laion/ablation-pymethods2test-seqmean-arm0-30-8B
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity
Files
main
ablation-pymethods2test-seq…/training_logs/ablation-pymethods2test-seqmean-arm0_630125.out

1553 lines
280 KiB
Plaintext
Raw Permalink Normal View History

初始化项目，由ModelHub XC社区提供模型 Model: laion/ablation-pymethods2test-seqmean-arm0-30-8B Source: Original Platform
2026-07-18 18:02:02 +08:00
The following have been reloaded with a version change:
1) GCCcore/.14.3.0 => GCCcore/14.3.0
Lmod is automatically replacing "GCC/14.3.0" with
"nvidia-compilers/25.9-CUDA-13".
Deactivating conda environment: /e/scratch/jureap59/feuer1/miniforge3/envs/otagent
Activating RL environment: /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl
Python executable: /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/bin/python
Python path check: /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/bin/python
[ray] RAY_TMPDIR=/tmp/ray/ray_630125
[triton_cache] Triton cache: /tmp/triton_cache_feuer1_630125
[triton_cache] TorchInductor cache: /tmp/torchinductor_cache_feuer1_630125
[proxy] ✓ Found proxychains binary at /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4
[proxy] Setting up SSH tunnel to jpbl-s01-01
[proxy] SSH key: /e/home/jusers/feuer1/jupiter/.ssh/authorized_keys/id_ed25519_jsc
[proxy] Tunnel port: 7003
[proxy] Node IP: 10.128.32.34 (workers will connect here)
[proxy] ✓ SSH tunnel started successfully
[proxy] ✓ Generated proxychains config at /e/home/jusers/feuer1/jupiter/.proxychains/proxychains_630125.conf
[proxy] - Internal traffic (10.x.x.x, 172.x.x.x, 169.254.x.x) → DIRECT
[proxy] - External traffic (internet) → PROXY via tunnel
[proxy] ✓ Daytona timeout settings configured
[proxy] Testing proxy connectivity...
[proxychains] config file found: /e/home/jusers/feuer1/jupiter/.proxychains/proxychains_630125.conf
[proxychains] preloading /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/lib/libproxychains4.so
[proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e
[proxy] ✓ Proxy connectivity test passed (huggingface.co reachable via wrapped binary)
[proxy] ⚠ Tunnel not accessible at 10.128.32.34:7003 (workers may fail)
[proxy] ✓ Proxy setup complete (using wrapped binary for Ray workers)
[container_runtime] Using cloud backend: daytona (no local container setup)
=== Universal RL Training Runner ===
Config: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/configs/ablation-pymethods2test-seqmean-arm0_rl_config.json
Working directory: /e/scratch/jureap59/feuer1/OpenThoughts-Agent
Python: /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/bin/python
Python version: Python 3.12.12
UV_USE_IO_URING: 0
Proxy: DISABLED (direct internet or not configured)
========================================
=== RLJobRunner: ablation-pymethods2test-seqmean-arm0 ===
[wandb_utils] Fixing permissions on: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/wandb
[wandb_utils] WandB directory ready: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/wandb
HF_TOKEN=****pDbg
HF_HUB_CACHE=/e/data1/datasets/playground/ot-baf/hf_hub
SUPABASE_URL=https://rpzmyuapoqilpghynmza.s... (direct Supabase config)
Environment configured:
TENSOR_PARALLEL_SIZE=1
NUM_INFERENCE_ENGINES=56
POLICY_NUM_NODES=14
WANDB_DIR=/e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/wandb
Starting Ray cluster with 14 nodes, 4 GPUs/node
Cleaning up existing Ray instances...
=== Starting Ray Cluster ===
Nodes: 14
GPUs per node: 4
CPUs per node: 288
Head node: jpbo-041-34 (10.128.32.34)
Ray port: 6379
============================
Starting Ray head on jpbo-041-34 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_head_jpbo-041-34.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.34 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-34 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --head --node-ip-address=10.128.32.34 --port=6379 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray head on jpbo-041-34
Starting Ray worker on jpbo-041-36 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-36.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.36 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-36 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.36 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 1 on jpbo-041-36
Starting Ray worker on jpbo-041-37 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-37.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.37 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-37 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.37 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 2 on jpbo-041-37
Starting Ray worker on jpbo-041-38 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-38.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.38 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-38 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.38 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 3 on jpbo-041-38
Starting Ray worker on jpbo-041-39 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-39.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.39 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-39 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.39 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 4 on jpbo-041-39
Starting Ray worker on jpbo-041-40 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-40.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.40 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-40 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.40 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 5 on jpbo-041-40
Starting Ray worker on jpbo-041-41 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-41.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.41 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-41 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.41 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 6 on jpbo-041-41
Starting Ray worker on jpbo-041-42 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-42.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.42 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-42 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.42 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 7 on jpbo-041-42
Starting Ray worker on jpbo-041-43 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-43.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.43 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-43 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.43 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 8 on jpbo-041-43
Starting Ray worker on jpbo-041-44 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-44.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.44 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-44 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.44 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 9 on jpbo-041-44
Starting Ray worker on jpbo-041-45 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-45.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.45 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-45 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.45 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 10 on jpbo-041-45
Starting Ray worker on jpbo-041-46 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-46.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.46 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-46 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.46 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 11 on jpbo-041-46
Starting Ray worker on jpbo-041-47 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-47.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.47 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-47 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.47 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 12 on jpbo-041-47
Starting Ray worker on jpbo-041-48 (logging to /e/data1/datasets/playground/ot-baf/experiments/_ray_logs/ray_worker_jpbo-041-48.log)...
Command: srun --export=ALL,VLLM_HOST_IP=10.128.32.48 --nodes=1 --ntasks=1 --gres=gpu:4 --gpu-bind=none --overlap --cpu-bind=none -w jpbo-041-48 bash -c /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f "$PROXYCHAINS_CONF_FILE" ray start --address=10.128.32.34:6379 --node-ip-address=10.128.32.48 --num-gpus=4 --num-cpus=288 --block --object-store-memory=42949672960
Started Ray worker 13 on jpbo-041-48
Waiting for cluster (56 GPUs, 14 nodes)...
Connecting to Ray at 10.128.32.34:6379 (expecting 14 nodes, 56.0 GPUs)
Ray connection established, polling for resources...
[Ray wait] nodes=14/14 GPUs=56.0/56.0 resources={'object_store_memory': 601295421440.0, 'memory': 10695797243904.0, 'GPU': 56.0, 'CPU': 4032.0, 'accelerator_type:GH200': 14.0, 'node:10.128.32.43': 1.0, 'node:10.128.32.39': 1.0, 'node:10.128.32.37': 1.0, 'node:10.128.32.46': 1.0, 'node:10.128.32.40': 1.0, 'node:10.128.32.38': 1.0, 'node:10.128.32.45': 1.0, 'node:10.128.32.36': 1.0, 'node:__internal_head__': 1.0, 'node:10.128.32.34': 1.0, 'node:10.128.32.44': 1.0, 'node:10.128.32.42': 1.0, 'node:10.128.32.41': 1.0, 'node:10.128.32.48': 1.0, 'node:10.128.32.47': 1.0}
✓ Ray cluster ready
=== Ray Cluster Ready ===
Address: 10.128.32.34:6379
Total GPUs: 56
=========================
Ray cluster ready at 10.128.32.34:6379
Total GPUs available: 56
[RLJobRunner] Pinggy check: url=False, token=False, needs_tunnel=False (agent=terminus-2, env=daytona)
[RLJobRunner] No Pinggy tunnel needed, using local vLLM
Running SkyRL:
Python: /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/bin/python
Entrypoint: examples.terminal_bench.entrypoints.main_tbench
Args: 120 Hydra arguments
Working dir: /e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train
Using proxychains binary: /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4
Executing command with srun: /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/bin/proxychains4 -f $PROXYCHAINS_CONF_FILE /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/bin/python -m examples.terminal_bench.entrypoints.main_tbench +terminal_bench_config=terminal_bench trainer.strategy=fsdp2 trainer.algorithm.advantage_estimator=rloo_n trainer.algorithm.use_kl_loss=false trainer.algorithm.kl_loss_coef=0.0 trainer.algorithm.eps_clip_low=0.2 trainer.algorithm.eps_clip_high=0.05 trainer.algorithm.loss_reduction=sequence_mean trainer.epochs=2 trainer.max_steps=80 trainer.update_epochs_per_batch=1 trainer.train_batch_size=64 trainer.policy_mini_batch_size=64 trainer.eval_batch_size=64 trainer.micro_forward_batch_size_per_gpu=4 trainer.micro_train_batch_size_per_gpu=1 trainer.max_prompt_length=999999 trainer.eval_interval=999999 trainer.eval_before_train=false trainer.ckpt_interval=2 trainer.resume_mode=latest trainer.hf_save_interval=5 ++trainer.hf_hub_repo_id=laion/ablation-pymethods2test-seqmean-arm0 ++trainer.hf_hub_private=false ++trainer.hf_hub_revision=main ++trainer.enable_db_registration=false trainer.project_name=OpenThoughts-Agent trainer.log_level=INFO trainer.tracker_commit_each_step=true trainer.logger=console trainer.run_name=ablation-pymethods2test-seqmean-arm0 trainer.ckpt_path=/e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints trainer.export_path=/e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/exports trainer.policy.optimizer_config.lr=8e-6 trainer.policy.optimizer_config.weight_decay=0.0 trainer.policy.optimizer_config.adam_betas=[0.9,0.999] trainer.policy.optimizer_config.max_grad_norm=0.9 trainer.policy.fsdp_config.cpu_offload=false trainer.policy.fsdp_config.reshard_after_forward=true trainer.policy.fsdp_config.fsdp_size=4 trainer.policy.model.path=/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6 trainer.ref.fsdp_config.cpu_offload=false trainer.ref.fsdp_config.reshard_after_forward=true trainer.ref.fsdp_config.fsdp_size=4 trainer.placement.colocate_all=false trainer.placement.policy_num_nodes=2 trainer.placement.ref_num_nodes=2 trainer.placement.policy_num_gpus_per_node=4 trainer.placement.ref_num_gpus_per_node=4 trainer.fully_async.max_staleness_steps=16 trainer.fully_async.num_parallel_generation_workers=338 generator.backend=vllm generator.timeout_multiplier=1.0 generator.model_dtype=bfloat16 generator.inference_engine_tensor_parallel_size=1 generator.num_inference_engines=48 generator.n_samples_per_prompt=8 generator.eval_n_samples_per_prompt=8 generator.gpu_memory_utilization=0.75 generator.max_num_seqs=24 generator.max_num_batched_tokens=65536 generator.enable_prefix_caching=true generator.enable_chunked_prefill=true generator.run_engines_locally=true generator.weight_sync_backend=nccl generator.async_engine=true generator.batched=false generator.enable_http_endpoint=true generator.enable_ray_prometheus_stats=false generator.vllm_stats_interval=1 generator.append_eos_token_after_stop_str_in_multi_turn=true generator.max_turns=999999 generator.sampling_params.max_generate_length=4096 generator.sampling_params.temperature=0.7 generator.sampling_params.top_p=0.95 generator.sampling_params.top_k=20 ++generator.engine_init_kwargs.max_model_len=32768 ++generator.engine_init_kwargs.custom_chat_template_chat_completion_path=chat_templates/qwen3_thinking_acc.jinja2 ++generator.engine_init_kwargs.served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6 data.train_data=["/e/scratch/jureap59/feuer1/tasks/exp_rpt_pymethods2test-large"] data.val_data=[] +terminal_bench_config.trials_dir=/e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/trace_jobs +terminal_bench_config.harbor.name=terminus-2 +terminal_bench_config.harbor.max_episodes=999999 +terminal_bench_config.har
[proxychains] config file found: /e/home/jusers/feuer1/jupiter/.proxychains/proxychains_630125.conf
[proxychains] preloading /e/scratch/jureap59/feuer1/proxychains-ng-aarch64/lib/libproxychains4.so
[proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e
2026-06-07 02:11:35.483 | INFO | skyrl_train.utils.utils:prepare_runtime_environment:686 - Exporting wandb api key to ray runtime env
2026-06-07 02:11:35.483 | INFO | skyrl_train.utils.utils:prepare_runtime_environment:705 - Exporting RAY_ADDRESS to ray runtime env
2026-06-07 02:11:35.483 | INFO | skyrl_train.utils.utils:prepare_runtime_environment:730 - Exporting `NCCL_SOCKET_IFNAME` to ray runtime env: ib0
2026-06-07 02:11:35.484 | INFO | skyrl_train.utils.utils:prepare_runtime_environment:730 - Exporting `NCCL_SOCKET_FAMILY` to ray runtime env: AF_INET
2026-06-07 02:11:35.484 | INFO | skyrl_train.utils.utils:prepare_runtime_environment:730 - Exporting `NCCL_DEBUG` to ray runtime env: WARN
2026-06-07 02:11:35,484 INFO worker.py:1680 -- Using address 10.128.32.34:6379 set in the environment variable RAY_ADDRESS
2026-06-07 02:11:35,519 INFO worker.py:1821 -- Connecting to existing Ray cluster at address: 10.128.32.34:6379...
