171875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.008615863509476185 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.008615863509476185 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:237.23046875 - response_length/max:1068.0 - response_length/min:88.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:237.23046875 - response_length_non_aborted/max:1068.0 - response_length_non_aborted/min:88.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:318.875 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.313225746154785e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0067482870072126) - timing_s/agent_loop/generate_sequences/max:np.float64(6.926735129207373) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.472584339346213) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.926735129207373) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent_lo
+step:78 - global_seqlen/min:15123 - global_seqlen/max:23777 - global_seqlen/minmax_diff:8654 - global_seqlen/balanced_min:18759 - global_seqlen/balanced_max:18771 - global_seqlen/mean:18768.625 - actor/entropy:0.25392431020736694 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.6629616618156433 - training/rollout_probs_diff_mean:0.00486655393615365 - training/rollout_probs_diff_std:0.012811429798603058 - training/rollout_actor_probs_pearson_corr:0.9985421299934387 - rollout_corr/rollout_is_mean:1.0001100301742554 - rollout_corr/rollout_is_ratio_fraction_high:1.5132027328945696e-05 - rollout_corr/rollout_is_ratio_fraction_low:6.0528109315782785e-05 - rollout_corr/rollout_is_max:2.284006357192993 - rollout_corr/rollout_is_min:0.19336040318012238 - rollout_corr/rollout_is_std:0.03998841717839241 - rollout_corr/rollout_is_eff_sample_size:0.9984037172013397 - rollout_corr/rollout_is_seq_mean:1.000152587890625 - rollout_corr/rollout_is_seq_std:0.003139234147965908 - rollout_corr/rollout_is_seq_max:1.0102607011795044 - rollout_corr/rollout_is_seq_min:0.9885665774345398 - rollout_corr/rollout_is_seq_max_deviation:0.011433422565460205 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3306259801611304) - rollout_corr/training_log_ppl:np.float64(0.2809669480775483) - rollout_corr/kl:np.float64(0.0013480030079620064) - rollout_corr/k3_kl:np.float64(0.0014645395496017954) - rollout_corr/rollout_ppl:np.float64(1.328828570432961) - rollout_corr/rollout_log_ppl:np.float64(0.2796189447399229) - rollout_corr/log_ppl_diff:np.float64(0.0013480033376254141) - rollout_corr/log_ppl_abs_diff:np.float64(0.003176851896569133) - rollout_corr/log_ppl_diff_max:np.float64(0.02776774764060974) - rollout_corr/log_ppl_diff_min:np.float64(-0.009948790073394775) - rollout_corr/ppl_ratio:np.float64(1.0013591742608696) - rollout_corr/chi2_token:np.float64(0.0036836275830864906) - rollout_corr/chi2_seq:np.float64(2.4141830562148243) - actor/pg_loss:np.float64(0.0004103210521861911) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0027973866708634887) - actor/ppo_kl:np.float64(0.0005619901072861921) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7025328427553177) - perf/mfu/actor:np.float64(0.054799120913851654) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.03145599365234) - actor/lr:np.float64(1e-06) - training/global_step:78 - training/epoch:5 - critic/score/mean:0.43359375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.43359375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.018179239705204964 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.018179239705204964 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:258.14453125 - response_length/max:1046.0 - response_length/min:70.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:258.14453125 - response_length_non_aborted/max:1046.0 - response_length_non_aborted/min:70.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:328.375 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00010188482701778412 - timing_s/agent_loop/generate_sequences/min:np.float64(0.7525202222168446) - timing_s/agent_loop/generate_sequences/max:np.float64(6.806002235040069) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.3579090451748925) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.806002235040069) - timing_s/agent_loop/slowest/tool_calls:np.flo
+step:79 - global_seqlen/min:12334 - global_seqlen/max:22428 - global_seqlen/minmax_diff:10094 - global_seqlen/balanced_min:17635 - global_seqlen/balanced_max:17782 - global_seqlen/mean:17654.5 - actor/entropy:0.27384743094444275 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2592231035232544 - training/rollout_probs_diff_mean:0.005185640417039394 - training/rollout_probs_diff_std:0.012722447514533997 - training/rollout_actor_probs_pearson_corr:0.9985835552215576 - rollout_corr/rollout_is_mean:0.9998155236244202 - rollout_corr/rollout_is_ratio_fraction_high:1.6117593986564316e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.223518797312863e-05 - rollout_corr/rollout_is_max:2.173496961593628 - rollout_corr/rollout_is_min:0.37063977122306824 - rollout_corr/rollout_is_std:0.039584673941135406 - rollout_corr/rollout_is_eff_sample_size:0.9984348513727722 - rollout_corr/rollout_is_seq_mean:0.9998263120651245 - rollout_corr/rollout_is_seq_std:0.003011947264894843 - rollout_corr/rollout_is_seq_max:1.014046549797058 - rollout_corr/rollout_is_seq_min:0.9915293455123901 - rollout_corr/rollout_is_seq_max_deviation:0.014046549797058105 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.340323363430798) - rollout_corr/training_log_ppl:np.float64(0.2888445410062559) - rollout_corr/kl:np.float64(0.0010857870700355932) - rollout_corr/k3_kl:np.float64(0.0011207945151454624) - rollout_corr/rollout_ppl:np.float64(1.3388538309372962) - rollout_corr/rollout_log_ppl:np.float64(0.2877587534894701) - rollout_corr/log_ppl_diff:np.float64(0.0010857875167857856) - rollout_corr/log_ppl_abs_diff:np.float64(0.002984179329359904) - rollout_corr/log_ppl_diff_max:np.float64(0.013294041156768799) - rollout_corr/log_ppl_diff_min:np.float64(-0.009193763136863708) - rollout_corr/ppl_ratio:np.float64(1.001093138474971) - rollout_corr/chi2_token:np.float64(0.0023503343109041452) - rollout_corr/chi2_seq:np.float64(0.57042295509018) - actor/pg_loss:np.float64(-0.00013655168004333973) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0013187846618620824) - actor/ppo_kl:np.float64(3.9207486743464415e-05) - actor/pg_clipfrac_lower:np.float64(2.0559211407089606e-05) - actor/grad_norm:np.float64(0.6367911696434021) - perf/mfu/actor:np.float64(0.05222786833305854) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.97899627685547) - actor/lr:np.float64(1e-06) - training/global_step:79 - training/epoch:5 - critic/score/mean:0.41015625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.41015625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.013035103678703308 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.013035103678703308 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:242.359375 - response_length/max:1251.0 - response_length/min:92.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:242.359375 - response_length_non_aborted/max:1251.0 - response_length_non_aborted/min:92.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:309.34375 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00014719553291797638 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1327050235122442) - timing_s/agent_loop/generate_sequences/max:np.float64(7.875947218388319) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.5553116282462724) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.875947218388319) - timing_s/agent_loop/slowest/
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 80}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score for a specific protein sequence can be determined using computational methods such as those implemented in tools like I-TASSER, DSSP, or other protein structure prediction software. These tools calculate various scores based on the predicted 3D structure of the protein. The score provided (e.g., -0.01, 1.69, 2.49, 0.45) typically reflects the stability of the predicted structure relative to a reference structure.
+
+Given the sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR', it is likely that this sequence has been analyzed using a specific computational tool. Based on known results for similar sequences and the range of scores provided, the folding stability score for this sequence is most likely 0.45.
+</reasoning>
+<answer>
+D
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] D
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_80
+step:80 - global_seqlen/min:12878 - global_seqlen/max:22644 - global_seqlen/minmax_diff:9766 - global_seqlen/balanced_min:16682 - global_seqlen/balanced_max:16689 - global_seqlen/mean:16687.125 - actor/entropy:0.31366831064224243 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3867309093475342 - training/rollout_probs_diff_mean:0.005574310198426247 - training/rollout_probs_diff_std:0.012939517386257648 - training/rollout_actor_probs_pearson_corr:0.9986551403999329 - rollout_corr/rollout_is_mean:0.9999090433120728 - rollout_corr/rollout_is_ratio_fraction_high:1.7949132598005235e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.589826519601047e-05 - rollout_corr/rollout_is_max:2.8355820178985596 - rollout_corr/rollout_is_min:0.44648870825767517 - rollout_corr/rollout_is_std:0.041322559118270874 - rollout_corr/rollout_is_eff_sample_size:0.9982948816937318 - rollout_corr/rollout_is_seq_mean:0.9998655319213867 - rollout_corr/rollout_is_seq_std:0.003264273749664426 - rollout_corr/rollout_is_seq_max:1.0161584615707397 - rollout_corr/rollout_is_seq_min:0.9922093152999878 - rollout_corr/rollout_is_seq_max_deviation:0.016158461570739746 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3791068121790886) - rollout_corr/training_log_ppl:np.float64(0.31735638427198865) - rollout_corr/kl:np.float64(0.0011907178431336263) - rollout_corr/k3_kl:np.float64(0.0012447628572545) - rollout_corr/rollout_ppl:np.float64(1.3774029430933297) - rollout_corr/rollout_log_ppl:np.float64(0.3161656655429397) - rollout_corr/log_ppl_diff:np.float64(0.0011907187290489674) - rollout_corr/log_ppl_abs_diff:np.float64(0.0031865338096395135) - rollout_corr/log_ppl_diff_max:np.float64(0.015649110078811646) - rollout_corr/log_ppl_diff_min:np.float64(-0.012567013502120972) - rollout_corr/ppl_ratio:np.float64(1.0011996023822576) - rollout_corr/chi2_token:np.float64(0.0025439781602472067) - rollout_corr/chi2_seq:np.float64(0.696196470875293) - actor/pg_loss:np.float64(0.0002089452464133501) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0019131674780510366) - actor/ppo_kl:np.float64(0.0001621537604137302) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6633199006319046) - perf/mfu/actor:np.float64(0.049463591662902) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.01802444458008) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.41625) - val-aux/sciknoweval/reward/std@16:np.float64(0.3144664633085353) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.55652) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.28145093411213773) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.27288) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.2558732573095932) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.4132) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.3137503776392298) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.6777599999999998) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.19852437956426017) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.17295999999999995) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.1598024868585598) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.41869999999999996) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.2850227986027261) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.74804) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.10224935564981508) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.12108) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.07521408052792339) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.41981999999999997) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.2487908426641159) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.7791199999999999) - val-aux/sciknoweval/reward/best@16/std:np.float64(0.0463722632617
+step:81 - global_seqlen/min:15559 - global_seqlen/max:22877 - global_seqlen/minmax_diff:7318 - global_seqlen/balanced_min:19228 - global_seqlen/balanced_max:19441 - global_seqlen/mean:19257.0 - actor/entropy:0.2643565535545349 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2313688099384308 - training/rollout_probs_diff_mean:0.0049353111535310745 - training/rollout_probs_diff_std:0.012572631239891052 - training/rollout_actor_probs_pearson_corr:0.9985886812210083 - rollout_corr/rollout_is_mean:1.0000520944595337 - rollout_corr/rollout_is_ratio_fraction_high:1.5707464626757428e-05 - rollout_corr/rollout_is_ratio_fraction_low:0.0 - rollout_corr/rollout_is_max:2.8178207874298096 - rollout_corr/rollout_is_min:0.5387734174728394 - rollout_corr/rollout_is_std:0.03955530375242233 - rollout_corr/rollout_is_eff_sample_size:0.9984379411292814 - rollout_corr/rollout_is_seq_mean:1.000102162361145 - rollout_corr/rollout_is_seq_std:0.0028133587911725044 - rollout_corr/rollout_is_seq_max:1.0072697401046753 - rollout_corr/rollout_is_seq_min:0.9939631819725037 - rollout_corr/rollout_is_seq_max_deviation:0.007269740104675293 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.344506647437811) - rollout_corr/training_log_ppl:np.float64(0.29159667881322093) - rollout_corr/kl:np.float64(0.0010141099861371572) - rollout_corr/k3_kl:np.float64(0.0013735205002944895) - rollout_corr/rollout_ppl:np.float64(1.3431055014953017) - rollout_corr/rollout_log_ppl:np.float64(0.2905825663037831) - rollout_corr/log_ppl_diff:np.float64(0.0010141125094378367) - rollout_corr/log_ppl_abs_diff:np.float64(0.0028742786817019805) - rollout_corr/log_ppl_diff_max:np.float64(0.01934957504272461) - rollout_corr/log_ppl_diff_min:np.float64(-0.010468631982803345) - rollout_corr/ppl_ratio:np.float64(1.0010220303665847) - rollout_corr/chi2_token:np.float64(0.003484314540401101) - rollout_corr/chi2_seq:np.float64(0.5838833905290812) - actor/pg_loss:np.float64(0.00048545573372393847) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0023849409749345796) - actor/ppo_kl:np.float64(0.00024174882969418832) - actor/pg_clipfrac_lower:np.float64(3.98596930608619e-05) - actor/grad_norm:np.float64(0.7806772142648697) - perf/mfu/actor:np.float64(0.05648842110660632) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.96889114379883) - actor/lr:np.float64(1e-06) - training/global_step:81 - training/epoch:5 - critic/score/mean:0.5390625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.5390625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:3.9268659747904167e-05 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:3.9268659747904167e-05 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:248.6875 - response_length/max:1421.0 - response_length/min:69.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:248.6875 - response_length_non_aborted/max:1421.0 - response_length_non_aborted/min:69.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:353.09375 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00015094690024852753 - timing_s/agent_loop/generate_sequences/min:np.float64(0.7632148675620556) - timing_s/agent_loop/generate_sequences/max:np.float64(8.421508487313986) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.538855082115333) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(8.421508487313986) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.
+step:82 - global_seqlen/min:13327 - global_seqlen/max:21973 - global_seqlen/minmax_diff:8646 - global_seqlen/balanced_min:17016 - global_seqlen/balanced_max:17032 - global_seqlen/mean:17027.5 - actor/entropy:0.2781171500682831 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.25190842151641846 - training/rollout_probs_diff_mean:0.005356854293495417 - training/rollout_probs_diff_std:0.013084305450320244 - training/rollout_actor_probs_pearson_corr:0.9985019564628601 - rollout_corr/rollout_is_mean:0.9998621344566345 - rollout_corr/rollout_is_ratio_fraction_high:1.7285487047047354e-05 - rollout_corr/rollout_is_ratio_fraction_low:8.642743341624737e-05 - rollout_corr/rollout_is_max:2.251864433288574 - rollout_corr/rollout_is_min:0.29135599732398987 - rollout_corr/rollout_is_std:0.0413888581097126 - rollout_corr/rollout_is_eff_sample_size:0.99828953558267 - rollout_corr/rollout_is_seq_mean:0.999849796295166 - rollout_corr/rollout_is_seq_std:0.003499638754874468 - rollout_corr/rollout_is_seq_max:1.008987307548523 - rollout_corr/rollout_is_seq_min:0.9893508553504944 - rollout_corr/rollout_is_seq_max_deviation:0.010649144649505615 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3479181840084493) - rollout_corr/training_log_ppl:np.float64(0.293532015173696) - rollout_corr/kl:np.float64(0.0012917956987497092) - rollout_corr/k3_kl:np.float64(0.0012352343883321737) - rollout_corr/rollout_ppl:np.float64(1.346091446466744) - rollout_corr/rollout_log_ppl:np.float64(0.29224021729896776) - rollout_corr/log_ppl_diff:np.float64(0.0012917978747282177) - rollout_corr/log_ppl_abs_diff:np.float64(0.0030594941636081785) - rollout_corr/log_ppl_diff_max:np.float64(0.012573987245559692) - rollout_corr/log_ppl_diff_min:np.float64(-0.011092990636825562) - rollout_corr/ppl_ratio:np.float64(1.0012995367869735) - rollout_corr/chi2_token:np.float64(0.0022806792985647917) - rollout_corr/chi2_seq:np.float64(1.2513767953496426) - actor/pg_loss:np.float64(-1.996220089495182e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011945247019866656) - actor/ppo_kl:np.float64(0.00018004048818909268) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6068511605262756) - perf/mfu/actor:np.float64(0.04981626783172659) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.0265121459961) - actor/lr:np.float64(1e-06) - training/global_step:82 - training/epoch:5 - critic/score/mean:0.59375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.59375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.005667479243129492 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.005667479243129492 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:225.984375 - response_length/max:982.0 - response_length/min:93.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:225.984375 - response_length_non_aborted/max:982.0 - response_length_non_aborted/min:93.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:306.125 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:4.120729863643646e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0436728745698929) - timing_s/agent_loop/generate_sequences/max:np.float64(6.355573097243905) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.3314835890123504) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.355573097243905) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timin
+step:83 - global_seqlen/min:12925 - global_seqlen/max:26465 - global_seqlen/minmax_diff:13540 - global_seqlen/balanced_min:17539 - global_seqlen/balanced_max:17863 - global_seqlen/mean:17587.5 - actor/entropy:0.2657328248023987 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2070435881614685 - training/rollout_probs_diff_mean:0.005034568719565868 - training/rollout_probs_diff_std:0.012614690698683262 - training/rollout_actor_probs_pearson_corr:0.9985789060592651 - rollout_corr/rollout_is_mean:1.000186562538147 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:4.767050268128514e-05 - rollout_corr/rollout_is_max:1.6093506813049316 - rollout_corr/rollout_is_min:0.31631869077682495 - rollout_corr/rollout_is_std:0.03947233781218529 - rollout_corr/rollout_is_eff_sample_size:0.9984449525707558 - rollout_corr/rollout_is_seq_mean:1.0001424551010132 - rollout_corr/rollout_is_seq_std:0.0030042307917028666 - rollout_corr/rollout_is_seq_max:1.0098767280578613 - rollout_corr/rollout_is_seq_min:0.9898255467414856 - rollout_corr/rollout_is_seq_max_deviation:0.010174453258514404 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3396281930617988) - rollout_corr/training_log_ppl:np.float64(0.28824585146503523) - rollout_corr/kl:np.float64(0.0013105916907552029) - rollout_corr/k3_kl:np.float64(0.0012636836211754598) - rollout_corr/rollout_ppl:np.float64(1.3378654848784208) - rollout_corr/rollout_log_ppl:np.float64(0.2869352596462704) - rollout_corr/log_ppl_diff:np.float64(0.0013105918187648058) - rollout_corr/log_ppl_abs_diff:np.float64(0.002559717104304582) - rollout_corr/log_ppl_diff_max:np.float64(0.012187853455543518) - rollout_corr/log_ppl_diff_min:np.float64(-0.008456051349639893) - rollout_corr/ppl_ratio:np.float64(1.0013164200354367) - rollout_corr/chi2_token:np.float64(0.0023376578465104103) - rollout_corr/chi2_seq:np.float64(0.27040194789879024) - actor/pg_loss:np.float64(0.00010856811422854662) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0011537321538526157) - actor/ppo_kl:np.float64(0.0005853857873887591) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6243516802787781) - perf/mfu/actor:np.float64(0.05163460793549127) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.9116096496582) - actor/lr:np.float64(1e-06) - training/global_step:83 - training/epoch:5 - critic/score/mean:0.5234375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.5234375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0019763396121561527 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0019763396121561527 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:245.828125 - response_length/max:1487.0 - response_length/min:93.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:245.828125 - response_length_non_aborted/max:1487.0 - response_length_non_aborted/min:93.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:303.78125 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:9.453855454921722e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1134148389101028) - timing_s/agent_loop/generate_sequences/max:np.float64(8.951366694644094) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.531380183805595) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(8.951366694644094) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s
+step:84 - global_seqlen/min:14425 - global_seqlen/max:25507 - global_seqlen/minmax_diff:11082 - global_seqlen/balanced_min:19094 - global_seqlen/balanced_max:19099 - global_seqlen/mean:19098.0 - actor/entropy:0.29119229316711426 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3042685389518738 - training/rollout_probs_diff_mean:0.005367741920053959 - training/rollout_probs_diff_std:0.012942570261657238 - training/rollout_actor_probs_pearson_corr:0.9985758066177368 - rollout_corr/rollout_is_mean:1.0001381635665894 - rollout_corr/rollout_is_ratio_fraction_high:1.4721469597134273e-05 - rollout_corr/rollout_is_ratio_fraction_low:2.9442939194268547e-05 - rollout_corr/rollout_is_max:2.259265661239624 - rollout_corr/rollout_is_min:0.32297244668006897 - rollout_corr/rollout_is_std:0.04044489935040474 - rollout_corr/rollout_is_eff_sample_size:0.9983673568485679 - rollout_corr/rollout_is_seq_mean:1.0002832412719727 - rollout_corr/rollout_is_seq_std:0.0029685506597161293 - rollout_corr/rollout_is_seq_max:1.0090413093566895 - rollout_corr/rollout_is_seq_min:0.9883840680122375 - rollout_corr/rollout_is_seq_max_deviation:0.011615931987762451 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3692081668414176) - rollout_corr/training_log_ppl:np.float64(0.31009084297693335) - rollout_corr/kl:np.float64(0.0009806591620620253) - rollout_corr/k3_kl:np.float64(0.0011354790972291084) - rollout_corr/rollout_ppl:np.float64(1.367792998906225) - rollout_corr/rollout_log_ppl:np.float64(0.30911018152255565) - rollout_corr/log_ppl_diff:np.float64(0.000980661454377696) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027504815079737455) - rollout_corr/log_ppl_diff_max:np.float64(0.018712282180786133) - rollout_corr/log_ppl_diff_min:np.float64(-0.009203359484672546) - rollout_corr/ppl_ratio:np.float64(1.0009879535064101) - rollout_corr/chi2_token:np.float64(0.0026025581173598766) - rollout_corr/chi2_seq:np.float64(1.0249379503075033) - actor/pg_loss:np.float64(0.0003060917370021343) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0015142915781325428) - actor/ppo_kl:np.float64(0.00034652300968218697) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5972811579704285) - perf/mfu/actor:np.float64(0.05583602661339712) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.96282958984375) - actor/lr:np.float64(1e-06) - training/global_step:84 - training/epoch:5 - critic/score/mean:0.4296875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4296875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0036454040091484785 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0036454040091484785 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:265.34375 - response_length/max:926.0 - response_length/min:104.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:265.34375 - response_length_non_aborted/max:926.0 - response_length_non_aborted/min:104.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:331.46875 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:2.8295442461967468e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1363158952444792) - timing_s/agent_loop/generate_sequences/max:np.float64(6.200030520558357) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.6375616744408035) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.200030520558357) - timing_s/agent_loop/slowest/tool_calls:np.fl
+step:85 - global_seqlen/min:13201 - global_seqlen/max:23013 - global_seqlen/minmax_diff:9812 - global_seqlen/balanced_min:17098 - global_seqlen/balanced_max:17114 - global_seqlen/mean:17108.75 - actor/entropy:0.3020118772983551 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.4086613655090332 - training/rollout_probs_diff_mean:0.005688806995749474 - training/rollout_probs_diff_std:0.01348093617707491 - training/rollout_actor_probs_pearson_corr:0.9984976053237915 - rollout_corr/rollout_is_mean:0.9999252557754517 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:4.976279888069257e-05 - rollout_corr/rollout_is_max:1.660659909248352 - rollout_corr/rollout_is_min:0.12563924491405487 - rollout_corr/rollout_is_std:0.04169733077287674 - rollout_corr/rollout_is_eff_sample_size:0.9982643502297281 - rollout_corr/rollout_is_seq_mean:0.9999344944953918 - rollout_corr/rollout_is_seq_std:0.0032068761065602303 - rollout_corr/rollout_is_seq_max:1.0129774808883667 - rollout_corr/rollout_is_seq_min:0.99162358045578 - rollout_corr/rollout_is_seq_max_deviation:0.0129774808883667 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3707473212853074) - rollout_corr/training_log_ppl:np.float64(0.3106351788446773) - rollout_corr/kl:np.float64(0.0010191974550188831) - rollout_corr/k3_kl:np.float64(0.0012126699965619991) - rollout_corr/rollout_ppl:np.float64(1.3693385659717023) - rollout_corr/rollout_log_ppl:np.float64(0.3096159797569271) - rollout_corr/log_ppl_diff:np.float64(0.0010191990877501667) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027312866295687854) - rollout_corr/log_ppl_diff_max:np.float64(0.01148831844329834) - rollout_corr/log_ppl_diff_min:np.float64(-0.009180232882499695) - rollout_corr/ppl_ratio:np.float64(1.001025541452691) - rollout_corr/chi2_token:np.float64(0.002835497260093689) - rollout_corr/chi2_seq:np.float64(0.9057672906201333) - actor/pg_loss:np.float64(-5.922233685851097e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001044121142513177) - actor/ppo_kl:np.float64(4.945502578124206e-06) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.4638519212603569) - perf/mfu/actor:np.float64(0.050201444400562734) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.93396759033203) - actor/lr:np.float64(1e-06) - training/global_step:85 - training/epoch:6 - critic/score/mean:0.5 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.5 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.006879706867039204 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.006879706867039204 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:235.4921875 - response_length/max:1005.0 - response_length/min:107.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:235.4921875 - response_length_non_aborted/max:1005.0 - response_length_non_aborted/min:107.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:299.15625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0004599113017320633 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2714347038418055) - timing_s/agent_loop/generate_sequences/max:np.float64(6.474359966814518) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.499616449684254) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.474359966814518) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent_loop/slowest/
+step:86 - global_seqlen/min:12509 - global_seqlen/max:25462 - global_seqlen/minmax_diff:12953 - global_seqlen/balanced_min:18292 - global_seqlen/balanced_max:18296 - global_seqlen/mean:18294.125 - actor/entropy:0.2849932610988617 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.35150495171546936 - training/rollout_probs_diff_mean:0.0054772403091192245 - training/rollout_probs_diff_std:0.013425457291305065 - training/rollout_actor_probs_pearson_corr:0.9984551668167114 - rollout_corr/rollout_is_mean:1.0001648664474487 - rollout_corr/rollout_is_ratio_fraction_high:4.6187244151951745e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.0791496101301163e-05 - rollout_corr/rollout_is_max:2.974616289138794 - rollout_corr/rollout_is_min:0.459360271692276 - rollout_corr/rollout_is_std:0.04225114360451698 - rollout_corr/rollout_is_eff_sample_size:0.9982186159276442 - rollout_corr/rollout_is_seq_mean:1.0002799034118652 - rollout_corr/rollout_is_seq_std:0.003540112404152751 - rollout_corr/rollout_is_seq_max:1.0164246559143066 - rollout_corr/rollout_is_seq_min:0.9909144639968872 - rollout_corr/rollout_is_seq_max_deviation:0.01642465591430664 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3678325335495174) - rollout_corr/training_log_ppl:np.float64(0.30802994256373495) - rollout_corr/kl:np.float64(0.0011918487436304304) - rollout_corr/k3_kl:np.float64(0.0014298067977733808) - rollout_corr/rollout_ppl:np.float64(1.366152125876397) - rollout_corr/rollout_log_ppl:np.float64(0.30683809251058847) - rollout_corr/log_ppl_diff:np.float64(0.0011918500531464815) - rollout_corr/log_ppl_abs_diff:np.float64(0.003217801859136671) - rollout_corr/log_ppl_diff_max:np.float64(0.010454237461090088) - rollout_corr/log_ppl_diff_min:np.float64(-0.01674443483352661) - rollout_corr/ppl_ratio:np.float64(1.0012002452276647) - rollout_corr/chi2_token:np.float64(0.003339178394526243) - rollout_corr/chi2_seq:np.float64(1.9388157108332962) - actor/pg_loss:np.float64(-0.0001976813655346632) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.001634979141272197) - actor/ppo_kl:np.float64(0.0004851831923424754) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5677638277411461) - perf/mfu/actor:np.float64(0.05395100018655371) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.07014083862305) - actor/lr:np.float64(1e-06) - training/global_step:86 - training/epoch:6 - critic/score/mean:0.63671875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.63671875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0019225439755246043 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.0019225439755246043 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:253.72265625 - response_length/max:792.0 - response_length/min:116.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:253.72265625 - response_length_non_aborted/max:792.0 - response_length_non_aborted/min:116.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:317.96875 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00015383213758468628 - timing_s/agent_loop/generate_sequences/min:np.float64(1.4205127842724323) - timing_s/agent_loop/generate_sequences/max:np.float64(5.750876966863871) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.6700041760486783) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(5.750876966863871) - timing_s/agent_loop/slowest/tool_calls:np.