2026-06-07 02:11:35,529 INFO worker.py:2007 -- Connected to Ray cluster.
/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/worker.py:2046: FutureWarning: Tip: In future versions of Ray, Ray will no longer override accelerator visible devices env var if num_gpus=0 or num_gpus=None (default). To enable this behavior and turn off this error message, set RAY_ACCEL_ENV_VAR_OVERRIDE_ON_ZERO=0
warnings.warn(
(raylet, ip=10.128.32.41) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e
2026-06-07 02:11:37.864 | INFO | skyrl_train.utils.ppo_utils:sync_registries:546 - Synced registries to ray actor
(skyrl_entrypoint pid=2701319) 2026-06-07 02:11:44.748 | INFO  | skyrl_train.entrypoints.main_base:_configure_log_level:212 - SkyRL log level set to: INFO
(raylet) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 17x across cluster] (Ray deduplicates logs by default. Set RAY_DEDUP_LOGS=0 to disable log deduplication, or see https://docs.ray.io/en/master/ray-observability/user-guides/configure-logging.html#log-deduplication for more options.)
(skyrl_entrypoint pid=2701319) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
(skyrl_entrypoint pid=2701319) 2026-06-07 02:11:45.133 | INFO  | examples.terminal_bench.dataset:_load_data_files:40 - Loading data from: /e/scratch/jureap59/feuer1/tasks/exp_rpt_pymethods2test-large
(skyrl_entrypoint pid=2701319) 2026-06-07 02:11:48.174 | INFO  | examples.terminal_bench.dataset:_load_data_files:50 - Found 5000 valid task directories out of 5000 total directories
(skyrl_entrypoint pid=2701319) 2026-06-07 02:11:48.174 | INFO  | examples.terminal_bench.dataset:__init__:27 - TerminalBenchTaskDataset initialized with 5000 task paths
(skyrl_entrypoint pid=2701319) 2026-06-07 02:11:48.187 | INFO  | skyrl_train.entrypoints.main_base:_setup_trainer:405 - data:
(skyrl_entrypoint pid=2701319) train_data:
(skyrl_entrypoint pid=2701319) - /e/scratch/jureap59/feuer1/tasks/exp_rpt_pymethods2test-large
(skyrl_entrypoint pid=2701319) val_data: []
(skyrl_entrypoint pid=2701319) trainer:
(skyrl_entrypoint pid=2701319) placement:
(skyrl_entrypoint pid=2701319) colocate_all: false
(skyrl_entrypoint pid=2701319) colocate_policy_ref: true
(skyrl_entrypoint pid=2701319) policy_num_nodes: 2
(skyrl_entrypoint pid=2701319) policy_num_gpus_per_node: 4
(skyrl_entrypoint pid=2701319) critic_num_nodes: 1
(skyrl_entrypoint pid=2701319) critic_num_gpus_per_node: 4
(skyrl_entrypoint pid=2701319) ref_num_nodes: 2
(skyrl_entrypoint pid=2701319) ref_num_gpus_per_node: 4
(skyrl_entrypoint pid=2701319) policy_strict_spread_pg: false
(skyrl_entrypoint pid=2701319) policy_per_gpu_bundles: false
(skyrl_entrypoint pid=2701319) sequence_parallel_backend: ulysses
(skyrl_entrypoint pid=2701319) strategy: fsdp2
(skyrl_entrypoint pid=2701319) policy:
(skyrl_entrypoint pid=2701319) model:
(skyrl_entrypoint pid=2701319) path: /e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6
(skyrl_entrypoint pid=2701319) lora:
(skyrl_entrypoint pid=2701319) rank: 0
(skyrl_entrypoint pid=2701319) alpha: 16
(skyrl_entrypoint pid=2701319) dropout: 0
(skyrl_entrypoint pid=2701319) lora_sync_path: /tmp/skyrl_lora_sync
(skyrl_entrypoint pid=2701319) target_modules: all-linear
(skyrl_entrypoint pid=2701319) exclude_modules: null
(skyrl_entrypoint pid=2701319) deepspeed_config: ${deepspeed_config.train}
(skyrl_entrypoint pid=2701319) optimizer_config:
(skyrl_entrypoint pid=2701319) optimizer: AdamW
(skyrl_entrypoint pid=2701319) lr: 8.0e-06
(skyrl_entrypoint pid=2701319) adam_betas:
(skyrl_entrypoint pid=2701319) - 0.9
(skyrl_entrypoint pid=2701319) - 0.999
(skyrl_entrypoint pid=2701319) weight_decay: 0.0
(skyrl_entrypoint pid=2701319) max_grad_norm: 0.9
(skyrl_entrypoint pid=2701319) offload_after_step: true
(skyrl_entrypoint pid=2701319) num_warmup_steps: 0
(skyrl_entrypoint pid=2701319) scheduler: constant_with_warmup
(skyrl_entrypoint pid=2701319) optimizer_kwargs: {}
(skyrl_entrypoint pid=2701319) fsdp_config:
(skyrl_entrypoint pid=2701319) cpu_offload: false
(skyrl_entrypoint pid=2701319) reshard_after_forward: true
(skyrl_entrypoint pid=2701319) fsdp_size: 4
(skyrl_entrypoint pid=2701319) expert_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) expert_tensor_parallel_size: 1
(skyrl_entrypoint pid=2701319) moe_token_dispatcher_type: alltoall
(skyrl_entrypoint pid=2701319) moe_router_replay: false
(skyrl_entrypoint pid=2701319) moe_grouped_gemm: false
(skyrl_entrypoint pid=2701319) ep_comm_backend: torch
(skyrl_entrypoint pid=2701319) deepep_num_sms: 20
(skyrl_entrypoint pid=2701319) deepep_token_chunk_size: null
(skyrl_entrypoint pid=2701319) sequence_parallel_size: 1
(skyrl_entrypoint pid=2701319) use_torch_compile: false
(skyrl_entrypoint pid=2701319) record_memory: false
(skyrl_entrypoint pid=2701319) megatron_config:
(skyrl_entrypoint pid=2701319) tensor_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) pipeline_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) context_parallel_size: 1
(skyrl_entrypoint pid=2701319) expert_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) expert_tensor_parallel_size: null
(skyrl_entrypoint pid=2701319) ddp_config:
(skyrl_entrypoint pid=2701319) grad_reduce_in_fp32: true
(skyrl_entrypoint pid=2701319) overlap_grad_reduce: false
(skyrl_entrypoint pid=2701319) overlap_param_gather: false
(skyrl_entrypoint pid=2701319) average_in_collective: true
(skyrl_entrypoint pid=2701319) model_config_kwargs: {}
(skyrl_entrypoint pid=2701319) torch_profiler_config:
(skyrl_entrypoint pid=2701319) enable: false
(skyrl_entrypoint pid=2701319) ranks: []
(skyrl_entrypoint pid=2701319) save_path: null
(skyrl_entrypoint pid=2701319) optimizer_config_kwargs:
(skyrl_entrypoint pid=2701319) overlap_cpu_optimizer_d2h_h2d: false
(skyrl_entrypoint pid=2701319) use_precision_aware_optimizer: false
(skyrl_entrypoint pid=2701319) optimizer_cpu_offload: false
(skyrl_entrypoint pid=2701319) optimizer_offload_fraction: 0.0
(skyrl_entrypoint pid=2701319) transformer_config_kwargs:
(skyrl_entrypoint pid=2701319) recompute_granularity: full
(skyrl_entrypoint pid=2701319) recompute_modules:
(skyrl_entrypoint pid=2701319) - core_attn
(skyrl_entrypoint pid=2701319) recompute_method: uniform
(skyrl_entrypoint pid=2701319) recompute_num_layers: 1
(skyrl_entrypoint pid=2701319) empty_cuda_cache: true
(skyrl_entrypoint pid=2701319) ref:
(skyrl_entrypoint pid=2701319) model:
(skyrl_entrypoint pid=2701319) path: ${trainer.policy.model.path}
(skyrl_entrypoint pid=2701319) sequence_parallel_size: 1
(skyrl_entrypoint pid=2701319) deepspeed_config: ${deepspeed_config.eval}
(skyrl_entrypoint pid=2701319) fsdp_config:
(skyrl_entrypoint pid=2701319) cpu_offload: false
(skyrl_entrypoint pid=2701319) reshard_after_forward: true
(skyrl_entrypoint pid=2701319) fsdp_size: 4
(skyrl_entrypoint pid=2701319) expert_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) expert_tensor_parallel_size: 1
(skyrl_entrypoint pid=2701319) moe_token_dispatcher_type: alltoall
(skyrl_entrypoint pid=2701319) moe_router_replay: false
(skyrl_entrypoint pid=2701319) moe_grouped_gemm: false
(skyrl_entrypoint pid=2701319) ep_comm_backend: torch
(skyrl_entrypoint pid=2701319) deepep_num_sms: 20
(skyrl_entrypoint pid=2701319) deepep_token_chunk_size: null
(skyrl_entrypoint pid=2701319) megatron_config:
(skyrl_entrypoint pid=2701319) tensor_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) pipeline_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) context_parallel_size: 1
(skyrl_entrypoint pid=2701319) expert_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) expert_tensor_parallel_size: 1
(skyrl_entrypoint pid=2701319) model_config_kwargs: {}
(skyrl_entrypoint pid=2701319) transformer_config_kwargs: {}
(skyrl_entrypoint pid=2701319) critic:
(skyrl_entrypoint pid=2701319) model:
(skyrl_entrypoint pid=2701319) path: null
(skyrl_entrypoint pid=2701319) lora:
(skyrl_entrypoint pid=2701319) rank: 0
(skyrl_entrypoint pid=2701319) alpha: 16
(skyrl_entrypoint pid=2701319) dropout: 0
(skyrl_entrypoint pid=2701319) target_modules: all-linear
(skyrl_entrypoint pid=2701319) exclude_modules: null
(skyrl_entrypoint pid=2701319) deepspeed_config: ${deepspeed_config.train}
(skyrl_entrypoint pid=2701319) optimizer_config:
(skyrl_entrypoint pid=2701319) optimizer: AdamW
(skyrl_entrypoint pid=2701319) lr: 5.0e-06
(skyrl_entrypoint pid=2701319) adam_betas:
(skyrl_entrypoint pid=2701319) - 0.9
(skyrl_entrypoint pid=2701319) - 0.999
(skyrl_entrypoint pid=2701319) weight_decay: 0.01
(skyrl_entrypoint pid=2701319) max_grad_norm: 1.0
(skyrl_entrypoint pid=2701319) offload_after_step: true
(skyrl_entrypoint pid=2701319) num_warmup_steps: 0
(skyrl_entrypoint pid=2701319) scheduler: constant_with_warmup
(skyrl_entrypoint pid=2701319) optimizer_kwargs: {}
(skyrl_entrypoint pid=2701319) fsdp_config:
(skyrl_entrypoint pid=2701319) cpu_offload: false
(skyrl_entrypoint pid=2701319) reshard_after_forward: true
(skyrl_entrypoint pid=2701319) fsdp_size: -1
(skyrl_entrypoint pid=2701319) expert_model_parallel_size: 1
(skyrl_entrypoint pid=2701319) expert_tensor_parallel_size: 1
(skyrl_entrypoint pid=2701319) moe_token_dispatcher_type: alltoall
(skyrl_entrypoint pid=2701319) moe_router_replay: false
(skyrl_entrypoint pid=2701319) moe_grouped_gemm: false
(skyrl_entrypoint pid=2701319) ep_comm_backend: torch
(skyrl_entrypoint pid=2701319) deepep_num_sms: 20
(skyrl_entrypoint pid=2701319) deepep_token_chunk_size: null
(skyrl_entrypoint pid=2701319) sequence_parallel_size: 1
(skyrl_entrypoint pid=2701319) algorithm:
(skyrl_entrypoint pid=2701319) advantage_estimator: rloo_n
(skyrl_entrypoint pid=2701319) kl_ctrl:
(skyrl_entrypoint pid=2701319) type: fixed
(skyrl_entrypoint pid=2701319) kl_target: 0.1
(skyrl_entrypoint pid=2701319) horizon: 10000
(skyrl_entrypoint pid=2701319) kl_estimator_type: k3
(skyrl_entrypoint pid=2701319) use_kl_estimator_k3: false
(skyrl_entrypoint pid=2701319) use_abs_kl: false
(skyrl_entrypoint pid=2701319) use_kl_in_reward: false
(skyrl_entrypoint pid=2701319) use_kl_loss: false
(skyrl_entrypoint pid=2701319) kl_loss_coef: 0.0
(skyrl_entrypoint pid=2701319) use_entropy_loss: false
(skyrl_entrypoint pid=2701319) entropy_loss_coef: 0.01
(skyrl_entrypoint pid=2701319) advantage_batch_normalize: false
(skyrl_entrypoint pid=2701319) value_head_prefix: value_head
(skyrl_entrypoint pid=2701319) policy_loss_type: regular
(skyrl_entrypoint pid=2701319) loss_reduction: sequence_mean
(skyrl_entrypoint pid=2701319) global_loss_denom: null
(skyrl_entrypoint pid=2701319) grpo_norm_by_std: true
(skyrl_entrypoint pid=2701319) rloo_n_min_group_size: 4
(skyrl_entrypoint pid=2701319) rloo_n_filter_zero_reward_groups: true
(skyrl_entrypoint pid=2701319) lambd: 1.0
(skyrl_entrypoint pid=2701319) gamma: 1.0
(skyrl_entrypoint pid=2701319) eps_clip_low: 0.2
(skyrl_entrypoint pid=2701319) eps_clip_high: 0.05
(skyrl_entrypoint pid=2701319) clip_ratio_c: 3.0
(skyrl_entrypoint pid=2701319) tis_imp_ratio_cap: -1.0
(skyrl_entrypoint pid=2701319) use_tis: false
(skyrl_entrypoint pid=2701319) sapo:
(skyrl_entrypoint pid=2701319) tau_pos: 1.0
(skyrl_entrypoint pid=2701319) tau_neg: 1.05
(skyrl_entrypoint pid=2701319) value_clip: 0.2
(skyrl_entrypoint pid=2701319) dynamic_sampling:
(skyrl_entrypoint pid=2701319) type: null
(skyrl_entrypoint pid=2701319) max_sample_batches: 30
(skyrl_entrypoint pid=2701319) min_replace_ratio: 0.3
(skyrl_entrypoint pid=2701319) clip_cov:
(skyrl_entrypoint pid=2701319) clip_ratio: 0.0002
(skyrl_entrypoint pid=2701319) clip_cov_lb: 1.0
(skyrl_entrypoint pid=2701319) clip_cov_ub: 5.0
(skyrl_entrypoint pid=2701319) kl_cov:
(skyrl_entrypoint pid=2701319) kl_cov_frac: 0.2
(skyrl_entrypoint pid=2701319) ppo_kl_coef: 1.0
(skyrl_entrypoint pid=2701319) cispo:
(skyrl_entrypoint pid=2701319) cispo_eps_clip_low: 0
(skyrl_entrypoint pid=2701319) cispo_eps_clip_high: 5
(skyrl_entrypoint pid=2701319) z_clip:
(skyrl_entrypoint pid=2701319) enabled: false
(skyrl_entrypoint pid=2701319) alpha: 0.97
(skyrl_entrypoint pid=2701319) z_thresh: 2.5
(skyrl_entrypoint pid=2701319) warmup_steps: 25
(skyrl_entrypoint pid=2701319) mode: zscore
(skyrl_entrypoint pid=2701319) clip_option: adaptive_scaling
(skyrl_entrypoint pid=2701319) clip_factor: 1.0
(skyrl_entrypoint pid=2701319) skip_update_on_spike: false
(skyrl_entrypoint pid=2701319) stale_clip:
(skyrl_entrypoint pid=2701319) enabled: false
(skyrl_entrypoint pid=2701319) alpha: 0.3
(skyrl_entrypoint pid=2701319) entropy_threshold: 0.15
(skyrl_entrypoint pid=2701319) entropy_window: 10
(skyrl_entrypoint pid=2701319) min_lr_scale: 0.1
(skyrl_entrypoint pid=2701319) max_seq_len: 1004095
(skyrl_entrypoint pid=2701319) fully_async:
(skyrl_entrypoint pid=2701319) max_staleness_steps: 16
(skyrl_entrypoint pid=2701319) num_parallel_generation_workers: 338
(skyrl_entrypoint pid=2701319) gradient_checkpointing: true
(skyrl_entrypoint pid=2701319) gradient_checkpointing_use_reentrant: false
(skyrl_entrypoint pid=2701319) seed: 42
(skyrl_entrypoint pid=2701319) resume_mode: latest
(skyrl_entrypoint pid=2701319) resume_path: null
(skyrl_entrypoint pid=2701319) ckpt_path: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints
(skyrl_entrypoint pid=2701319) max_ckpts_to_keep: -1
(skyrl_entrypoint pid=2701319) ckpt_interval: 2
(skyrl_entrypoint pid=2701319) hf_save_interval: 5
(skyrl_entrypoint pid=2701319) hf_upload_mode: latest
(skyrl_entrypoint pid=2701319) export_path: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/exports
(skyrl_entrypoint pid=2701319) bf16: true
(skyrl_entrypoint pid=2701319) epochs: 2
(skyrl_entrypoint pid=2701319) max_steps: 80
(skyrl_entrypoint pid=2701319) update_epochs_per_batch: 1
(skyrl_entrypoint pid=2701319) train_batch_size: 64
(skyrl_entrypoint pid=2701319) policy_mini_batch_size: 64
(skyrl_entrypoint pid=2701319) critic_mini_batch_size: 256
(skyrl_entrypoint pid=2701319) micro_train_batch_size_per_gpu: 1
(skyrl_entrypoint pid=2701319) micro_forward_batch_size_per_gpu: 4
(skyrl_entrypoint pid=2701319) update_ref_every_epoch: false
(skyrl_entrypoint pid=2701319) use_sample_packing: true
(skyrl_entrypoint pid=2701319) eval_batch_size: 64
(skyrl_entrypoint pid=2701319) eval_before_train: false
(skyrl_entrypoint pid=2701319) eval_interval: 999999
(skyrl_entrypoint pid=2701319) max_prompt_length: 999999
(skyrl_entrypoint pid=2701319) flash_attn: true
(skyrl_entrypoint pid=2701319) disable_fast_tokenizer: false
(skyrl_entrypoint pid=2701319) target_modules: null
(skyrl_entrypoint pid=2701319) exclude_modules: null
(skyrl_entrypoint pid=2701319) project_name: OpenThoughts-Agent
(skyrl_entrypoint pid=2701319) run_name: ablation-pymethods2test-seqmean-arm0
(skyrl_entrypoint pid=2701319) logger: console
(skyrl_entrypoint pid=2701319) tracker_commit_each_step: true
(skyrl_entrypoint pid=2701319) dump_data_batch: false
(skyrl_entrypoint pid=2701319) dump_eval_results: true
(skyrl_entrypoint pid=2701319) log_level: INFO
(skyrl_entrypoint pid=2701319) rope_scaling: null
(skyrl_entrypoint pid=2701319) rope_theta: null
(skyrl_entrypoint pid=2701319) step_wise_training: false
(skyrl_entrypoint pid=2701319) hf_hub_repo_id: laion/ablation-pymethods2test-seqmean-arm0
(skyrl_entrypoint pid=2701319) hf_hub_private: false
(skyrl_entrypoint pid=2701319) hf_hub_revision: main
(skyrl_entrypoint pid=2701319) enable_db_registration: false
(skyrl_entrypoint pid=2701319) generator:
(skyrl_entrypoint pid=2701319) model_name: ${trainer.policy.model.path}
(skyrl_entrypoint pid=2701319) model_dtype: bfloat16
(skyrl_entrypoint pid=2701319) timeout_multiplier: 1.0
(skyrl_entrypoint pid=2701319) run_engines_locally: true
(skyrl_entrypoint pid=2701319) num_inference_engines: 48
(skyrl_entrypoint pid=2701319) backend: vllm
(skyrl_entrypoint pid=2701319) weight_sync_backend: nccl
(skyrl_entrypoint pid=2701319) fuse_weights: false
(skyrl_entrypoint pid=2701319) weight_transfer_threshold_cuda_ipc_GB: 1.0
(skyrl_entrypoint pid=2701319) inference_engine_tensor_parallel_size: 1
(skyrl_entrypoint pid=2701319) inference_engine_pipeline_parallel_size: 1
(skyrl_entrypoint pid=2701319) inference_engine_expert_parallel_size: 1
(skyrl_entrypoint pid=2701319) inference_engine_data_parallel_size: 1
(skyrl_entrypoint pid=2701319) n_samples_per_prompt: 8
(skyrl_entrypoint pid=2701319) async_engine: true
(skyrl_entrypoint pid=2701319) batched: false
(skyrl_entrypoint pid=2701319) max_input_length: ${trainer.max_prompt_length}
(skyrl_entrypoint pid=2701319) vllm_v1_disable_multiproc: true
(skyrl_entrypoint pid=2701319) enable_prefix_caching: true
(skyrl_entrypoint pid=2701319) enable_chunked_prefill: true
(skyrl_entrypoint pid=2701319) max_num_batched_tokens: 65536