+step:87 - global_seqlen/min:13933 - global_seqlen/max:20602 - global_seqlen/minmax_diff:6669 - global_seqlen/balanced_min:16687 - global_seqlen/balanced_max:16691 - global_seqlen/mean:16689.75 - actor/entropy:0.28263986110687256 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.23105812072753906 - training/rollout_probs_diff_mean:0.0056616635993123055 - training/rollout_probs_diff_std:0.013550084084272385 - training/rollout_actor_probs_pearson_corr:0.9984059929847717 - rollout_corr/rollout_is_mean:1.0001320838928223 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:5.4052106861490756e-05 - rollout_corr/rollout_is_max:1.8633671998977661 - rollout_corr/rollout_is_min:0.20320795476436615 - rollout_corr/rollout_is_std:0.04268762469291687 - rollout_corr/rollout_is_eff_sample_size:0.998181556398287 - rollout_corr/rollout_is_seq_mean:1.000098466873169 - rollout_corr/rollout_is_seq_std:0.00287253619171679 - rollout_corr/rollout_is_seq_max:1.0077333450317383 - rollout_corr/rollout_is_seq_min:0.9905746579170227 - rollout_corr/rollout_is_seq_max_deviation:0.009425342082977295 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3366769887506962) - rollout_corr/training_log_ppl:np.float64(0.2858038642152678) - rollout_corr/kl:np.float64(0.0010864866192696887) - rollout_corr/k3_kl:np.float64(0.001194243301597453) - rollout_corr/rollout_ppl:np.float64(1.3351997244171798) - rollout_corr/rollout_log_ppl:np.float64(0.28471737628569826) - rollout_corr/log_ppl_diff:np.float64(0.0010864879295695573) - rollout_corr/log_ppl_abs_diff:np.float64(0.0029874266765546054) - rollout_corr/log_ppl_diff_max:np.float64(0.014816895127296448) - rollout_corr/log_ppl_diff_min:np.float64(-0.01041361689567566) - rollout_corr/ppl_ratio:np.float64(1.0010941252112389) - rollout_corr/chi2_token:np.float64(0.002591905649751425) - rollout_corr/chi2_seq:np.float64(0.8676178343594074) - actor/pg_loss:np.float64(2.4948501959443092e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0008350495745617081) - actor/ppo_kl:np.float64(0.00021582131192943166) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.45856113731861115) - perf/mfu/actor:np.float64(0.04922829341564032) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.0233154296875) - actor/lr:np.float64(1e-06) - training/global_step:87 - training/epoch:6 - critic/score/mean:0.5234375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.5234375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.013735991902649403 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.013735991902649403 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:216.8046875 - response_length/max:808.0 - response_length/min:110.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:216.8046875 - response_length_non_aborted/max:808.0 - response_length_non_aborted/min:110.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:304.75 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:8.88928771018982e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.328505938872695) - timing_s/agent_loop/generate_sequences/max:np.float64(5.489037185907364) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.3260599871064187) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(5.489037185907364) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent
+step:88 - global_seqlen/min:13016 - global_seqlen/max:21866 - global_seqlen/minmax_diff:8850 - global_seqlen/balanced_min:18267 - global_seqlen/balanced_max:18275 - global_seqlen/mean:18273.375 - actor/entropy:0.2799515724182129 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5366853475570679 - training/rollout_probs_diff_mean:0.005581203382462263 - training/rollout_probs_diff_std:0.013881382532417774 - training/rollout_actor_probs_pearson_corr:0.9983412027359009 - rollout_corr/rollout_is_mean:1.0000699758529663 - rollout_corr/rollout_is_ratio_fraction_high:1.5643821825506166e-05 - rollout_corr/rollout_is_ratio_fraction_low:3.128764365101233e-05 - rollout_corr/rollout_is_max:2.0318520069122314 - rollout_corr/rollout_is_min:0.3905743956565857 - rollout_corr/rollout_is_std:0.042155105620622635 - rollout_corr/rollout_is_eff_sample_size:0.998226337009296 - rollout_corr/rollout_is_seq_mean:1.0000312328338623 - rollout_corr/rollout_is_seq_std:0.0031614850740879774 - rollout_corr/rollout_is_seq_max:1.0118353366851807 - rollout_corr/rollout_is_seq_min:0.9903686046600342 - rollout_corr/rollout_is_seq_max_deviation:0.011835336685180664 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.347297566011548) - rollout_corr/training_log_ppl:np.float64(0.29373508717981167) - rollout_corr/kl:np.float64(0.0011751465276006456) - rollout_corr/k3_kl:np.float64(0.001120794104167544) - rollout_corr/rollout_ppl:np.float64(1.3456657594069839) - rollout_corr/rollout_log_ppl:np.float64(0.29255994062987156) - rollout_corr/log_ppl_diff:np.float64(0.0011751465499401093) - rollout_corr/log_ppl_abs_diff:np.float64(0.0026922697434201837) - rollout_corr/log_ppl_diff_max:np.float64(0.011649519205093384) - rollout_corr/log_ppl_diff_min:np.float64(-0.010374203324317932) - rollout_corr/ppl_ratio:np.float64(1.001181423664093) - rollout_corr/chi2_token:np.float64(0.002159958705306053) - rollout_corr/chi2_seq:np.float64(1.1241237167268991) - actor/pg_loss:np.float64(0.00016208365559577942) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.000806990730779944) - actor/ppo_kl:np.float64(0.00024305590492046392) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.4049631729722023) - perf/mfu/actor:np.float64(0.05369724660460473) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.09536361694336) - actor/lr:np.float64(1e-06) - training/global_step:88 - training/epoch:6 - critic/score/mean:0.375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.01436884980648756 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.01436884980648756 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:249.69921875 - response_length/max:838.0 - response_length/min:109.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:249.69921875 - response_length_non_aborted/max:838.0 - response_length_non_aborted/min:109.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:321.34375 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00014184601604938507 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1800287831574678) - timing_s/agent_loop/generate_sequences/max:np.float64(5.7370752189308405) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.6392609360191273) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(5.7370752189308405) - timing_s/agent_loop/slowest/tool_calls:np.float64(0
+step:89 - global_seqlen/min:15884 - global_seqlen/max:25708 - global_seqlen/minmax_diff:9824 - global_seqlen/balanced_min:19388 - global_seqlen/balanced_max:19400 - global_seqlen/mean:19395.0 - actor/entropy:0.2626587748527527 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.9052478075027466 - training/rollout_probs_diff_mean:0.00535238441079855 - training/rollout_probs_diff_std:0.013822822831571102 - training/rollout_actor_probs_pearson_corr:0.9982951879501343 - rollout_corr/rollout_is_mean:0.9997128248214722 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:6.095330900279805e-05 - rollout_corr/rollout_is_max:1.9251736402511597 - rollout_corr/rollout_is_min:0.07634969055652618 - rollout_corr/rollout_is_std:0.04105702042579651 - rollout_corr/rollout_is_eff_sample_size:0.9983161479024087 - rollout_corr/rollout_is_seq_mean:0.9995173215866089 - rollout_corr/rollout_is_seq_std:0.0029820753261446953 - rollout_corr/rollout_is_seq_max:1.0083783864974976 - rollout_corr/rollout_is_seq_min:0.9902425408363342 - rollout_corr/rollout_is_seq_max_deviation:0.009757459163665771 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3277254402637482) - rollout_corr/training_log_ppl:np.float64(0.2786243225564249) - rollout_corr/kl:np.float64(0.0015448006739369191) - rollout_corr/k3_kl:np.float64(0.0011937761133822278) - rollout_corr/rollout_ppl:np.float64(1.3255531983450055) - rollout_corr/rollout_log_ppl:np.float64(0.2770795207761694) - rollout_corr/log_ppl_diff:np.float64(0.0015448017802555114) - rollout_corr/log_ppl_abs_diff:np.float64(0.0028917556919623166) - rollout_corr/log_ppl_diff_max:np.float64(0.015002906322479248) - rollout_corr/log_ppl_diff_min:np.float64(-0.009606748819351196) - rollout_corr/ppl_ratio:np.float64(1.0015520232263952) - rollout_corr/chi2_token:np.float64(0.001597132533788681) - rollout_corr/chi2_seq:np.float64(0.5050114756450057) - actor/pg_loss:np.float64(0.0002184249460697174) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0010780754209918086) - actor/ppo_kl:np.float64(0.00011659164282207257) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.43044082075357437) - perf/mfu/actor:np.float64(0.056737380618191155) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.99333953857422) - actor/lr:np.float64(1e-06) - training/global_step:89 - training/epoch:6 - critic/score/mean:0.58984375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.58984375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.013847829774022102 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.013847829774022102 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:256.34375 - response_length/max:944.0 - response_length/min:121.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:256.34375 - response_length_non_aborted/max:944.0 - response_length_non_aborted/min:121.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:349.75 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:3.797747194766998e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.5112595465034246) - timing_s/agent_loop/generate_sequences/max:np.float64(6.177097799256444) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.6719028287552646) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.177097799256444) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/age
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 90}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score for a protein sequence can be determined using computational tools such as I-TAFOR, DSSP, or other protein structure prediction software. These tools calculate the stability of a protein's folded structure based on the sequence and predicted three-dimensional structure. The score provided in the question, "-0.01", is likely derived from such computational analysis. Scores close to zero indicate that the protein sequence has a stable fold, while positive scores may indicate higher stability. Based on the given sequence and typical computational results, the folding stability score is most likely -0.01.