(skyrl_entrypoint pid=2701319) enforce_eager: true
(skyrl_entrypoint pid=2701319) fully_sharded_loras: false
(skyrl_entrypoint pid=2701319) enable_ray_prometheus_stats: false
(skyrl_entrypoint pid=2701319) vllm_stats_interval: 1
(skyrl_entrypoint pid=2701319) gpu_memory_utilization: 0.75
(skyrl_entrypoint pid=2701319) max_num_seqs: 24
(skyrl_entrypoint pid=2701319) remote_inference_engine_urls:
(skyrl_entrypoint pid=2701319) - 127.0.0.1:8001
(skyrl_entrypoint pid=2701319) enable_http_endpoint: true
(skyrl_entrypoint pid=2701319) http_endpoint_host: 127.0.0.1
(skyrl_entrypoint pid=2701319) http_endpoint_port: 8000
(skyrl_entrypoint pid=2701319) max_turns: 999999
(skyrl_entrypoint pid=2701319) chat_template:
(skyrl_entrypoint pid=2701319) source: name
(skyrl_entrypoint pid=2701319) name_or_path: null
(skyrl_entrypoint pid=2701319) chat_template_kwargs: {}
(skyrl_entrypoint pid=2701319) engine_init_kwargs:
(skyrl_entrypoint pid=2701319) max_model_len: 32768
(skyrl_entrypoint pid=2701319) custom_chat_template_chat_completion_path: chat_templates/qwen3_thinking_acc.jinja2
(skyrl_entrypoint pid=2701319) served_model_name: 0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6
(skyrl_entrypoint pid=2701319) override_existing_update_group: disable
(skyrl_entrypoint pid=2701319) sampling_params:
(skyrl_entrypoint pid=2701319) max_generate_length: 4096
(skyrl_entrypoint pid=2701319) repetition_penalty: 1.0
(skyrl_entrypoint pid=2701319) temperature: 0.7
(skyrl_entrypoint pid=2701319) top_p: 0.95
(skyrl_entrypoint pid=2701319) min_p: 0.0
(skyrl_entrypoint pid=2701319) top_k: 20
(skyrl_entrypoint pid=2701319) logprobs: null
(skyrl_entrypoint pid=2701319) stop: null
(skyrl_entrypoint pid=2701319) use_conversation_multi_turn: true
(skyrl_entrypoint pid=2701319) append_eos_token_after_stop_str_in_multi_turn: true
(skyrl_entrypoint pid=2701319) eval_sampling_params:
(skyrl_entrypoint pid=2701319) max_generate_length: ${generator.sampling_params.max_generate_length}
(skyrl_entrypoint pid=2701319) repetition_penalty: 1.0
(skyrl_entrypoint pid=2701319) temperature: 0.0
(skyrl_entrypoint pid=2701319) top_p: 1.0
(skyrl_entrypoint pid=2701319) min_p: 0.0
(skyrl_entrypoint pid=2701319) top_k: -1
(skyrl_entrypoint pid=2701319) logprobs: null
(skyrl_entrypoint pid=2701319) stop: null
(skyrl_entrypoint pid=2701319) eval_n_samples_per_prompt: 8
(skyrl_entrypoint pid=2701319) zero_reward_on_non_stop: false
(skyrl_entrypoint pid=2701319) apply_overlong_filtering: false
(skyrl_entrypoint pid=2701319) rope_scaling: ${trainer.rope_scaling}
(skyrl_entrypoint pid=2701319) rope_theta: ${trainer.rope_theta}
(skyrl_entrypoint pid=2701319) teacher:
(skyrl_entrypoint pid=2701319) model_path: null
(skyrl_entrypoint pid=2701319) top_k_logprobs: 256
(skyrl_entrypoint pid=2701319) num_inference_engines: 1
(skyrl_entrypoint pid=2701319) inference_engine_tensor_parallel_size: 1
(skyrl_entrypoint pid=2701319) inference_engine_pipeline_parallel_size: 1
(skyrl_entrypoint pid=2701319) gpu_memory_utilization: 0.9
(skyrl_entrypoint pid=2701319) enforce_eager: false
(skyrl_entrypoint pid=2701319) backend: vllm
(skyrl_entrypoint pid=2701319) engine_init_kwargs: {}
(skyrl_entrypoint pid=2701319) environment:
(skyrl_entrypoint pid=2701319) env_class: gsm8k
(skyrl_entrypoint pid=2701319) skyrl_gym:
(skyrl_entrypoint pid=2701319) max_env_workers: 32
(skyrl_entrypoint pid=2701319) text2sql:
(skyrl_entrypoint pid=2701319) db_path: /home/ray/default/sql_data
(skyrl_entrypoint pid=2701319) llm_as_a_judge:
(skyrl_entrypoint pid=2701319) model: gpt-4o-mini
(skyrl_entrypoint pid=2701319) base_url: null
(skyrl_entrypoint pid=2701319) search:
(skyrl_entrypoint pid=2701319) log_requests: false
(skyrl_entrypoint pid=2701319) search_url: http://127.0.0.1:8000/retrieve
(skyrl_entrypoint pid=2701319) topk: 3
(skyrl_entrypoint pid=2701319) timeout: 30
(skyrl_entrypoint pid=2701319) rollout:
(skyrl_entrypoint pid=2701319) fanout:
(skyrl_entrypoint pid=2701319) enabled: true
(skyrl_entrypoint pid=2701319) num_coordinators: 4
(skyrl_entrypoint pid=2701319) cpus_per_coordinator: 8
(skyrl_entrypoint pid=2701319) deepspeed_config:
(skyrl_entrypoint pid=2701319) train:
(skyrl_entrypoint pid=2701319) zero_optimization:
(skyrl_entrypoint pid=2701319) stage: 3
(skyrl_entrypoint pid=2701319) offload_param:
(skyrl_entrypoint pid=2701319) device: none
(skyrl_entrypoint pid=2701319) offload_optimizer:
(skyrl_entrypoint pid=2701319) device: none
(skyrl_entrypoint pid=2701319) pin_memory: true
(skyrl_entrypoint pid=2701319) sub_group_size: auto
(skyrl_entrypoint pid=2701319) reduce_bucket_size: auto
(skyrl_entrypoint pid=2701319) stage3_param_persistence_threshold: auto
(skyrl_entrypoint pid=2701319) stage3_prefetch_bucket_size: auto
(skyrl_entrypoint pid=2701319) stage3_max_live_parameters: auto
(skyrl_entrypoint pid=2701319) stage3_max_reuse_distance: auto
(skyrl_entrypoint pid=2701319) round_robin_gradients: true
(skyrl_entrypoint pid=2701319) zero_hpz_partition_size: 1
(skyrl_entrypoint pid=2701319) zero_quantized_weights: false
(skyrl_entrypoint pid=2701319) zero_quantized_gradients: false
(skyrl_entrypoint pid=2701319) torch_autocast:
(skyrl_entrypoint pid=2701319) enabled: true
(skyrl_entrypoint pid=2701319) dtype: bfloat16
(skyrl_entrypoint pid=2701319) disable_trace_cache: false
(skyrl_entrypoint pid=2701319) data_types:
(skyrl_entrypoint pid=2701319) grad_accum_dtype: fp32
(skyrl_entrypoint pid=2701319) gradient_clipping: 1.0
(skyrl_entrypoint pid=2701319) wall_clock_breakdown: false
(skyrl_entrypoint pid=2701319) prescale_gradient: false
(skyrl_entrypoint pid=2701319) eval:
(skyrl_entrypoint pid=2701319) zero_optimization:
(skyrl_entrypoint pid=2701319) stage: 3
(skyrl_entrypoint pid=2701319) stage3_param_persistence_threshold: auto
(skyrl_entrypoint pid=2701319) offload_param:
(skyrl_entrypoint pid=2701319) device: cpu
(skyrl_entrypoint pid=2701319) pin_memory: true
(skyrl_entrypoint pid=2701319) torch_autocast:
(skyrl_entrypoint pid=2701319) enabled: true
(skyrl_entrypoint pid=2701319) dtype: bfloat16
(skyrl_entrypoint pid=2701319) gradient_clipping: 1.0
(skyrl_entrypoint pid=2701319) prescale_gradient: false
(skyrl_entrypoint pid=2701319) wall_clock_breakdown: false
(skyrl_entrypoint pid=2701319) terminal_bench_config:
(skyrl_entrypoint pid=2701319) trials_dir: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/trace_jobs
(skyrl_entrypoint pid=2701319) harbor:
(skyrl_entrypoint pid=2701319) name: terminus-2
(skyrl_entrypoint pid=2701319) max_episodes: 999999
(skyrl_entrypoint pid=2701319) enable_summarize: false
(skyrl_entrypoint pid=2701319) store_all_messages: true
(skyrl_entrypoint pid=2701319) trajectory_config:
(skyrl_entrypoint pid=2701319) raw_content: true
(skyrl_entrypoint pid=2701319) enable_episode_logging: false
(skyrl_entrypoint pid=2701319) record_terminal_session: false
(skyrl_entrypoint pid=2701319) enable_pane_logging: false
(skyrl_entrypoint pid=2701319) strict_json_parser: true
(skyrl_entrypoint pid=2701319) interleaved_thinking: true
(skyrl_entrypoint pid=2701319) extra_body:
(skyrl_entrypoint pid=2701319) chat_template_kwargs:
(skyrl_entrypoint pid=2701319) enable_thinking: true
(skyrl_entrypoint pid=2701319) override_timeout_sec: 900
(skyrl_entrypoint pid=2701319) override_cpus: 1
(skyrl_entrypoint pid=2701319) override_memory_mb: 2048
(skyrl_entrypoint pid=2701319) override_storage_mb: 2048
(skyrl_entrypoint pid=2701319) auto_snapshot: true
(skyrl_entrypoint pid=2701319) verifier_override_timeout_sec: 120
(skyrl_entrypoint pid=2701319) max_retries: 3
(skyrl_entrypoint pid=2701319) min_wait_sec: 60.0
(skyrl_entrypoint pid=2701319) max_wait_sec: 600.0
(skyrl_entrypoint pid=2701319) wait_multiplier: 2.0
(skyrl_entrypoint pid=2701319) exclude_exceptions:
(skyrl_entrypoint pid=2701319) - VerifierTimeoutError
(skyrl_entrypoint pid=2701319) - VerifierRuntimeError
(skyrl_entrypoint pid=2701319) - RewardFileNotFoundError
(skyrl_entrypoint pid=2701319) - RewardFileEmptyError
(skyrl_entrypoint pid=2701319) - VerifierOutputParseError
(skyrl_entrypoint pid=2701319) n_concurrent_trials: 675
(skyrl_entrypoint pid=2701319) log_level: INFO
(skyrl_entrypoint pid=2701319) enable_reward_shaping: false
(skyrl_entrypoint pid=2701319) enable_error_classification: true
(skyrl_entrypoint pid=2701319) mask_exceptions:
(skyrl_entrypoint pid=2701319) - DaytonaError
(skyrl_entrypoint pid=2701319) - EnvironmentStartTimeoutError
(skyrl_entrypoint pid=2701319) - NetworkError
(skyrl_entrypoint pid=2701319) - ConnectionError
(skyrl_entrypoint pid=2701319) - RewardFileNotFoundError
(skyrl_entrypoint pid=2701319) - RewardFileEmptyError
(skyrl_entrypoint pid=2701319) - AgentEnvironmentTimeoutError
(skyrl_entrypoint pid=2701319) - ContextLengthExceededError
(skyrl_entrypoint pid=2701319) default_error_treatment: zero
(skyrl_entrypoint pid=2701319) passthrough_exceptions:
(skyrl_entrypoint pid=2701319) - AgentTimeoutError
(skyrl_entrypoint pid=2701319) zero_exceptions: []
(skyrl_entrypoint pid=2701319) model_info:
(skyrl_entrypoint pid=2701319) max_input_tokens: 32000
(skyrl_entrypoint pid=2701319) max_output_tokens: 4096
(skyrl_entrypoint pid=2701319) archiving:
(skyrl_entrypoint pid=2701319) enabled: false
(skyrl_entrypoint pid=2701319) trace_upload:
(skyrl_entrypoint pid=2701319) enabled: true
(skyrl_entrypoint pid=2701319) repo_org: DCAgent
(skyrl_entrypoint pid=2701319) episodes: last
(skyrl_entrypoint pid=2701319) dataset_type: SFT
(skyrl_entrypoint pid=2701319) cleanup: true
(skyrl_entrypoint pid=2701319) 
(skyrl_entrypoint pid=2701319) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash:
(skyrl_entrypoint pid=2701319) No module named 'vllm._version'
(skyrl_entrypoint pid=2701319) from .version import __version__, __version_tuple__ # isort:skip
(skyrl_entrypoint pid=2701319) W0607 02:12:04.505000 2701319 envs/rl/lib/python3.12/site-packages/torch/utils/cpp_extension.py:117] No CUDA runtime is found, using CUDA_HOME='/e/software/default/stages/2026/software/CUDA/13'
[2026-06-07 02:12:05,742 E 2700889 2701293] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(RegistryActor pid=2575022, ip=10.128.32.41) [2026-06-07 02:12:06,659 E 2575022 2575062] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(raylet) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e
(skyrl_entrypoint pid=2701319) [2026-06-07 02:12:08,855 E 2701319 2701362] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14 [repeated 2x across cluster]
(raylet, ip=10.128.32.37) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 31x across cluster]
(pid=2666524, ip=10.128.32.38) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash:
(pid=2666524, ip=10.128.32.38) No module named 'vllm._version'
(pid=2666524, ip=10.128.32.38) from .version import __version__, __version_tuple__ # isort:skip
(raylet, ip=10.128.32.48) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 64x across cluster]
(pid=2539107, ip=10.128.32.40) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 7x across cluster]
(pid=2539107, ip=10.128.32.40) No module named 'vllm._version' [repeated 7x across cluster]
(pid=2539107, ip=10.128.32.40) from .version import __version__, __version_tuple__ # isort:skip [repeated 7x across cluster]
(raylet, ip=10.128.32.47) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 116x across cluster]
(pid=2593412, ip=10.128.32.36) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 16x across cluster]
(pid=2593412, ip=10.128.32.36) No module named 'vllm._version' [repeated 16x across cluster]
(pid=2593412, ip=10.128.32.36) from .version import __version__, __version_tuple__ # isort:skip [repeated 16x across cluster]
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:28.309 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:123 - setup_envvars_for_vllm: distributed_executor_backend=uni, SKYRL_ENABLE_NUMA_AFFINITY=1, CUDA_VISIBLE_DEVICES=0, VLLM_ENABLE_V1_MULTIPROCESSING=<unset>
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:28.334 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:133 - setup_envvars_for_vllm: numa_enabled=True
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:28.335 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:136 - setup_envvars_for_vllm: set VLLM_ENABLE_V1_MULTIPROCESSING=0 for NUMA affinity
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:28.982 | INFO | skyrl_train.utils.numa:set_numa_affinity_for_gpu:340 - NUMA affinity: bound process to CPUs 0-71 (NUMA node 0) for GPU 0
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:29.008 | DEBUG | skyrl_train.utils.numa:_set_membind_via_libnuma:375 - NUMA affinity: set memory preferred to NUMA node 0
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:29.008 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:584 - BaseVLLMInferenceEngine: vllm_v1_disable_multiproc=True, vllm.__version__=dev, VLLM_ENABLE_V1_MULTIPROCESSING=0
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:29.008 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:592 - BaseVLLMInferenceEngine: set VLLM_ENABLE_V1_MULTIPROCESSING=0
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:29.008 | WARNING | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1158 - OpenAI API sampling params overridden: temperature=0.7, top_p=0.95, top_k=20
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) 2026-06-07 02:12:29.059 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1210 - Engine startup stagger: sleeping 1.94s (attempt 1/5) to avoid port collisions
(raylet, ip=10.128.32.44) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 106x across cluster]
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
(pid=2659898, ip=10.128.32.43) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 13x across cluster]
(pid=2659898, ip=10.128.32.43) No module named 'vllm._version' [repeated 13x across cluster]
(pid=2659898, ip=10.128.32.43) from .version import __version__, __version_tuple__ # isort:skip [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:31.812 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:123 - setup_envvars_for_vllm: distributed_executor_backend=uni, SKYRL_ENABLE_NUMA_AFFINITY=1, CUDA_VISIBLE_DEVICES=3, VLLM_ENABLE_V1_MULTIPROCESSING=<unset> [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:31.812 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:133 - setup_envvars_for_vllm: numa_enabled=True [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:31.812 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:136 - setup_envvars_for_vllm: set VLLM_ENABLE_V1_MULTIPROCESSING=0 for NUMA affinity [repeated 3x across cluster]
(skyrl_entrypoint pid=2701319) [2026-06-07 02:12:33] INFO inference_engine_client_http_endpoint.py:350: Starting server on 0.0.0.0:8000
(skyrl_entrypoint pid=2701319) [2026-06-07 02:12:34] INFO inference_engine_client_http_endpoint.py:242: Starting inference HTTP endpoint...
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:32.760 | INFO | skyrl_train.utils.numa:set_numa_affinity_for_gpu:340 - NUMA affinity: bound process to CPUs 216-287 (NUMA node 3) for GPU 3 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:32.763 | DEBUG | skyrl_train.utils.numa:_set_membind_via_libnuma:375 - NUMA affinity: set memory preferred to NUMA node 3 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:32.763 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:584 - BaseVLLMInferenceEngine: vllm_v1_disable_multiproc=True, vllm.__version__=dev, VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:32.763 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:592 - BaseVLLMInferenceEngine: set VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:32.763 | WARNING | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1158 - OpenAI API sampling params overridden: temperature=0.7, top_p=0.95, top_k=20 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) 2026-06-07 02:12:32.784 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1210 - Engine startup stagger: sleeping 1.51s (attempt 1/5) to avoid port collisions [repeated 3x across cluster]
(skyrl_entrypoint pid=2701319) [2026-06-07 02:12:34] INFO inference_engine_client_http_endpoint.py:229: Server ready after 2 attempts (2 seconds)
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.803 | INFO  | skyrl_train.inference_engines.inference_engine_client:_spin_up_http_endpoint:969 - InferenceEngineClient HTTP endpoint started on 127.0.0.1:8000
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.803 | INFO  | skyrl_train.inference_engines.inference_engine_client:__init__:61 - InferenceEngineClient initialized with 48 engines.