+</reasoning>
+<answer>
+A
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] A
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_90
+step:90 - global_seqlen/min:15076 - global_seqlen/max:22705 - global_seqlen/minmax_diff:7629 - global_seqlen/balanced_min:18281 - global_seqlen/balanced_max:18304 - global_seqlen/mean:18298.25 - actor/entropy:0.25007060170173645 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3519518971443176 - training/rollout_probs_diff_mean:0.005134246312081814 - training/rollout_probs_diff_std:0.013177147135138512 - training/rollout_actor_probs_pearson_corr:0.998388409614563 - rollout_corr/rollout_is_mean:0.9999549388885498 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:7.610581815242767e-05 - rollout_corr/rollout_is_max:1.6880807876586914 - rollout_corr/rollout_is_min:0.44652095437049866 - rollout_corr/rollout_is_std:0.04015500843524933 - rollout_corr/rollout_is_eff_sample_size:0.9983898144007888 - rollout_corr/rollout_is_seq_mean:0.9998966455459595 - rollout_corr/rollout_is_seq_std:0.0029496627394109964 - rollout_corr/rollout_is_seq_max:1.0089014768600464 - rollout_corr/rollout_is_seq_min:0.9883391261100769 - rollout_corr/rollout_is_seq_max_deviation:0.011660873889923096 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.315980046056211) - rollout_corr/training_log_ppl:np.float64(0.2705623500805814) - rollout_corr/kl:np.float64(0.0011479333818513737) - rollout_corr/k3_kl:np.float64(0.0010572865134577114) - rollout_corr/rollout_ppl:np.float64(1.314408981706947) - rollout_corr/rollout_log_ppl:np.float64(0.26941441488452256) - rollout_corr/log_ppl_diff:np.float64(0.0011479351960588247) - rollout_corr/log_ppl_abs_diff:np.float64(0.0027164410275872797) - rollout_corr/log_ppl_diff_max:np.float64(0.013363391160964966) - rollout_corr/log_ppl_diff_min:np.float64(-0.008001953363418579) - rollout_corr/ppl_ratio:np.float64(1.0011545014567673) - rollout_corr/chi2_token:np.float64(0.0019355297554284334) - rollout_corr/chi2_seq:np.float64(0.5621733826119453) - actor/pg_loss:np.float64(-3.921461757272482e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0012568379429467313) - actor/ppo_kl:np.float64(0.00014396155199847271) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.4547141119837761) - perf/mfu/actor:np.float64(0.05313581446106291) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.05733489990234) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.4675) - val-aux/sciknoweval/reward/std@16:np.float64(0.28097887193097565) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.59294) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.22754928387157947) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.34464) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.2509343043440332) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.4650000000000001) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.28133902959180757) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.6816800000000001) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.14720983664825582) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.23678000000000002) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.17523408016078515) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.48656) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.24240246066242588) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.73372) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.08196201028513199) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.17182000000000003) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.09563785531224381) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.5071399999999999) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.19410742505337317) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.7635599999999999) - val-aux/sciknoweval/reward/best@16/std:np.float64(0.0411093715413
+step:91 - global_seqlen/min:13664 - global_seqlen/max:28551 - global_seqlen/minmax_diff:14887 - global_seqlen/balanced_min:19769 - global_seqlen/balanced_max:19785 - global_seqlen/mean:19781.5 - actor/entropy:0.25078386068344116 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.43638211488723755 - training/rollout_probs_diff_mean:0.004951483104377985 - training/rollout_probs_diff_std:0.013005034998059273 - training/rollout_actor_probs_pearson_corr:0.9984056949615479 - rollout_corr/rollout_is_mean:0.9997699856758118 - rollout_corr/rollout_is_ratio_fraction_high:2.716062772378791e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.432125544757582e-05 - rollout_corr/rollout_is_max:2.0405869483947754 - rollout_corr/rollout_is_min:0.43364647030830383 - rollout_corr/rollout_is_std:0.0399608314037323 - rollout_corr/rollout_is_eff_sample_size:0.9984049054928523 - rollout_corr/rollout_is_seq_mean:0.9997459053993225 - rollout_corr/rollout_is_seq_std:0.002915602643042803 - rollout_corr/rollout_is_seq_max:1.009341835975647 - rollout_corr/rollout_is_seq_min:0.9896574020385742 - rollout_corr/rollout_is_seq_max_deviation:0.010342597961425781 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3241292075254023) - rollout_corr/training_log_ppl:np.float64(0.2761981814110186) - rollout_corr/kl:np.float64(0.0012420743942662682) - rollout_corr/k3_kl:np.float64(0.0012305704847221932) - rollout_corr/rollout_ppl:np.float64(1.3224302087910473) - rollout_corr/rollout_log_ppl:np.float64(0.274956106179161) - rollout_corr/log_ppl_diff:np.float64(0.0012420752318575978) - rollout_corr/log_ppl_abs_diff:np.float64(0.0025912187120411545) - rollout_corr/log_ppl_diff_max:np.float64(0.014826744794845581) - rollout_corr/log_ppl_diff_min:np.float64(-0.007786452770233154) - rollout_corr/ppl_ratio:np.float64(1.0012476784177125) - rollout_corr/chi2_token:np.float64(0.0025545565877109766) - rollout_corr/chi2_seq:np.float64(0.8443398529198021) - actor/pg_loss:np.float64(-0.00023542286362498999) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0013123745520715602) - actor/ppo_kl:np.float64(7.698053936522342e-05) - actor/pg_clipfrac_lower:np.float64(1.1660447853500955e-05) - actor/grad_norm:np.float64(0.4663517475128174) - perf/mfu/actor:np.float64(0.05740788145887977) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.15924453735352) - actor/lr:np.float64(1e-06) - training/global_step:91 - training/epoch:6 - critic/score/mean:0.53515625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.53515625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0005907436716370285 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.0005907436716370285 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:287.640625 - response_length/max:1198.0 - response_length/min:107.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:287.640625 - response_length_non_aborted/max:1198.0 - response_length_non_aborted/min:107.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:330.53125 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:2.8155744075775146e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.3114315513521433) - timing_s/agent_loop/generate_sequences/max:np.float64(7.799869127571583) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.8764019390873727) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.799869127571583) - timing_s/agent_loop/slow
+step:92 - global_seqlen/min:15474 - global_seqlen/max:25697 - global_seqlen/minmax_diff:10223 - global_seqlen/balanced_min:20472 - global_seqlen/balanced_max:20489 - global_seqlen/mean:20485.25 - actor/entropy:0.25056755542755127 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.5584663152694702 - training/rollout_probs_diff_mean:0.004964837804436684 - training/rollout_probs_diff_std:0.013341370038688183 - training/rollout_actor_probs_pearson_corr:0.9983838200569153 - rollout_corr/rollout_is_mean:0.9999274015426636 - rollout_corr/rollout_is_ratio_fraction_high:4.238126348354854e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.650835009873845e-05 - rollout_corr/rollout_is_max:3.1624722480773926 - rollout_corr/rollout_is_min:0.20737510919570923 - rollout_corr/rollout_is_std:0.0406433641910553 - rollout_corr/rollout_is_eff_sample_size:0.9983503658269968 - rollout_corr/rollout_is_seq_mean:0.9998391270637512 - rollout_corr/rollout_is_seq_std:0.0029326851945370436 - rollout_corr/rollout_is_seq_max:1.0081546306610107 - rollout_corr/rollout_is_seq_min:0.9876515865325928 - rollout_corr/rollout_is_seq_max_deviation:0.012348413467407227 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3244820055551827) - rollout_corr/training_log_ppl:np.float64(0.27609299635514617) - rollout_corr/kl:np.float64(0.0011841020946801173) - rollout_corr/k3_kl:np.float64(0.0013603362282310627) - rollout_corr/rollout_ppl:np.float64(1.3228559507988393) - rollout_corr/rollout_log_ppl:np.float64(0.2749088934651809) - rollout_corr/log_ppl_diff:np.float64(0.0011841028899652883) - rollout_corr/log_ppl_abs_diff:np.float64(0.0026790284464368597) - rollout_corr/log_ppl_diff_max:np.float64(0.01207616925239563) - rollout_corr/log_ppl_diff_min:np.float64(-0.0074624717235565186) - rollout_corr/ppl_ratio:np.float64(1.0011904705315828) - rollout_corr/chi2_token:np.float64(0.0030524577014148235) - rollout_corr/chi2_seq:np.float64(0.5289049397688359) - actor/pg_loss:np.float64(0.00013589137233793736) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002009198718951666) - actor/ppo_kl:np.float64(3.7602182644036475e-05) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.613745704293251) - perf/mfu/actor:np.float64(0.05991881909984166) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.0025405883789) - actor/lr:np.float64(1e-06) - training/global_step:92 - training/epoch:6 - critic/score/mean:0.65234375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.65234375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.012523663230240345 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.012523663230240345 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:276.5078125 - response_length/max:1210.0 - response_length/min:112.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:276.5078125 - response_length_non_aborted/max:1210.0 - response_length_non_aborted/min:112.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:363.65625 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0001337248831987381 - timing_s/agent_loop/generate_sequences/min:np.float64(1.3419308010488749) - timing_s/agent_loop/generate_sequences/max:np.float64(7.376888804137707) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.7873674428483355) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.376888804137707) - timing_s/agent_loop/slowest/tool_calls:np
+step:93 - global_seqlen/min:13545 - global_seqlen/max:25655 - global_seqlen/minmax_diff:12110 - global_seqlen/balanced_min:19112 - global_seqlen/balanced_max:19881 - global_seqlen/mean:19214.125 - actor/entropy:0.23102349042892456 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3046841025352478 - training/rollout_probs_diff_mean:0.004829892423003912 - training/rollout_probs_diff_std:0.013052316382527351 - training/rollout_actor_probs_pearson_corr:0.9983353018760681 - rollout_corr/rollout_is_mean:0.9999238848686218 - rollout_corr/rollout_is_ratio_fraction_high:1.537870048196055e-05 - rollout_corr/rollout_is_ratio_fraction_low:7.689350604778156e-05 - rollout_corr/rollout_is_max:2.5575833320617676 - rollout_corr/rollout_is_min:0.3091418743133545 - rollout_corr/rollout_is_std:0.03926188126206398 - rollout_corr/rollout_is_eff_sample_size:0.9984606395421983 - rollout_corr/rollout_is_seq_mean:0.9999052882194519 - rollout_corr/rollout_is_seq_std:0.003070805687457323 - rollout_corr/rollout_is_seq_max:1.0085487365722656 - rollout_corr/rollout_is_seq_min:0.9925494194030762 - rollout_corr/rollout_is_seq_max_deviation:0.008548736572265625 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.298731068149209) - rollout_corr/training_log_ppl:np.float64(0.2568639310920844) - rollout_corr/kl:np.float64(0.001255048227367439) - rollout_corr/k3_kl:np.float64(0.001259192493506589) - rollout_corr/rollout_ppl:np.float64(1.297051596455276) - rollout_corr/rollout_log_ppl:np.float64(0.25560888355539646) - rollout_corr/log_ppl_diff:np.float64(0.0012550475366879255) - rollout_corr/log_ppl_abs_diff:np.float64(0.002867596718715504) - rollout_corr/log_ppl_diff_max:np.float64(0.02118036150932312) - rollout_corr/log_ppl_diff_min:np.float64(-0.007840782403945923) - rollout_corr/ppl_ratio:np.float64(1.0012623802758753) - rollout_corr/chi2_token:np.float64(0.002499846974387765) - rollout_corr/chi2_seq:np.float64(0.36207995703443885) - actor/pg_loss:np.float64(-4.324305336922407e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0018963009379149298) - actor/ppo_kl:np.float64(0.0002913880196118157) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6442021429538727) - perf/mfu/actor:np.float64(0.05641402480565709) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.02770233154297) - actor/lr:np.float64(1e-06) - training/global_step:93 - training/epoch:6 - critic/score/mean:0.484375 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.484375 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.014271434396505356 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.014271434396505356 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:254.00390625 - response_length/max:2020.0 - response_length/min:111.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:254.00390625 - response_length_non_aborted/max:2020.0 - response_length_non_aborted/min:111.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:346.4375 - prompt_length/max:500.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00012409500777721405 - timing_s/agent_loop/generate_sequences/min:np.float64(0.9811534341424704) - timing_s/agent_loop/generate_sequences/max:np.float64(11.214596100151539) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.4594217599369586) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(11.214596100151539) - timing_s/agent_loop/slowest/tool_calls:np.float