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.805 | INFO  | examples.terminal_bench.terminal_bench_generator:_configure_harbor_logging:236 - Harbor logging level set to INFO
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.806 | INFO  | examples.terminal_bench.terminal_bench_generator:__init__:136 - TerminalBenchGenerator initialized with HarborConfigBuilder. Exposed fields: ['name', 'max_episodes', 'enable_summarize', 'store_all_messages', 'trajectory_config', 'enable_episode_logging', 'record_terminal_session', 'enable_pane_logging', 'strict_json_parser', 'interleaved_thinking', 'extra_body', 'override_timeout_sec', 'override_cpus', 'override_memory_mb', 'override_storage_mb', 'auto_snapshot', 'verifier_override_timeout_sec', 'max_retries', 'min_wait_sec', 'max_wait_sec', 'wait_multiplier', 'exclude_exceptions', 'n_concurrent_trials', 'log_level', 'enable_reward_shaping', 'enable_error_classification', 'mask_exceptions', 'default_error_treatment', 'passthrough_exceptions', 'zero_exceptions']. Retry config: max_retries=3, backoff=60.0-600.0s. Concurrent trials: 675. Reward shaping: enabled=False, shaper=pass_ratio. Error classification: enabled=True
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.817 | INFO  | examples.terminal_bench.terminal_bench_generator:__init__:152 - TerminalBenchGenerator initialized with custom chat template read from: chat_templates/qwen3_thinking_acc.jinja2
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.818 | INFO  | skyrl_train.utils.trainer_utils:build_dataloader:656 - Total steps: 156
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.818 | INFO  | skyrl_train.fully_async_trainer:_build_train_dataloader_and_compute_training_steps:357 - Length of train_dataloader: 5000
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.818 | INFO  | skyrl_train.fully_async_trainer:_build_train_dataloader_and_compute_training_steps:358 - Number of steps per epoch: 78
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:34.818 | INFO  | skyrl_train.fully_async_trainer:_build_train_dataloader_and_compute_training_steps:359 - Total training steps: 80
(raylet, ip=10.128.32.46) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 63x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored. [repeated 6x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 3x across cluster]
(pid=2955859, ip=10.128.32.39) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 15x across cluster]
(pid=2955859, ip=10.128.32.39) No module named 'vllm._version' [repeated 15x across cluster]
(pid=2955859, ip=10.128.32.39) from .version import __version__, __version_tuple__ # isort:skip [repeated 15x across cluster]
(pid=2623829, ip=10.128.32.46) Using blocking ray.get inside async actor. This blocks the event loop. Please use `await` on object ref with asyncio.gather if you want to yield execution to the event loop instead.
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:39.814 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:123 - setup_envvars_for_vllm: distributed_executor_backend=uni, SKYRL_ENABLE_NUMA_AFFINITY=1, CUDA_VISIBLE_DEVICES=1, VLLM_ENABLE_V1_MULTIPROCESSING=<unset> [repeated 2x across cluster]
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:39.815 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:133 - setup_envvars_for_vllm: numa_enabled=True [repeated 2x across cluster]
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:39.815 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:136 - setup_envvars_for_vllm: set VLLM_ENABLE_V1_MULTIPROCESSING=0 for NUMA affinity [repeated 2x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) 2026-06-07 02:12:38.211 | INFO | skyrl_train.utils.numa:set_numa_affinity_for_gpu:340 - NUMA affinity: bound process to CPUs 0-71 (NUMA node 0) for GPU 0
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) 2026-06-07 02:12:38.234 | DEBUG | skyrl_train.utils.numa:_set_membind_via_libnuma:375 - NUMA affinity: set memory preferred to NUMA node 0
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) 2026-06-07 02:12:38.234 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:584 - BaseVLLMInferenceEngine: vllm_v1_disable_multiproc=True, vllm.__version__=dev, VLLM_ENABLE_V1_MULTIPROCESSING=0
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) 2026-06-07 02:12:38.234 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:592 - BaseVLLMInferenceEngine: set VLLM_ENABLE_V1_MULTIPROCESSING=0
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) 2026-06-07 02:12:38.234 | WARNING | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1158 - OpenAI API sampling params overridden: temperature=0.7, top_p=0.95, top_k=20
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) 2026-06-07 02:12:38.253 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1210 - Engine startup stagger: sleeping 2.33s (attempt 1/5) to avoid port collisions
(bundle_reservation_check_func pid=2701396) [2026-06-07 02:12:39,951 E 2701396 2701436] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:40.760 | INFO | skyrl_train.utils.numa:set_numa_affinity_for_gpu:340 - NUMA affinity: bound process to CPUs 72-143 (NUMA node 1) for GPU 1
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:40.764 | DEBUG | skyrl_train.utils.numa:_set_membind_via_libnuma:375 - NUMA affinity: set memory preferred to NUMA node 1
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:40.764 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:584 - BaseVLLMInferenceEngine: vllm_v1_disable_multiproc=True, vllm.__version__=dev, VLLM_ENABLE_V1_MULTIPROCESSING=0
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:40.765 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:592 - BaseVLLMInferenceEngine: set VLLM_ENABLE_V1_MULTIPROCESSING=0
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:40.765 | WARNING | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1158 - OpenAI API sampling params overridden: temperature=0.7, top_p=0.95, top_k=20
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) 2026-06-07 02:12:40.793 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1210 - Engine startup stagger: sleeping 2.14s (attempt 1/5) to avoid port collisions
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 2x across cluster]
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored. [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 2x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash:
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) No module named 'vllm._version'
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) from .version import __version__, __version_tuple__ # isort:skip
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash:
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) No module named 'vllm._version'
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) from .version import __version__, __version_tuple__ # isort:skip
(AsyncVLLMInferenceEngine pid=3076043, ip=10.128.32.42) 2026-06-07 02:12:45.362 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:123 - setup_envvars_for_vllm: distributed_executor_backend=uni, SKYRL_ENABLE_NUMA_AFFINITY=1, CUDA_VISIBLE_DEVICES=0, VLLM_ENABLE_V1_MULTIPROCESSING=<unset> [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076043, ip=10.128.32.42) 2026-06-07 02:12:45.364 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:133 - setup_envvars_for_vllm: numa_enabled=True [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076043, ip=10.128.32.42) 2026-06-07 02:12:45.364 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:136 - setup_envvars_for_vllm: set VLLM_ENABLE_V1_MULTIPROCESSING=0 for NUMA affinity [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) [2026-06-07 02:12:41,839 E 2666524 2666626] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) 2026-06-07 02:12:45.860 | INFO | skyrl_train.utils.numa:set_numa_affinity_for_gpu:340 - NUMA affinity: bound process to CPUs 0-71 (NUMA node 0) for GPU 0 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) 2026-06-07 02:12:45.883 | DEBUG | skyrl_train.utils.numa:_set_membind_via_libnuma:375 - NUMA affinity: set memory preferred to NUMA node 0 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) 2026-06-07 02:12:45.883 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:584 - BaseVLLMInferenceEngine: vllm_v1_disable_multiproc=True, vllm.__version__=dev, VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) 2026-06-07 02:12:45.883 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:592 - BaseVLLMInferenceEngine: set VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) 2026-06-07 02:12:45.883 | WARNING | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1158 - OpenAI API sampling params overridden: temperature=0.7, top_p=0.95, top_k=20 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) 2026-06-07 02:12:45.916 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1210 - Engine startup stagger: sleeping 2.38s (attempt 1/5) to avoid port collisions [repeated 4x across cluster]
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) 2026-06-07 02:12:45 INFO [ipv4-debug] hostname=jpbo-041-46.jupiter.internal
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) 2026-06-07 02:12:45 INFO [ipv4-debug] _global_node.node_ip_address=10.128.32.46
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) 2026-06-07 02:12:45 INFO [ipv4-debug] get_node_ip_address()=10.128.32.46
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:46.122 | INFO  | skyrl_train.workers.worker:_initiate_actors:563 - Initializing process group for RayActorGroup
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [W607 02:12:46.969120626 socket.cpp:767] [c10d] The client socket cannot be initialized to connect to [jpbo-041-46.jupiter.internal]:46781 (errno: 97 - Address family not supported by protocol).
(raylet, ip=10.128.32.46) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2793258, ip=10.128.32.45) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored. [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2557696, ip=10.128.32.48) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2557696, ip=10.128.32.48) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2557696, ip=10.128.32.48) No module named 'vllm._version' [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2557696, ip=10.128.32.48) from .version import __version__, __version_tuple__ # isort:skip [repeated 3x across cluster]
(pid=2623903, ip=10.128.32.46) Using blocking ray.get inside async actor. This blocks the event loop. Please use `await` on object ref with asyncio.gather if you want to yield execution to the event loop instead.
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) (EngineCore_DP0 pid=2666891) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/__init__.py:1617: UserWarning: Please use the new API settings to control TF32 behavior, such as torch.backends.cudnn.conv.fp32_precision = 'tf32' or torch.backends.cuda.matmul.fp32_precision = 'ieee'. Old settings, e.g, torch.backends.cuda.matmul.allow_tf32 = True, torch.backends.cudnn.allow_tf32 = True, allowTF32CuDNN() and allowTF32CuBLAS() will be deprecated after Pytorch 2.9. Please see https://pytorch.org/docs/main/notes/cuda.html#tensorfloat-32-tf32-on-ampere-and-later-devices (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:80.)