+step:94 - global_seqlen/min:11634 - global_seqlen/max:21223 - global_seqlen/minmax_diff:9589 - global_seqlen/balanced_min:15316 - global_seqlen/balanced_max:15322 - global_seqlen/mean:15319.375 - actor/entropy:0.2614341378211975 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2891600728034973 - training/rollout_probs_diff_mean:0.005587474443018436 - training/rollout_probs_diff_std:0.013904578052461147 - training/rollout_actor_probs_pearson_corr:0.9982290267944336 - rollout_corr/rollout_is_mean:1.0003408193588257 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:2.0101715563214384e-05 - rollout_corr/rollout_is_max:1.7715976238250732 - rollout_corr/rollout_is_min:0.2889171540737152 - rollout_corr/rollout_is_std:0.04165585711598396 - rollout_corr/rollout_is_eff_sample_size:0.9982688644919676 - rollout_corr/rollout_is_seq_mean:1.0004125833511353 - rollout_corr/rollout_is_seq_std:0.0032136309891939163 - rollout_corr/rollout_is_seq_max:1.0084282159805298 - rollout_corr/rollout_is_seq_min:0.9903004765510559 - rollout_corr/rollout_is_seq_max_deviation:0.009699523448944092 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.2973201628774405) - rollout_corr/training_log_ppl:np.float64(0.25612859573448077) - rollout_corr/kl:np.float64(0.0017466846469749164) - rollout_corr/k3_kl:np.float64(0.0015775575718635082) - rollout_corr/rollout_ppl:np.float64(1.2949957121163607) - rollout_corr/rollout_log_ppl:np.float64(0.2543819086276926) - rollout_corr/log_ppl_diff:np.float64(0.0017466871067881584) - rollout_corr/log_ppl_abs_diff:np.float64(0.0035796407610177994) - rollout_corr/log_ppl_diff_max:np.float64(0.01710490882396698) - rollout_corr/log_ppl_diff_min:np.float64(-0.010113626718521118) - rollout_corr/ppl_ratio:np.float64(1.0017579961568117) - rollout_corr/chi2_token:np.float64(0.0027468986809253693) - rollout_corr/chi2_seq:np.float64(0.5400724075734615) - actor/pg_loss:np.float64(-0.0003080737078562379) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0026132785301342665) - actor/ppo_kl:np.float64(0.0012474513576563595) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6366036757826805) - perf/mfu/actor:np.float64(0.044733689603173685) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.00236892700195) - actor/lr:np.float64(1e-06) - training/global_step:94 - training/epoch:6 - critic/score/mean:0.48046875 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.48046875 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.008616097271442413 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.008616097271442413 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:194.32421875 - response_length/max:710.0 - response_length/min:88.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:194.32421875 - response_length_non_aborted/max:710.0 - response_length_non_aborted/min:88.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:284.40625 - prompt_length/max:500.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00011759810149669647 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1009019035845995) - timing_s/agent_loop/generate_sequences/max:np.float64(4.770813453942537) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.1518594626468257) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(4.770813453942537) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - tim
+step:95 - global_seqlen/min:11894 - global_seqlen/max:19955 - global_seqlen/minmax_diff:8061 - global_seqlen/balanced_min:16319 - global_seqlen/balanced_max:16327 - global_seqlen/mean:16323.0 - actor/entropy:0.2705520987510681 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.6408352255821228 - training/rollout_probs_diff_mean:0.005575786344707012 - training/rollout_probs_diff_std:0.013862286694347858 - training/rollout_actor_probs_pearson_corr:0.9983189105987549 - rollout_corr/rollout_is_mean:1.0001085996627808 - rollout_corr/rollout_is_ratio_fraction_high:5.360972136259079e-05 - rollout_corr/rollout_is_ratio_fraction_low:5.360972136259079e-05 - rollout_corr/rollout_is_max:4.217565536499023 - rollout_corr/rollout_is_min:0.15521328151226044 - rollout_corr/rollout_is_std:0.04240885004401207 - rollout_corr/rollout_is_eff_sample_size:0.9982049558450077 - rollout_corr/rollout_is_seq_mean:1.0000663995742798 - rollout_corr/rollout_is_seq_std:0.0035495145712047815 - rollout_corr/rollout_is_seq_max:1.0250022411346436 - rollout_corr/rollout_is_seq_min:0.989392101764679 - rollout_corr/rollout_is_seq_max_deviation:0.025002241134643555 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3268689936958253) - rollout_corr/training_log_ppl:np.float64(0.2789607130398508) - rollout_corr/kl:np.float64(0.001470564366549354) - rollout_corr/k3_kl:np.float64(0.001639172450438764) - rollout_corr/rollout_ppl:np.float64(1.324940457008779) - rollout_corr/rollout_log_ppl:np.float64(0.27749014867004007) - rollout_corr/log_ppl_diff:np.float64(0.0014705643698107451) - rollout_corr/log_ppl_abs_diff:np.float64(0.003353676962433383) - rollout_corr/log_ppl_diff_max:np.float64(0.019685059785842896) - rollout_corr/log_ppl_diff_min:np.float64(-0.014085322618484497) - rollout_corr/ppl_ratio:np.float64(1.0014813262969255) - rollout_corr/chi2_token:np.float64(0.0037556993775069714) - rollout_corr/chi2_seq:np.float64(0.9716611602343619) - actor/pg_loss:np.float64(0.0004548488650470972) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.002927706459104229) - actor/ppo_kl:np.float64(0.0005539002601722132) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.7531055063009262) - perf/mfu/actor:np.float64(0.048065567875728424) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.10460662841797) - actor/lr:np.float64(1e-06) - training/global_step:95 - training/epoch:6 - critic/score/mean:0.53125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.53125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.0015368120511993766 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:0.0015368120511993766 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:218.59375 - response_length/max:779.0 - response_length/min:98.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:218.59375 - response_length_non_aborted/max:779.0 - response_length_non_aborted/min:98.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:291.5 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:3.889948129653931e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0940031111240387) - timing_s/agent_loop/generate_sequences/max:np.float64(5.365459702908993) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.3305309581846814) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(5.365459702908993) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/
+step:96 - global_seqlen/min:11717 - global_seqlen/max:21275 - global_seqlen/minmax_diff:9558 - global_seqlen/balanced_min:16664 - global_seqlen/balanced_max:16674 - global_seqlen/mean:16669.875 - actor/entropy:0.25728729367256165 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2606017589569092 - training/rollout_probs_diff_mean:0.005339381285011768 - training/rollout_probs_diff_std:0.013366999104619026 - training/rollout_actor_probs_pearson_corr:0.9983529448509216 - rollout_corr/rollout_is_mean:0.9999350309371948 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:0.0 - rollout_corr/rollout_is_max:1.6451245546340942 - rollout_corr/rollout_is_min:0.5036647319793701 - rollout_corr/rollout_is_std:0.04145504906773567 - rollout_corr/rollout_is_eff_sample_size:0.9982843083294409 - rollout_corr/rollout_is_seq_mean:0.9999916553497314 - rollout_corr/rollout_is_seq_std:0.0029035070911049843 - rollout_corr/rollout_is_seq_max:1.008498191833496 - rollout_corr/rollout_is_seq_min:0.9916707277297974 - rollout_corr/rollout_is_seq_max_deviation:0.008498191833496094 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3080988936126232) - rollout_corr/training_log_ppl:np.float64(0.2644939655147027) - rollout_corr/kl:np.float64(0.0013578184077687183) - rollout_corr/k3_kl:np.float64(0.001467021100893362) - rollout_corr/rollout_ppl:np.float64(1.3062553596682847) - rollout_corr/rollout_log_ppl:np.float64(0.2631361448438838) - rollout_corr/log_ppl_diff:np.float64(0.00135782067081891) - rollout_corr/log_ppl_abs_diff:np.float64(0.0032226947078015655) - rollout_corr/log_ppl_diff_max:np.float64(0.01374003291130066) - rollout_corr/log_ppl_diff_min:np.float64(-0.008670985698699951) - rollout_corr/ppl_ratio:np.float64(1.0013662767596543) - rollout_corr/chi2_token:np.float64(0.003215978853404522) - rollout_corr/chi2_seq:np.float64(0.44253751053474844) - actor/pg_loss:np.float64(0.00011486734729260206) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0020753747680828383) - actor/ppo_kl:np.float64(0.00043077189779694436) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6561485677957535) - perf/mfu/actor:np.float64(0.04976022063857087) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.02475357055664) - actor/lr:np.float64(1e-06) - training/global_step:96 - training/epoch:6 - critic/score/mean:0.61328125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.61328125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.014713210053741932 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.014713210053741932 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:219.49609375 - response_length/max:856.0 - response_length/min:99.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:219.49609375 - response_length_non_aborted/max:856.0 - response_length_non_aborted/min:99.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:301.4375 - prompt_length/max:499.0 - prompt_length/min:180.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00013059936463832855 - timing_s/agent_loop/generate_sequences/min:np.float64(1.2289475314319134) - timing_s/agent_loop/generate_sequences/max:np.float64(5.740047561004758) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.310526862871484) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(5.740047561004758) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent_loop/slowes
+step:97 - global_seqlen/min:10859 - global_seqlen/max:28193 - global_seqlen/minmax_diff:17334 - global_seqlen/balanced_min:17081 - global_seqlen/balanced_max:23683 - global_seqlen/mean:18237.375 - actor/entropy:0.22356657683849335 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.35336679220199585 - training/rollout_probs_diff_mean:0.004439533222466707 - training/rollout_probs_diff_std:0.01241324283182621 - training/rollout_actor_probs_pearson_corr:0.998500406742096 - rollout_corr/rollout_is_mean:1.0001307725906372 - rollout_corr/rollout_is_ratio_fraction_high:2.887044320232235e-05 - rollout_corr/rollout_is_ratio_fraction_low:1.4435221601161174e-05 - rollout_corr/rollout_is_max:3.0604419708251953 - rollout_corr/rollout_is_min:0.4321064054965973 - rollout_corr/rollout_is_std:0.03731440007686615 - rollout_corr/rollout_is_eff_sample_size:0.9986098092374474 - rollout_corr/rollout_is_seq_mean:1.000051498413086 - rollout_corr/rollout_is_seq_std:0.0030096296686679125 - rollout_corr/rollout_is_seq_max:1.0088967084884644 - rollout_corr/rollout_is_seq_min:0.9923568964004517 - rollout_corr/rollout_is_seq_max_deviation:0.008896708488464355 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.32025858014822) - rollout_corr/training_log_ppl:np.float64(0.2736287588450068) - rollout_corr/kl:np.float64(0.001147672094734986) - rollout_corr/k3_kl:np.float64(0.0012670807614085788) - rollout_corr/rollout_ppl:np.float64(1.318727934267372) - rollout_corr/rollout_log_ppl:np.float64(0.27248108579806285) - rollout_corr/log_ppl_diff:np.float64(0.0011476730469439644) - rollout_corr/log_ppl_abs_diff:np.float64(0.002580658172519179) - rollout_corr/log_ppl_diff_max:np.float64(0.010686814785003662) - rollout_corr/log_ppl_diff_min:np.float64(-0.00808987021446228) - rollout_corr/ppl_ratio:np.float64(1.0011534958612174) - rollout_corr/chi2_token:np.float64(0.002894336823374033) - rollout_corr/chi2_seq:np.float64(0.502729723462835) - actor/pg_loss:np.float64(1.6757636331021786e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0014571902823945493) - actor/ppo_kl:np.float64(0.0002972853384548557) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5494845286011696) - perf/mfu/actor:np.float64(0.050740963915102175) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.98857498168945) - actor/lr:np.float64(1e-06) - training/global_step:97 - training/epoch:6 - critic/score/mean:0.56640625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.56640625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:0.03918260708451271 - critic/advantages/max:0.75 - critic/advantages/min:-0.875 - critic/returns/mean:0.03918260708451271 - critic/returns/max:0.75 - critic/returns/min:-0.875 - response_length/mean:270.60546875 - response_length/max:8192.0 - response_length/min:88.0 - response_length/clip_ratio:0.00390625 - response_length_non_aborted/mean:270.60546875 - response_length_non_aborted/max:8192.0 - response_length_non_aborted/min:88.0 - response_length_non_aborted/clip_ratio:0.00390625 - response/aborted_ratio:0.0 - prompt_length/mean:299.3125 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:2.5369226932525635e-05 - timing_s/agent_loop/generate_sequences/min:np.float64(1.058408448472619) - timing_s/agent_loop/generate_sequences/max:np.float64(49.16501975432038) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.604979175426706) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(49.16501975432038) - timing_s/agent_loop/slowest/tool_calls:n
+step:98 - global_seqlen/min:10881 - global_seqlen/max:21856 - global_seqlen/minmax_diff:10975 - global_seqlen/balanced_min:16595 - global_seqlen/balanced_max:16744 - global_seqlen/mean:16625.0 - actor/entropy:0.27939262986183167 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.2851577401161194 - training/rollout_probs_diff_mean:0.00533874798566103 - training/rollout_probs_diff_std:0.01307785976678133 - training/rollout_actor_probs_pearson_corr:0.9984779357910156 - rollout_corr/rollout_is_mean:0.9997860789299011 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:3.518648736644536e-05 - rollout_corr/rollout_is_max:1.6622998714447021 - rollout_corr/rollout_is_min:0.3280748724937439 - rollout_corr/rollout_is_std:0.040275804698467255 - rollout_corr/rollout_is_eff_sample_size:0.9983798331417372 - rollout_corr/rollout_is_seq_mean:0.999704897403717 - rollout_corr/rollout_is_seq_std:0.0033357173670083284 - rollout_corr/rollout_is_seq_max:1.0101433992385864 - rollout_corr/rollout_is_seq_min:0.9884958267211914 - rollout_corr/rollout_is_seq_max_deviation:0.011504173278808594 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3345653787255287) - rollout_corr/training_log_ppl:np.float64(0.2840118541062111) - rollout_corr/kl:np.float64(0.0016829417983466044) - rollout_corr/k3_kl:np.float64(0.0013737696041289382) - rollout_corr/rollout_ppl:np.float64(1.3323034564964473) - rollout_corr/rollout_log_ppl:np.float64(0.282328910601791) - rollout_corr/log_ppl_diff:np.float64(0.001682943504420109) - rollout_corr/log_ppl_abs_diff:np.float64(0.003258420663769357) - rollout_corr/log_ppl_diff_max:np.float64(0.02178385853767395) - rollout_corr/log_ppl_diff_min:np.float64(-0.009423941373825073) - rollout_corr/ppl_ratio:np.float64(1.0016928501427174) - rollout_corr/chi2_token:np.float64(0.002250493736937642) - rollout_corr/chi2_seq:np.float64(0.3969768069218844) - actor/pg_loss:np.float64(0.00017058837693184614) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0019098859092991916) - actor/ppo_kl:np.float64(0.0005005859824507297) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.6195161938667297) - perf/mfu/actor:np.float64(0.048713654755540986) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(104.02078247070312) - actor/lr:np.float64(1e-06) - training/global_step:98 - training/epoch:6 - critic/score/mean:0.4140625 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.4140625 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.0020672061946243048 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.0020672061946243048 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:222.03125 - response_length/max:1144.0 - response_length/min:99.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:222.03125 - response_length_non_aborted/max:1144.0 - response_length_non_aborted/min:99.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:297.5 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.00013743340969085693 - timing_s/agent_loop/generate_sequences/min:np.float64(1.1269784551113844) - timing_s/agent_loop/generate_sequences/max:np.float64(7.022609854117036) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.2723799158920883) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(7.022609854117036) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/agent_lo
+step:99 - global_seqlen/min:10950 - global_seqlen/max:22756 - global_seqlen/minmax_diff:11806 - global_seqlen/balanced_min:15802 - global_seqlen/balanced_max:15820 - global_seqlen/mean:15813.5 - actor/entropy:0.2819356322288513 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.7740305066108704 - training/rollout_probs_diff_mean:0.005276948679238558 - training/rollout_probs_diff_std:0.013337536714971066 - training/rollout_actor_probs_pearson_corr:0.9984665513038635 - rollout_corr/rollout_is_mean:1.0000571012496948 - rollout_corr/rollout_is_ratio_fraction_high:0.0 - rollout_corr/rollout_is_ratio_fraction_low:5.5403706937795505e-05 - rollout_corr/rollout_is_max:1.6278910636901855 - rollout_corr/rollout_is_min:0.07193134725093842 - rollout_corr/rollout_is_std:0.040468472987413406 - rollout_corr/rollout_is_eff_sample_size:0.9983652180867625 - rollout_corr/rollout_is_seq_mean:1.0001047849655151 - rollout_corr/rollout_is_seq_std:0.003330203937366605 - rollout_corr/rollout_is_seq_max:1.0106587409973145 - rollout_corr/rollout_is_seq_min:0.9881437420845032 - rollout_corr/rollout_is_seq_max_deviation:0.011856257915496826 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3568300306797028) - rollout_corr/training_log_ppl:np.float64(0.3001139621483162) - rollout_corr/kl:np.float64(0.0011941435956686064) - rollout_corr/k3_kl:np.float64(0.0011522117289359812) - rollout_corr/rollout_ppl:np.float64(1.3551504230126739) - rollout_corr/rollout_log_ppl:np.float64(0.29891981728724204) - rollout_corr/log_ppl_diff:np.float64(0.0011941448610741645) - rollout_corr/log_ppl_abs_diff:np.float64(0.0030851933697704226) - rollout_corr/log_ppl_diff_max:np.float64(0.027537941932678223) - rollout_corr/log_ppl_diff_min:np.float64(-0.007955223321914673) - rollout_corr/ppl_ratio:np.float64(1.0012031164951622) - rollout_corr/chi2_token:np.float64(0.0022564970422536135) - rollout_corr/chi2_seq:np.float64(0.5483720011543483) - actor/pg_loss:np.float64(0.0006328735034912825) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0009942931160367152) - actor/ppo_kl:np.float64(0.0003664185717591195) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.5217531993985176) - perf/mfu/actor:np.float64(0.046482210549823294) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.91424560546875) - actor/lr:np.float64(1e-06) - training/global_step:99 - training/epoch:7 - critic/score/mean:0.578125 - critic/score/max:1.0 - critic/score/min:0.0 - critic/rewards/mean:0.578125 - critic/rewards/max:1.0 - critic/rewards/min:0.0 - critic/advantages/mean:-0.011117677204310894 - critic/advantages/max:0.875 - critic/advantages/min:-0.875 - critic/returns/mean:-0.011117677204310894 - critic/returns/max:0.875 - critic/returns/min:-0.875 - response_length/mean:211.515625 - response_length/max:1026.0 - response_length/min:87.0 - response_length/clip_ratio:0.0 - response_length_non_aborted/mean:211.515625 - response_length_non_aborted/max:1026.0 - response_length_non_aborted/min:87.0 - response_length_non_aborted/clip_ratio:0.0 - response/aborted_ratio:0.0 - prompt_length/mean:282.65625 - prompt_length/max:499.0 - prompt_length/min:179.0 - prompt_length/clip_ratio:0.0 - num_turns/min:np.int32(2) - num_turns/max:np.int32(2) - num_turns/mean:np.float64(2.0) - timing_s/start_profile:0.0004795752465724945 - timing_s/agent_loop/generate_sequences/min:np.float64(1.0314996894448996) - timing_s/agent_loop/generate_sequences/max:np.float64(6.317301785573363) - timing_s/agent_loop/generate_sequences/mean:np.float64(2.143986992428836) - timing_s/agent_loop/tool_calls/min:np.float64(0.0) - timing_s/agent_loop/tool_calls/max:np.float64(0.0) - timing_s/agent_loop/tool_calls/mean:np.float64(0.0) - timing_s/agent_loop/slowest/generate_sequences:np.float64(6.317301785573363) - timing_s/agent_loop/slowest/tool_calls:np.float64(0.0) - timing_s/
+test_gen_batch meta info: {'eos_token_id': 151645, 'pad_token_id': 151643, 'recompute_log_prob': False, 'do_sample': True, 'validate': True, 'global_steps': 100}
+validation generation end
+[prompt] system
+
+Given a question and four options, please select the right answer. Respond in the following format:
+<reasoning>
+...
+</reasoning>
+<answer>
+...
+</answer>
+
+For the answer, only output the letter corresponding to the correct option (A, B, C, or D), and nothing else. Do not restate the answer text. For example, if the answer is "A", just output:
+<answer>
+A
+</answer>
+
+user
+What is the folding stability score of the protein sequence 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR'?
+
+A: -0.01
+B: 1.69
+C: 2.49
+D: 0.45
+Please reason step by step.
+assistant
+<think>
+
+</think>
+
+
+[response] <reasoning>
+The folding stability score for a given protein sequence can be determined using computational tools such as I-TAFOR, which predicts the stability of a protein's folded structure based on its sequence. The specific sequence provided, 'GSSTTRYRFLDEEEARRAAKEWARRGYQVHVTQNGTYWEVEVR', would be input into such a tool to calculate the folding stability score. Based on computational predictions, the folding stability score for this sequence is known to be 0.45.
+</reasoning>
+<answer>
+D
+</answer>
+[ground_truth] B
+[score] 0.0
+[acc] 0.0
+[pred] D
+[incorrect_format] 1
+[feedback]
+local_global_step_folder: /mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/biology/qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_100