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) (EngineCore_DP0 pid=2666891) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2666657, ip=10.128.32.38) (EngineCore_DP0 pid=2666896) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) (EngineCore_DP0 pid=2666900) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:49.608 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:123 - setup_envvars_for_vllm: distributed_executor_backend=uni, SKYRL_ENABLE_NUMA_AFFINITY=1, CUDA_VISIBLE_DEVICES=1, VLLM_ENABLE_V1_MULTIPROCESSING=<unset> [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:49.610 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:133 - setup_envvars_for_vllm: numa_enabled=True [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:49.610 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:setup_envvars_for_vllm:136 - setup_envvars_for_vllm: set VLLM_ENABLE_V1_MULTIPROCESSING=0 for NUMA affinity [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2557696, ip=10.128.32.48) [2026-06-07 02:12:49,438 E 2557696 2557862] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14 [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) [W607 02:12:50.707628733 Utils.hpp:166] Warning: Environment variable NCCL_BLOCKING_WAIT is deprecated; use TORCH_NCCL_BLOCKING_WAIT instead (function operator())
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) [rank0]:[W607 02:12:50.713854317 Utils.hpp:137] Warning: Environment variable TORCH_NCCL_TRACE_BUFFER_SIZE is deprecated; use TORCH_FR_BUFFER_SIZE instead (function operator())
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) 2026-06-07 02:12:50.870 | INFO | skyrl_train.utils.numa:set_numa_affinity_for_gpu:340 - NUMA affinity: bound process to CPUs 0-71 (NUMA node 0) for GPU 0 [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) 2026-06-07 02:12:50.889 | DEBUG | skyrl_train.utils.numa:_set_membind_via_libnuma:375 - NUMA affinity: set memory preferred to NUMA node 0 [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) 2026-06-07 02:12:50.889 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:584 - BaseVLLMInferenceEngine: vllm_v1_disable_multiproc=True, vllm.__version__=dev, VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) 2026-06-07 02:12:50.889 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:592 - BaseVLLMInferenceEngine: set VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) 2026-06-07 02:12:50.889 | WARNING | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1158 - OpenAI API sampling params overridden: temperature=0.7, top_p=0.95, top_k=20 [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) 2026-06-07 02:12:50.920 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1210 - Engine startup stagger: sleeping 2.99s (attempt 1/5) to avoid port collisions [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) [W607 02:12:50.703217538 socket.cpp:767] [c10d] The client socket cannot be initialized to connect to [jpbo-041-38-interconnect-1.jupiter.internal]:38761 (errno: 97 - Address family not supported by protocol). [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 107x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2955860, ip=10.128.32.39) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored. [repeated 71x across cluster]
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) (EngineCore_DP0 pid=2660131) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) No module named 'vllm._version' [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) from .version import __version__, __version_tuple__ # isort:skip [repeated 35x across cluster]
(pid=2701488) Using blocking ray.get inside async actor. This blocks the event loop. Please use `await` on object ref with asyncio.gather if you want to yield execution to the event loop instead. [repeated 6x across cluster]
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) (EngineCore_DP0 pid=2660139) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) Loading safetensors checkpoint shards: 0% Completed | 0/4 [00:00<?, ?it/s]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/__init__.py:1617: UserWarning: Please use the new API settings to control TF32 behavior, such as torch.backends.cudnn.conv.fp32_precision = 'tf32' or torch.backends.cuda.matmul.fp32_precision = 'ieee'. Old settings, e.g, torch.backends.cuda.matmul.allow_tf32 = True, torch.backends.cudnn.allow_tf32 = True, allowTF32CuDNN() and allowTF32CuBLAS() will be deprecated after Pytorch 2.9. Please see https://pytorch.org/docs/main/notes/cuda.html#tensorfloat-32-tf32-on-ampere-and-later-devices (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:80.) [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2593412, ip=10.128.32.36) [2026-06-07 02:12:55,179 E 2593412 2593580] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14 [repeated 19x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) [W607 02:12:55.527964802 Utils.hpp:166] Warning: Environment variable NCCL_BLOCKING_WAIT is deprecated; use TORCH_NCCL_BLOCKING_WAIT instead (function operator()) [repeated 10x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) [rank0]:[W607 02:12:55.530401044 Utils.hpp:137] Warning: Environment variable TORCH_NCCL_TRACE_BUFFER_SIZE is deprecated; use TORCH_FR_BUFFER_SIZE instead (function operator()) [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:50.861 | INFO | skyrl_train.utils.numa:set_numa_affinity_for_gpu:340 - NUMA affinity: bound process to CPUs 72-143 (NUMA node 1) for GPU 1 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:50.889 | DEBUG | skyrl_train.utils.numa:_set_membind_via_libnuma:375 - NUMA affinity: set memory preferred to NUMA node 1 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:50.889 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:584 - BaseVLLMInferenceEngine: vllm_v1_disable_multiproc=True, vllm.__version__=dev, VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:50.889 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:__init__:592 - BaseVLLMInferenceEngine: set VLLM_ENABLE_V1_MULTIPROCESSING=0 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:50.889 | WARNING | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1158 - OpenAI API sampling params overridden: temperature=0.7, top_p=0.95, top_k=20 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:12:50.917 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1210 - Engine startup stagger: sleeping 1.98s (attempt 1/5) to avoid port collisions [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) [W607 02:12:55.527535851 socket.cpp:767] [c10d] The client socket cannot be initialized to connect to [jpbo-041-43.jupiter.internal]:50889 (errno: 97 - Address family not supported by protocol). [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 8x across cluster]
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:57.653 | INFO  | skyrl_train.workers.worker:_initiate_actors:565 - Initialized process group for RayActorGroup
(skyrl_entrypoint pid=2701319) 2026-06-07 02:12:57.664 | INFO  | skyrl_train.workers.worker:_initiate_actors:567 - Mesh Ranks: [MeshRank(dp=0, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1), MeshRank(dp=1, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1), MeshRank(dp=2, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1), MeshRank(dp=3, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1), MeshRank(dp=4, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1), MeshRank(dp=5, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1), MeshRank(dp=6, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1), MeshRank(dp=7, sp=0, tp=0, pp=0, world_size=8, dp_size=8, pp_size=1)]
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) `torch_dtype` is deprecated! Use `dtype` instead!
(FSDPPolicyWorkerBase pid=2701489) Loading checkpoint shards: 0%| | 0/4 [00:00<?, ?it/s]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored. [repeated 6x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 4x across cluster]
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) Loading checkpoint shards: 25%|██▌ | 1/4 [00:00<00:00, 3.50it/s]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) No module named 'vllm._version' [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) from .version import __version__, __version_tuple__ # isort:skip [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) Loading safetensors checkpoint shards: 25% Completed | 1/4 [00:03<00:10, 3.51s/it]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) Loading safetensors checkpoint shards: 0% Completed | 0/4 [00:00<?, ?it/s] [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2793258, ip=10.128.32.45) (EngineCore_DP0 pid=2793487) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2593412, ip=10.128.32.36) (EngineCore_DP0 pid=2593646) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2593411, ip=10.128.32.36) (EngineCore_DP0 pid=2593642) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2593411, ip=10.128.32.36) (EngineCore_DP0 pid=2593642) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/__init__.py:1617: UserWarning: Please use the new API settings to control TF32 behavior, such as torch.backends.cudnn.conv.fp32_precision = 'tf32' or torch.backends.cuda.matmul.fp32_precision = 'ieee'. Old settings, e.g, torch.backends.cuda.matmul.allow_tf32 = True, torch.backends.cudnn.allow_tf32 = True, allowTF32CuDNN() and allowTF32CuBLAS() will be deprecated after Pytorch 2.9. Please see https://pytorch.org/docs/main/notes/cuda.html#tensorfloat-32-tf32-on-ampere-and-later-devices (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:80.) [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=3076043, ip=10.128.32.42) [2026-06-07 02:12:58,732 E 3076043 3076146] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14 [repeated 6x across cluster]
(AsyncVLLMInferenceEngine pid=290611, ip=10.128.32.47) (EngineCore_DP0 pid=290849) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2575228, ip=10.128.32.41) (EngineCore_DP0 pid=2575453) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=290737, ip=10.128.32.47) (EngineCore_DP0 pid=290838) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2557696, ip=10.128.32.48) (EngineCore_DP0 pid=2558131) _C._set_float32_matmul_precision(precision)
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [W607 02:12:57.986763802 Utils.hpp:166] Warning: Environment variable NCCL_BLOCKING_WAIT is deprecated; use TORCH_NCCL_BLOCKING_WAIT instead (function operator()) [repeated 5x across cluster]
(AsyncVLLMInferenceEngine pid=2557898, ip=10.128.32.48) (EngineCore_DP0 pid=2558139) _C._set_float32_matmul_precision(precision)
(FSDPPolicyWorkerBase pid=2701488) [W607 02:12:57.876876876 socket.cpp:767] [c10d] The client socket cannot be initialized to connect to [jpbo-041-46.jupiter.internal]:46781 (errno: 97 - Address family not supported by protocol). [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2793257, ip=10.128.32.45) (EngineCore_DP0 pid=2793503) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2575228, ip=10.128.32.41) [rank0]:[W607 02:13:01.642740161 Utils.hpp:137] Warning: Environment variable TORCH_NCCL_TRACE_BUFFER_SIZE is deprecated; use TORCH_FR_BUFFER_SIZE instead (function operator())
(AsyncVLLMInferenceEngine pid=2557899, ip=10.128.32.48) (EngineCore_DP0 pid=2558147) _C._set_float32_matmul_precision(precision)
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) Loading checkpoint shards: 100%|██████████| 4/4 [00:03<00:00, 1.05it/s]
(AsyncVLLMInferenceEngine pid=2593413, ip=10.128.32.36) (EngineCore_DP0 pid=2593666) _C._set_float32_matmul_precision(precision)
(FSDPPolicyWorkerBase pid=2701489) Loading checkpoint shards: 100%|██████████| 4/4 [00:03<00:00, 1.14it/s] Loading checkpoint shards: 100%|██████████| 4/4 [00:03<00:00, 1.06it/s]
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) (EngineCore_DP0 pid=2575477) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) (EngineCore_DP0 pid=2558151) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=290612, ip=10.128.32.47) (EngineCore_DP0 pid=290842) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2793126, ip=10.128.32.45) (EngineCore_DP0 pid=2793507) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) Process EngineCore_DP0:
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) Traceback (most recent call last):
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) self.run()
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/multiprocessing/process.py", line 108, in run
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) self._target(*self._args, **self._kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 1010, in run_engine_core
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) raise e
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 996, in run_engine_core
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) engine_core = EngineCoreProc(*args, engine_index=dp_rank, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 740, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) super().__init__(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 106, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) self.model_executor = executor_class(vllm_config)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/executor/abstract.py", line 103, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) self._init_executor()
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/executor/uniproc_executor.py", line 47, in _init_executor
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) self.driver_worker.init_device()
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/worker/worker_base.py", line 332, in init_device
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) self.worker.init_device() # type: ignore
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/worker/gpu_worker.py", line 232, in init_device
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) init_worker_distributed_environment(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/worker/gpu_worker.py", line 1040, in init_worker_distributed_environment
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) init_distributed_environment(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/distributed/parallel_state.py", line 1256, in init_distributed_environment
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) torch.distributed.init_process_group(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/c10d_logger.py", line 81, in wrapper
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) return func(*args, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/c10d_logger.py", line 95, in wrapper
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) func_return = func(*args, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/distributed_c10d.py", line 1762, in init_process_group