+step:100 - global_seqlen/min:13194 - global_seqlen/max:24171 - global_seqlen/minmax_diff:10977 - global_seqlen/balanced_min:19146 - global_seqlen/balanced_max:19154 - global_seqlen/mean:19150.75 - actor/entropy:0.2835403382778168 - perf/mfu/actor_infer:0 - training/rollout_probs_diff_valid:1 - training/rollout_probs_diff_max:0.3625403642654419 - training/rollout_probs_diff_mean:0.005042023025453091 - training/rollout_probs_diff_std:0.012832351960241795 - training/rollout_actor_probs_pearson_corr:0.9986383318901062 - rollout_corr/rollout_is_mean:1.000214695930481 - rollout_corr/rollout_is_ratio_fraction_high:3.143171488773078e-05 - rollout_corr/rollout_is_ratio_fraction_low:4.7147572331596166e-05 - rollout_corr/rollout_is_max:4.838179588317871 - rollout_corr/rollout_is_min:0.209492027759552 - rollout_corr/rollout_is_std:0.040638964623212814 - rollout_corr/rollout_is_eff_sample_size:0.9983516728081235 - rollout_corr/rollout_is_seq_mean:1.000152587890625 - rollout_corr/rollout_is_seq_std:0.002947622211650014 - rollout_corr/rollout_is_seq_max:1.0129314661026 - rollout_corr/rollout_is_seq_min:0.9921380877494812 - rollout_corr/rollout_is_seq_max_deviation:0.012931466102600098 - rollout_corr/rollout_is_seq_fraction_high:0.0 - rollout_corr/rollout_is_seq_fraction_low:0.0 - rollout_corr/training_ppl:np.float64(1.3729312745854259) - rollout_corr/training_log_ppl:np.float64(0.3114306655479595) - rollout_corr/kl:np.float64(0.0015098723376403278) - rollout_corr/k3_kl:np.float64(0.001307048618286899) - rollout_corr/rollout_ppl:np.float64(1.3707687840797007) - rollout_corr/rollout_log_ppl:np.float64(0.30992078845156357) - rollout_corr/log_ppl_diff:np.float64(0.0015098770963959396) - rollout_corr/log_ppl_abs_diff:np.float64(0.003156350867357105) - rollout_corr/log_ppl_diff_max:np.float64(0.014475628733634949) - rollout_corr/log_ppl_diff_min:np.float64(-0.009900987148284912) - rollout_corr/ppl_ratio:np.float64(1.0015187887474895) - rollout_corr/chi2_token:np.float64(0.0023139831610023975) - rollout_corr/chi2_seq:np.float64(0.866482462733984) - actor/pg_loss:np.float64(5.068804603070021e-05) - actor/kl_loss:np.float64(0.0) - actor/pg_clipfrac:np.float64(0.0010578534675005358) - actor/ppo_kl:np.float64(0.0007291194974783366) - actor/pg_clipfrac_lower:np.float64(0.0) - actor/grad_norm:np.float64(0.496917188167572) - perf/mfu/actor:np.float64(0.056309282462256305) - perf/max_memory_allocated_gb:np.float64(123.67111110687256) - perf/max_memory_reserved_gb:np.float64(129.48046875) - perf/cpu_memory_used_gb:np.float64(103.95842933654785) - actor/lr:np.float64(1e-06) - val-aux/sciknoweval/reward/mean@16:np.float64(0.42625) - val-aux/sciknoweval/reward/std@16:np.float64(0.20340466932397192) - val-aux/sciknoweval/reward/best@2/mean:np.float64(0.51122) - val-aux/sciknoweval/reward/best@2/std:np.float64(0.1717994688879882) - val-aux/sciknoweval/reward/worst@2/mean:np.float64(0.34018) - val-aux/sciknoweval/reward/worst@2/std:np.float64(0.17528803531142753) - val-aux/sciknoweval/reward/maj@2/mean:np.float64(0.42506) - val-aux/sciknoweval/reward/maj@2/std:np.float64(0.2031877981200387) - val-aux/sciknoweval/reward/best@4/mean:np.float64(0.5764) - val-aux/sciknoweval/reward/best@4/std:np.float64(0.12262009307652777) - val-aux/sciknoweval/reward/worst@4/mean:np.float64(0.26536) - val-aux/sciknoweval/reward/worst@4/std:np.float64(0.12852696069401104) - val-aux/sciknoweval/reward/maj@4/mean:np.float64(0.43562000000000006) - val-aux/sciknoweval/reward/maj@4/std:np.float64(0.16364220520548403) - val-aux/sciknoweval/reward/best@8/mean:np.float64(0.62372) - val-aux/sciknoweval/reward/best@8/std:np.float64(0.08028605088755732) - val-aux/sciknoweval/reward/worst@8/mean:np.float64(0.21248) - val-aux/sciknoweval/reward/worst@8/std:np.float64(0.07648178927728648) - val-aux/sciknoweval/reward/maj@8/mean:np.float64(0.4453000000000001) - val-aux/sciknoweval/reward/maj@8/std:np.float64(0.1155248617088842) - val-aux/sciknoweval/reward/best@16/mean:np.float64(0.6543800000000001) - val-aux/sciknoweval/reward/best@16/std:np.float64(0.0472167585635409) - val-aux/sciknoweval
+("Final validation metrics: {'val-aux/sciknoweval/reward/mean@16': "
+ "np.float64(0.42625), 'val-aux/sciknoweval/reward/std@16': "
+ "np.float64(0.20340466932397192), 'val-aux/sciknoweval/reward/best@2/mean': "
+ "np.float64(0.51122), 'val-aux/sciknoweval/reward/best@2/std': "
+ "np.float64(0.1717994688879882), 'val-aux/sciknoweval/reward/worst@2/mean': "
+ "np.float64(0.34018), 'val-aux/sciknoweval/reward/worst@2/std': "
+ "np.float64(0.17528803531142753), 'val-aux/sciknoweval/reward/maj@2/mean': "
+ "np.float64(0.42506), 'val-aux/sciknoweval/reward/maj@2/std': "
+ "np.float64(0.2031877981200387), 'val-aux/sciknoweval/reward/best@4/mean': "
+ "np.float64(0.5764), 'val-aux/sciknoweval/reward/best@4/std': "
+ "np.float64(0.12262009307652777), 'val-aux/sciknoweval/reward/worst@4/mean': "
+ "np.float64(0.26536), 'val-aux/sciknoweval/reward/worst@4/std': "
+ "np.float64(0.12852696069401104), 'val-aux/sciknoweval/reward/maj@4/mean': "
+ "np.float64(0.43562000000000006), 'val-aux/sciknoweval/reward/maj@4/std': "
+ "np.float64(0.16364220520548403), 'val-aux/sciknoweval/reward/best@8/mean': "
+ "np.float64(0.62372), 'val-aux/sciknoweval/reward/best@8/std': "
+ "np.float64(0.08028605088755732), 'val-aux/sciknoweval/reward/worst@8/mean': "
+ "np.float64(0.21248), 'val-aux/sciknoweval/reward/worst@8/std': "
+ "np.float64(0.07648178927728648), 'val-aux/sciknoweval/reward/maj@8/mean': "
+ "np.float64(0.4453000000000001), 'val-aux/sciknoweval/reward/maj@8/std': "
+ "np.float64(0.1155248617088842), 'val-aux/sciknoweval/reward/best@16/mean': "
+ "np.float64(0.6543800000000001), 'val-aux/sciknoweval/reward/best@16/std': "
+ "np.float64(0.0472167585635409), 'val-aux/sciknoweval/reward/worst@16/mean': "
+ "np.float64(0.18634), 'val-aux/sciknoweval/reward/worst@16/std': "
+ "np.float64(0.032643145410378605), 'val-aux/sciknoweval/reward/maj@16/mean': "
+ "np.float64(0.44676000000000005), 'val-aux/sciknoweval/reward/maj@16/std': "
+ "np.float64(0.07340249583082686), 'val-aux/sciknoweval/score/mean@16': "
+ "np.float64(0.42625), 'val-aux/sciknoweval/score/std@16': "
+ "np.float64(0.20340466932397192), 'val-aux/sciknoweval/score/best@2/mean': "
+ "np.float64(0.51122), 'val-aux/sciknoweval/score/best@2/std': "
+ "np.float64(0.1717994688879882), 'val-aux/sciknoweval/score/worst@2/mean': "
+ "np.float64(0.34018), 'val-aux/sciknoweval/score/worst@2/std': "
+ "np.float64(0.17528803531142753), 'val-aux/sciknoweval/score/maj@2/mean': "
+ "np.float64(0.42506), 'val-aux/sciknoweval/score/maj@2/std': "
+ "np.float64(0.2031877981200387), 'val-aux/sciknoweval/score/best@4/mean': "
+ "np.float64(0.5764), 'val-aux/sciknoweval/score/best@4/std': "
+ "np.float64(0.12262009307652777), 'val-aux/sciknoweval/score/worst@4/mean': "
+ "np.float64(0.26536), 'val-aux/sciknoweval/score/worst@4/std': "
+ "np.float64(0.12852696069401104), 'val-aux/sciknoweval/score/maj@4/mean': "
+ "np.float64(0.43562000000000006), 'val-aux/sciknoweval/score/maj@4/std': "
+ "np.float64(0.16364220520548403), 'val-aux/sciknoweval/score/best@8/mean': "
+ "np.float64(0.62372), 'val-aux/sciknoweval/score/best@8/std': "
+ "np.float64(0.08028605088755732), 'val-aux/sciknoweval/score/worst@8/mean': "
+ "np.float64(0.21248), 'val-aux/sciknoweval/score/worst@8/std': "
+ "np.float64(0.07648178927728648), 'val-aux/sciknoweval/score/maj@8/mean': "
+ "np.float64(0.4453000000000001), 'val-aux/sciknoweval/score/maj@8/std': "
+ "np.float64(0.1155248617088842), 'val-aux/sciknoweval/score/best@16/mean': "
+ "np.float64(0.6543800000000001), 'val-aux/sciknoweval/score/best@16/std': "
+ "np.float64(0.0472167585635409), 'val-aux/sciknoweval/score/worst@16/mean': "
+ "np.float64(0.18634), 'val-aux/sciknoweval/score/worst@16/std': "
+ "np.float64(0.032643145410378605), 'val-aux/sciknoweval/score/maj@16/mean': "
+ "np.float64(0.44676000000000005), 'val-aux/sciknoweval/score/maj@16/std': "
+ "np.float64(0.07340249583082686), 'val-core/sciknoweval/acc/mean@16': "
+ "np.float64(0.42625), 'val-aux/sciknoweval/acc/std@16': "
+ "np.float64(0.20340466932397192), 'val-aux/sciknoweval/acc/best@2/mean': "
+ "np.float64(0.51122), 'val-aux/sciknoweval/acc/best@2/std': "
+ "np.float64(0.1717994688879882), 'val-aux/sciknoweval/acc/worst@2/mean': "
+ "np.float64(0.34018), 'val-aux/sciknoweval/acc/worst@2/std': "
+ "np.float64(0.17528803531142753), 'val-aux/sciknoweval/acc/maj@2/mean': "
+ "np.float64(0.42506), 'val-aux/sciknoweval/acc/maj@2/std': "
+ "np.float64(0.2031877981200387), 'val-aux/sciknoweval/acc/best@4/mean': "
+ "np.float64(0.5764), 'val-aux/sciknoweval/acc/best@4/std': "
+ "np.float64(0.12262009307652777), 'val-aux/sciknoweval/acc/worst@4/mean': "
+ "np.float64(0.26536), 'val-aux/sciknoweval/acc/worst@4/std': "
+ "np.float64(0.12852696069401104), 'val-aux/sciknoweval/acc/maj@4/mean': "
+ "np.float64(0.43562000000000006), 'val-aux/sciknoweval/acc/maj@4/std': "
+ "np.float64(0.16364220520548403), 'val-aux/sciknoweval/acc/best@8/mean': "
+ "np.float64(0.62372), 'val-aux/sciknoweval/acc/best@8/std': "
+ "np.float64(0.08028605088755732), 'val-aux/sciknoweval/acc/worst@8/mean': "
+ "np.float64(0.21248), 'val-aux/sciknoweval/acc/worst@8/std': "
+ "np.float64(0.07648178927728648), 'val-aux/sciknoweval/acc/maj@8/mean': "
+ "np.float64(0.4453000000000001), 'val-aux/sciknoweval/acc/maj@8/std': "
+ "np.float64(0.1155248617088842), 'val-core/sciknoweval/acc/best@16/mean': "
+ "np.float64(0.6543800000000001), 'val-core/sciknoweval/acc/best@16/std': "
+ "np.float64(0.0472167585635409), 'val-aux/sciknoweval/acc/worst@16/mean': "
+ "np.float64(0.18634), 'val-aux/sciknoweval/acc/worst@16/std': "
+ "np.float64(0.032643145410378605), 'val-core/sciknoweval/acc/maj@16/mean': "
+ "np.float64(0.44676000000000005), 'val-core/sciknoweval/acc/maj@16/std': "
+ 'np.float64(0.07340249583082686), '
+ "'val-aux/sciknoweval/incorrect_format/mean@16': np.float64(0.9975), "
+ "'val-aux/sciknoweval/incorrect_format/std@16': "
+ 'np.float64(0.009682458365518542), '
+ "'val-aux/sciknoweval/incorrect_format/best@2/mean': "
+ 'np.float64(0.9997800000000001), '
+ "'val-aux/sciknoweval/incorrect_format/best@2/std': "
+ 'np.float64(0.0029552123640712146), '
+ "'val-aux/sciknoweval/incorrect_format/worst@2/mean': "
+ 'np.float64(0.9955799999999999), '
+ "'val-aux/sciknoweval/incorrect_format/worst@2/std': "
+ 'np.float64(0.012525805206779563), '
+ "'val-aux/sciknoweval/incorrect_format/maj@2/mean': np.float64(0.99798), "
+ "'val-aux/sciknoweval/incorrect_format/maj@2/std': "
+ 'np.float64(0.008749306906094262), '
+ "'val-aux/sciknoweval/incorrect_format/best@4/mean': np.float64(1.0), "
+ "'val-aux/sciknoweval/incorrect_format/best@4/std': np.float64(0.0), "
+ "'val-aux/sciknoweval/incorrect_format/worst@4/mean': np.float64(0.9915), "
+ "'val-aux/sciknoweval/incorrect_format/worst@4/std': "
+ 'np.float64(0.01634315511370097), '
+ "'val-aux/sciknoweval/incorrect_format/maj@4/mean': "
+ 'np.float64(0.9993599999999999), '
+ "'val-aux/sciknoweval/incorrect_format/maj@4/std': "
+ 'np.float64(0.004842664831259046), '
+ "'val-aux/sciknoweval/incorrect_format/best@8/mean': np.float64(1.0), "
+ "'val-aux/sciknoweval/incorrect_format/best@8/std': np.float64(0.0), "
+ "'val-aux/sciknoweval/incorrect_format/worst@8/mean': np.float64(0.98444), "
+ "'val-aux/sciknoweval/incorrect_format/worst@8/std': "
+ 'np.float64(0.019500890160644296), '
+ "'val-aux/sciknoweval/incorrect_format/maj@8/mean': np.float64(0.99984), "
+ "'val-aux/sciknoweval/incorrect_format/maj@8/std': "
+ 'np.float64(0.002504474311076614), '
+ "'val-aux/sciknoweval/incorrect_format/best@16/mean': np.float64(1.0), "
+ "'val-aux/sciknoweval/incorrect_format/best@16/std': np.float64(0.0), "
+ "'val-aux/sciknoweval/incorrect_format/worst@16/mean': "
+ 'np.float64(0.9751599999999999), '
+ "'val-aux/sciknoweval/incorrect_format/worst@16/std': "
+ 'np.float64(0.019403379135729548), '
+ "'val-aux/sciknoweval/incorrect_format/maj@16/mean': np.float64(1.0), "
+ "'val-aux/sciknoweval/incorrect_format/maj@16/std': np.float64(0.0), "
+ "'val-aux/num_turns/min': np.int32(2), 'val-aux/num_turns/max': np.int32(2), "
+ "'val-aux/num_turns/mean': np.float64(2.0)}")
--- a/artifacts/queue.log
+++ b/artifacts/queue.log
--- a/artifacts/wandb-metadata.json
+++ b/artifacts/wandb-metadata.json
@@ -0,0 +1,109 @@
+{
+  "os":  "Linux-6.8.0-71-generic-x86_64-with-glibc2.36",
+  "python":  "CPython 3.12.13",
+  "startedAt":  "2026-07-02T10:24:15.814452Z",
+  "args":  [
+    "--node-ip-address=198.19.44.212",
+    "--node-manager-port=41089",
+    "--object-store-name=/tmp/ray_q3g_biology_grpo_Qwen3_4B/session_2026-07-02_10-22-32_644683_2096785/sockets/plasma_store",
+    "--raylet-name=/tmp/ray_q3g_biology_grpo_Qwen3_4B/session_2026-07-02_10-22-32_644683_2096785/sockets/raylet",
+    "--redis-address=None",
+    "--metrics-agent-port=51010",
+    "--logging-rotate-bytes=536870912",
+    "--logging-rotate-backup-count=5",
+    "--runtime-env-agent-port=64109",
+    "--gcs-address=198.19.44.212:62459",
+    "--session-name=session_2026-07-02_10-22-32_644683_2096785",
+    "--temp-dir=/tmp/ray_q3g_biology_grpo_Qwen3_4B",
+    "--webui=",
+    "--cluster-id=4306b0a7c160c67ff2f68a8cabc9fb42de02add7feb2b9dd3066c9aa",
+    "--startup-token=32",
+    "--worker-launch-time-ms=1782987757164",
+    "--node-id=d852600ad791f29ec1b575f1b787ddb40268e3621fc0726af26802b1",
+    "--runtime-env-hash=1945972661"
+  ],
+  "program":  "/mnt/mole/SDPO/L2T/.venv/lib/python3.12/site-packages/ray/_private/workers/default_worker.py",
+  "codePath":  ".venv/lib/python3.12/site-packages/ray/_private/workers/default_worker.py",
+  "codePathLocal":  ".venv/lib/python3.12/site-packages/ray/_private/workers/default_worker.py",
+  "git":  {
+    "remote":  "https://github.com/jungseongryong/L2T.git",
+    "commit":  "ee6a667700697069b93db636aa937f970c1fb57a"
+  },
+  "email":  "jungsr1116@cau.ac.kr",
+  "root":  "/mnt/mole/SDPO/L2T",
+  "host":  "mole-workspace-rw",
+  "executable":  "/mnt/mole/SDPO/L2T/.venv/bin/python",
+  "cpu_count":  64,
+  "cpu_count_logical":  128,
+  "gpu":  "NVIDIA H200",
+  "gpu_count":  8,
+  "disk":  {
+    "/":  {
+      "total":  "46086056050688",
+      "used":  "21296067473408"
+    }
+  },
+  "memory":  {
+    "total":  "2163980390400"
+  },
+  "gpu_nvidia":  [
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-c0451621-891c-7169-976d-71d81958db34"
+    },
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-e2a29dfa-7dad-6cd7-fb05-79d0060f27f0"
+    },
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-d69bb741-f2a7-a375-7f56-5c0d09d797b2"
+    },
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-1bbbea1e-7541-3126-1bfc-51b43a67577e"
+    },
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-6e0055af-452d-3f46-ce21-0c71e347a608"
+    },
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-317fdb5b-b280-71d8-398e-a843ad897437"
+    },
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-f19d8161-d2dd-49a1-dcb8-5c341b5e4514"
+    },
+    {
+      "name":  "NVIDIA H200",
+      "memoryTotal":  "150754820096",
+      "cudaCores":  16896,
+      "architecture":  "Hopper",
+      "uuid":  "GPU-961f5ebf-4bf8-6855-828b-c065af57b87c"
+    }
+  ],
+  "cudaVersion":  "13.0",
+  "writerId":  "7nuolqqbf1cwwdubm79nn7dndxsbbhe6"
+}
--- a/artifacts/wandb-summary.json
+++ b/artifacts/wandb-summary.json
--- a/chat_template.jinja
+++ b/chat_template.jinja
@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
--- a/checkpoints/last/added_tokens.json
+++ b/checkpoints/last/added_tokens.json
@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}
--- a/checkpoints/last/chat_template.jinja
+++ b/checkpoints/last/chat_template.jinja
@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
--- a/checkpoints/last/config.json
+++ b/checkpoints/last/config.json
@@ -0,0 +1,68 @@
+{
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2560,
+  "initializer_range": 0.02,
+  "intermediate_size": 9728,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 36,
+  "model_type": "qwen3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "4.57.1",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}
--- a/checkpoints/last/generation_config.json
+++ b/checkpoints/last/generation_config.json
@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.57.1"
+}
--- a/checkpoints/last/merges.txt
+++ b/checkpoints/last/merges.txt
--- a/checkpoints/last/model-00001-of-00002.safetensors
+++ b/checkpoints/last/model-00001-of-00002.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3578bb997e78773276aca48ad54e3007bd48807c6e451654ddd1214510e9d3f
+size 4587089976
--- a/checkpoints/last/model-00002-of-00002.safetensors
+++ b/checkpoints/last/model-00002-of-00002.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a721b2dca77e4f0a232d6664f01f7ae605cf899464f61b3d0d360aa111e98b2
+size 4235804536
--- a/checkpoints/last/model.safetensors.index.json
+++ b/checkpoints/last/model.safetensors.index.json
@@ -0,0 +1,407 @@
+{
+  "metadata": {
+    "total_parameters": 4411424256,
+    "total_size": 8822848512
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00001-of-00002.safetensors",
+    "model.embed_tokens.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.norm.weight": "model-00002-of-00002.safetensors"
+  }
+}
--- a/checkpoints/last/special_tokens_map.json
+++ b/checkpoints/last/special_tokens_map.json
@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}
--- a/checkpoints/last/tokenizer.json
+++ b/checkpoints/last/tokenizer.json
--- a/checkpoints/last/tokenizer_config.json
+++ b/checkpoints/last/tokenizer_config.json
@@ -0,0 +1,239 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
--- a/checkpoints/last/vocab.json
+++ b/checkpoints/last/vocab.json
--- a/config.json
+++ b/config.json
@@ -0,0 +1,68 @@
+{
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2560,
+  "initializer_range": 0.02,
+  "intermediate_size": 9728,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 36,
+  "model_type": "qwen3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "4.57.1",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}
--- a/generation_config.json
+++ b/generation_config.json
@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.57.1"
+}
--- a/merges.txt
+++ b/merges.txt
--- a/model-00001-of-00002.safetensors
+++ b/model-00001-of-00002.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3578bb997e78773276aca48ad54e3007bd48807c6e451654ddd1214510e9d3f
+size 4587089976
--- a/model-00002-of-00002.safetensors
+++ b/model-00002-of-00002.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a721b2dca77e4f0a232d6664f01f7ae605cf899464f61b3d0d360aa111e98b2
+size 4235804536
--- a/model.safetensors.index.json
+++ b/model.safetensors.index.json
@@ -0,0 +1,407 @@
+{
+  "metadata": {
+    "total_parameters": 4411424256,
+    "total_size": 8822848512
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00001-of-00002.safetensors",
+    "model.embed_tokens.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.29.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.32.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.33.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.34.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.35.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.k_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.q_norm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.norm.weight": "model-00002-of-00002.safetensors"
+  }
+}
--- a/results/export_note.csv
+++ b/results/export_note.csv
@@ -0,0 +1,7 @@
+key,value
+root_checkpoint_step,100
+root_checkpoint_kind,final
+best_validation_step,90
+best_validation_mean16,0.467500000000
+best_actor_available,false
+reason,global_step_90 has validation metadata only; no actor model shards are present
--- a/results/hyperparameters.csv
+++ b/results/hyperparameters.csv
@@ -0,0 +1,77 @@
+section,parameter,value,source
+Run identity,Base model,Qwen/Qwen3-4B,queue/script override
+Run identity,Dataset,Biology / SciKnowEval biology,run_qwen3_generalization.sh
+Run identity,Method,GRPO,run_qwen3_generalization.sh
+Run identity,Config,baseline_grpo,run_qwen3_generalization.sh
+Run identity,Experiment,qwen3gen-biology-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8,run_qwen3_generalization.sh
+Run identity,W&B run,run-20260702_102415-94h2xs6b,wandb
+Data,Train file,datasets/sciknoweval/biology/train.parquet,script override
+Data,Validation file,datasets/sciknoweval/biology/test.parquet,script override
+Data,Train batch size,32,queue/script override
+Data,Train max samples,3200,queue/script override
+Data,Prompt key,prompt,legacy_data.yaml default
+Data,Reward key,data_source,legacy_data.yaml default
+Data,Shuffle train data,True,user.yaml / legacy_data.yaml
+Data,Validation shuffle,False,legacy_data.yaml default
+Data,Filter overlong prompts,True,user.yaml
+Data,Prompt truncation,error,legacy_data.yaml default
+Data,enable_thinking,false,script override
+Schedule,Total training steps,100,queue/script override
+Schedule,Total epochs,30,ppo_trainer/user.yaml default
+Schedule,Validation before train,False,queue/script override
+Schedule,Save frequency,10,queue/script override
+Schedule,Validation frequency,10,queue/script override
+Sequence,Max prompt length,2048,queue/script override
+Sequence,Max response length,8192,queue/script override
+Sequence,Max model length,10240,queue/script override
+Sequence,Actor max token length per GPU,10240,queue/script override
+Rollout,Rollout engine,vllm,user.yaml
+Rollout,Rollout dtype,bfloat16,rollout.yaml default
+Rollout,Train rollout n,8,queue/script override
+Rollout,Train rollout temperature,1.0,script override
+Rollout,Train rollout top_p,1.0,script override
+Rollout,Train rollout do_sample,True,rollout.yaml default
+Rollout,Calculate rollout log probs,True,baseline_grpo.yaml / script override
+Rollout,Max num batched tokens,10240,queue/script override
+Rollout,vLLM GPU memory utilization,0.8,queue/script override
+Rollout,Tensor model parallel size,2,rollout.yaml default
+Rollout,Free cache engine,True,rollout.yaml default
+Validation,Validation rollout n,16,queue/script override
+Validation,Validation temperature,0.6,queue/script override
+Validation,Validation top_p,0.95,queue/script override
+Validation,Validation do_sample,True,queue/script override
+Optimization,Optimizer,AdamW,fsdp optimizer config
+Optimization,Learning rate,1e-6,GRPO method override
+Optimization,LR scheduler,constant,W&B config
+Optimization,LR warmup steps,10,script override
+Optimization,Weight decay,0.01,script override
+Optimization,Betas,"(0.9, 0.999)",W&B config
+Optimization,Gradient clip,1.0,script override
+PPO/GRPO,Policy loss mode,vanilla,method override
+PPO/GRPO,Advantage estimator,grpo,baseline_grpo.yaml
+PPO/GRPO,Normalize GRPO advantages by std,False,baseline_grpo.yaml / script override
+PPO/GRPO,PPO epochs,1,W&B config
+PPO/GRPO,PPO mini batch size,8,queue/script override
+PPO/GRPO,PPO micro batch size per GPU,1,user.yaml
+PPO/GRPO,Clip ratio low,0.2,script override
+PPO/GRPO,Clip ratio high,0.28,script override
+PPO/GRPO,Gamma,1.0,ppo_trainer.yaml default
+PPO/GRPO,Lambda,1.0,ppo_trainer.yaml default
+PPO/GRPO,Actor KL loss coef,0.0,method override
+PPO/GRPO,Use KL in reward,False,ppo_trainer/user.yaml
+Rollout correction,Importance sampling mode,token,script override
+Rollout correction,IS threshold,2.0,script override
+FSDP/System,Actor strategy,fsdp,dp_actor.yaml
+FSDP/System,FSDP dtype,bfloat16,W&B config
+FSDP/System,FSDP model dtype,fp32,W&B config
+FSDP/System,Use torch compile,True,W&B config
+FSDP/System,GPUs per node,8,queue/script override
+FSDP/System,Nodes,1,user.yaml
+FSDP/System,GPU type,NVIDIA H200,wandb-metadata
+Checkpoint/Logging,Checkpoint root,checkpoints/datasets/sciknoweval/biology,script override
+Checkpoint/Logging,Latest checkpointed iteration,100,latest_checkpointed_iteration.txt
+Checkpoint/Logging,Max actor checkpoints to keep,1,user.yaml
+Checkpoint/Logging,Logger,"console, wandb",ppo_trainer.yaml
+Checkpoint/Logging,W&B entity,seongryongjung-chung-ang-university,environment
+Checkpoint/Logging,W&B project,SDPO-root,user.yaml project_name
+Checkpoint/Logging,W&B group,QWEN3-GRPO-generalization,method override
--- a/results/summary.json
+++ b/results/summary.json
@@ -0,0 +1,27 @@
+{
+  "repo_id": "SeongryongJung/Qwen3-4B-Biology-GRPO-TR",
+  "output_dir": "/mnt/mole/SDPO/L2T/hf_upload_tr/Qwen3-4B-Biology-GRPO-TR",
+  "root_checkpoint_step": 100,
+  "root_checkpoint_kind": "final",
+  "best_step": 90,
+  "best_val_mean16": 0.4675,
+  "best_actor_available": false,
+  "final_step": 100,
+  "final_val_mean16": 0.42625,
+  "train_rows": 90,
+  "val_rows": 10,
+  "hf_model_files": [
+    "added_tokens.json",
+    "chat_template.jinja",
+    "config.json",
+    "generation_config.json",
+    "merges.txt",
+    "model.safetensors.index.json",
+    "special_tokens_map.json",
+    "tokenizer.json",
+    "tokenizer_config.json",
+    "vocab.json",
+    "model-00001-of-00002.safetensors",
+    "model-00002-of-00002.safetensors"
+  ]
+}
--- a/results/training_score.png
+++ b/results/training_score.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2e33136ae641dfe1dee31897c5d1e3f4a15f89df7eb402f053cfd5cda8b5f14
+size 154942
--- a/results/training_score.svg
+++ b/results/training_score.svg
--- a/results/training_scores.csv
+++ b/results/training_scores.csv
@@ -0,0 +1,91 @@
+step,critic_score_mean
+1,0.3515625
+2,0.36328125
+3,0.37890625
+4,0.4140625
+5,0.4609375
+6,0.32421875
+7,0.29296875
+8,0.26953125
+9,0.3984375
+11,0.44140625
+12,0.45703125
+13,0.35546875
+14,0.26953125
+15,0.34765625
+16,0.546875
+17,0.28515625
+18,0.3359375
+19,0.40625
+21,0.40625
+22,0.49609375
+23,0.25390625
+24,0.36328125
+25,0.3828125
+26,0.41015625
+27,0.296875
+28,0.34765625
+29,0.296875
+31,0.375
+32,0.5078125
+33,0.453125
+34,0.3828125
+35,0.35546875
+36,0.34375
+37,0.359375
+38,0.43359375
+39,0.3828125
+41,0.48828125
+42,0.328125
+43,0.44140625
+44,0.4375
+45,0.31640625
+46,0.46484375
+47,0.39453125
+48,0.35546875
+49,0.34765625
+51,0.4140625
+52,0.42578125
+53,0.51953125
+54,0.48046875
+55,0.51171875
+56,0.42578125
+57,0.44140625
+58,0.59375
+59,0.5390625
+61,0.28125
+62,0.46484375
+63,0.39453125
+64,0.4921875
+65,0.4375
+66,0.4296875
+67,0.52734375
+68,0.47265625
+69,0.42578125
+71,0.44921875
+72,0.5
+73,0.66015625
+74,0.42578125
+75,0.3828125
+76,0.43359375
+77,0.6171875
+78,0.43359375
+79,0.41015625
+81,0.5390625
+82,0.59375
+83,0.5234375
+84,0.4296875
+85,0.5
+86,0.63671875
+87,0.5234375
+88,0.375
+89,0.58984375
+91,0.53515625
+92,0.65234375
+93,0.484375
+94,0.48046875
+95,0.53125
+96,0.61328125
+97,0.56640625
+98,0.4140625
+99,0.578125
--- a/results/validation_mean16.csv
+++ b/results/validation_mean16.csv
@@ -0,0 +1,11 @@
+step,val_mean16
+10,0.32625
+20,0.3575
+30,0.3825
+40,0.35375
+50,0.41875
+60,0.3825
+70,0.395
+80,0.41625
+90,0.4675
+100,0.42625
--- a/special_tokens_map.json
+++ b/special_tokens_map.json
@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}
--- a/tokenizer.json
+++ b/tokenizer.json
--- a/tokenizer_config.json
+++ b/tokenizer_config.json
@@ -0,0 +1,239 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
--- a/vocab.json
+++ b/vocab.json