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) store, rank, world_size = next(rendezvous_iterator)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/rendezvous.py", line 230, in _tcp_rendezvous_handler
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) store = _create_c10d_store(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/rendezvous.py", line 198, in _create_c10d_store
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) return TCPStore(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) torch.distributed.DistNetworkError: The server socket has failed to listen on any local network address. port: 37945, useIpv6: false, code: -98, name: EADDRINUSE, message: address already in use
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) Exception raised from makeWithPort at /pytorch/torch/csrc/distributed/c10d/TCPStoreLibUvBackend.cpp:307 (most recent call first):
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0xb0 (0x4000730cc700 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libc10.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #1: <unknown function> + 0x5f29220 (0x400053269220 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #2: <unknown function> + 0x5f4326c (0x40005328326c in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #3: <unknown function> + 0x5f49074 (0x400053289074 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #4: <unknown function> + 0x5f49138 (0x400053289138 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #5: <unknown function> + 0x5f2ccc4 (0x40005326ccc4 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #6: c10d::TCPStore::TCPStore(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, c10d::TCPStoreOptions const&) + 0x104 (0x400053271564 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #7: <unknown function> + 0x109a094 (0x40004d01a094 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #8: <unknown function> + 0x113236c (0x40004d0b236c in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #9: <unknown function> + 0x5d6d60 (0x40004c556d60 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #10: <unknown function> + 0x1b7a38 (0xaaaabc307a38 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #11: _PyObject_MakeTpCall + 0x98 (0xaaaabc2b5db8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #12: <unknown function> + 0x169f50 (0xaaaabc2b9f50 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #13: <unknown function> + 0x1682e4 (0xaaaabc2b82e4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #14: <unknown function> + 0x1e0ce8 (0xaaaabc330ce8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #15: <unknown function> + 0x1d7ddc (0xaaaabc327ddc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #16: <unknown function> + 0x646b0c (0x40004c5c6b0c in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #17: _PyObject_MakeTpCall + 0x98 (0xaaaabc2b5db8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #18: _PyEval_EvalFrameDefault + 0x280c (0xaaaabc3bae54 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #19: <unknown function> + 0x1808c0 (0xaaaabc2d08c0 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #20: <unknown function> + 0x182bf8 (0xaaaabc2d2bf8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #21: <unknown function> + 0x25fd30 (0xaaaabc3afd30 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #22: <unknown function> + 0x1b7d20 (0xaaaabc307d20 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #23: PyObject_Vectorcall + 0x54 (0xaaaabc2b60e4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #24: _PyEval_EvalFrameDefault + 0x280c (0xaaaabc3bae54 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #25: _PyObject_FastCallDictTstate + 0x80 (0xaaaabc2b7f20 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #26: _PyObject_Call_Prepend + 0x140 (0xaaaabc2b81ec in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #27: <unknown function> + 0x1e0d80 (0xaaaabc330d80 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #28: <unknown function> + 0x1d7ddc (0xaaaabc327ddc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #29: _PyObject_MakeTpCall + 0x98 (0xaaaabc2b5db8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #30: _PyEval_EvalFrameDefault + 0x280c (0xaaaabc3bae54 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #31: _PyObject_FastCallDictTstate + 0x10c (0xaaaabc2b7fac in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #32: _PyObject_Call_Prepend + 0x140 (0xaaaabc2b81ec in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #33: <unknown function> + 0x1e0d80 (0xaaaabc330d80 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #34: <unknown function> + 0x1d7ddc (0xaaaabc327ddc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #35: _PyObject_Call + 0x68 (0xaaaabc2b8488 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #36: _PyEval_EvalFrameDefault + 0x52ac (0xaaaabc3bd8f4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #37: PyEval_EvalCode + 0xb4 (0xaaaabc3c2eb4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #38: <unknown function> + 0x2ccdcc (0xaaaabc41cdcc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #39: <unknown function> + 0x2ccef4 (0xaaaabc41cef4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #40: PyRun_StringFlags + 0x90 (0xaaaabc421050 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #41: PyRun_SimpleStringFlags + 0x58 (0xaaaabc4210f8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #42: Py_RunMain + 0x2c8 (0xaaaabc449190 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #43: Py_BytesMain + 0x64 (0xaaaabc449fb4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #44: <unknown function> + 0x27540 (0x40003b5c7540 in /lib64/libc.so.6)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #45: __libc_start_main + 0x98 (0x40003b5c7618 in /lib64/libc.so.6)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) frame #46: <unknown function> + 0x10e0c0 (0xaaaabc25e0c0 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) [pynccl] dumped 0 entries to /e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps/nccl_trace_pynccl_pid2593652 (reason=atexit)
(FSDPPolicyWorkerBase pid=2701488) `torch_dtype` is deprecated! Use `dtype` instead! [repeated 7x across cluster]
(FSDPPolicyWorkerBase pid=2623904, ip=10.128.32.46) Loading checkpoint shards: 0%| | 0/4 [00:00<?, ?it/s] [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) (EngineCore_DP0 pid=3076408) _C._set_float32_matmul_precision(precision)
(FSDPPolicyWorkerBase pid=2701488) Loading checkpoint shards: 75%|███████▌ | 3/4 [00:03<00:01, 1.38s/it] [repeated 23x across cluster]
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) (EngineCore_DP0 pid=290850) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2575093, ip=10.128.32.41) (EngineCore_DP0 pid=2575462) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) (EngineCore_DP0 pid=2666891) Loading safetensors checkpoint shards: 50% Completed | 2/4 [00:08<00:07, 4.00s/it] [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=3076172, ip=10.128.32.42) (EngineCore_DP0 pid=3076398) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2972527, ip=10.128.32.44) (EngineCore_DP0 pid=2972899) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2539108, ip=10.128.32.40) (EngineCore_DP0 pid=2539361) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2538980, ip=10.128.32.40) (EngineCore_DP0 pid=2539337) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2972656, ip=10.128.32.44) (EngineCore_DP0 pid=2972886) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=3076173, ip=10.128.32.42) (EngineCore_DP0 pid=3076402) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2972657, ip=10.128.32.44) (EngineCore_DP0 pid=2972891) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2539109, ip=10.128.32.40) (EngineCore_DP0 pid=2539345) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) Exception raised in creation task: The actor died because of an error raised in its creation task, ray::AsyncVLLMInferenceEngine.__init__() (pid=2593284, ip=10.128.32.36, actor_id=a9ee7a598f6c230f2b02b9e202000000, repr=<skyrl_train.inference_engines.vllm.vllm_engine.AsyncVLLMInferenceEngine object at 0x400c0f140ec0>)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 449, in result
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) return self.__get_result()
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) raise self._exception
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1139, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) super().__init__(*args, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 603, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) self.llm = self._create_engine(*args, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1216, in _create_engine
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) engine = vllm.AsyncLLMEngine.from_engine_args(engine_args, stat_loggers=stat_loggers)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 251, in from_engine_args
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) return cls(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 148, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) self.engine_core = EngineCoreClient.make_async_mp_client(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 124, in make_async_mp_client
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) return AsyncMPClient(*client_args)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 835, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) super().__init__(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 490, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) with launch_core_engines(vllm_config, executor_class, log_stats) as (
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/contextlib.py", line 144, in __exit__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) next(self.gen)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 936, in launch_core_engines
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) wait_for_engine_startup(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 995, in wait_for_engine_startup
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) raise RuntimeError(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) [pynccl] dumped 0 entries to /e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps/nccl_trace_pynccl_pid2593284 (reason=atexit)
(AsyncVLLMInferenceEngine pid=984183, ip=10.128.32.37) (EngineCore_DP0 pid=984423) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=3076043, ip=10.128.32.42) (EngineCore_DP0 pid=3076412) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=290611, ip=10.128.32.47) (EngineCore_DP0 pid=290849) Loading safetensors checkpoint shards: 0% Completed | 0/4 [00:00<?, ?it/s] [repeated 14x across cluster]
(AsyncVLLMInferenceEngine pid=2955861, ip=10.128.32.39) (EngineCore_DP0 pid=2956093) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=984182, ip=10.128.32.37) (EngineCore_DP0 pid=984418) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2539107, ip=10.128.32.40) (EngineCore_DP0 pid=2539341) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2972655, ip=10.128.32.44) (EngineCore_DP0 pid=2972907) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=984055, ip=10.128.32.37) (EngineCore_DP0 pid=984427) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=984055, ip=10.128.32.37) (EngineCore_DP0 pid=984427) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/__init__.py:1617: UserWarning: Please use the new API settings to control TF32 behavior, such as torch.backends.cudnn.conv.fp32_precision = 'tf32' or torch.backends.cuda.matmul.fp32_precision = 'ieee'. Old settings, e.g, torch.backends.cuda.matmul.allow_tf32 = True, torch.backends.cudnn.allow_tf32 = True, allowTF32CuDNN() and allowTF32CuBLAS() will be deprecated after Pytorch 2.9. Please see https://pytorch.org/docs/main/notes/cuda.html#tensorfloat-32-tf32-on-ampere-and-later-devices (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:80.) [repeated 33x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) [2026-06-07 02:13:04,817 E 2955859 2955971] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14 [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955860, ip=10.128.32.39) (EngineCore_DP0 pid=2956089) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=984310, ip=10.128.32.37) (EngineCore_DP0 pid=984419) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=984183, ip=10.128.32.37) [W607 02:13:05.754325616 Utils.hpp:166] Warning: Environment variable NCCL_BLOCKING_WAIT is deprecated; use TORCH_NCCL_BLOCKING_WAIT instead (function operator()) [repeated 34x across cluster]
(AsyncVLLMInferenceEngine pid=984310, ip=10.128.32.37) [W607 02:13:06.082184402 socket.cpp:767] [c10d] The client socket cannot be initialized to connect to [jpbo-041-37.jupiter.internal]:34951 (errno: 97 - Address family not supported by protocol). [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=984310, ip=10.128.32.37) [rank0]:[W607 02:13:06.084447273 Utils.hpp:137] Warning: Environment variable TORCH_NCCL_TRACE_BUFFER_SIZE is deprecated; use TORCH_FR_BUFFER_SIZE instead (function operator()) [repeated 34x across cluster]
(FSDPPolicyWorkerBase pid=2623904, ip=10.128.32.46) Loading checkpoint shards: 100%|██████████| 4/4 [00:03<00:00, 1.05it/s] [repeated 3x across cluster]
(FSDPPolicyWorkerBase pid=2701488) Loading checkpoint shards: 100%|██████████| 4/4 [00:03<00:00, 1.14it/s] Loading checkpoint shards: 100%|██████████| 4/4 [00:03<00:00, 1.06it/s] [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113) _C._set_float32_matmul_precision(precision)
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123)
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) (EngineCore_DP0 pid=2660139)
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) (EngineCore_DP0 pid=2660131)
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148)
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870)
(AsyncVLLMInferenceEngine pid=2666657, ip=10.128.32.38) (EngineCore_DP0 pid=2666896)
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) (EngineCore_DP0 pid=2666900)
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) (EngineCore_DP0 pid=2666891)
(AsyncVLLMInferenceEngine pid=290737, ip=10.128.32.47) (EngineCore_DP0 pid=290838) Loading safetensors checkpoint shards: 25% Completed | 1/4 [00:03<00:09, 3.22s/it] [repeated 47x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) Loading safetensors checkpoint shards: 0% Completed | 0/4 [00:00<?, ?it/s] [repeated 25x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/__init__.py:1617: UserWarning: Please use the new API settings to control TF32 behavior, such as torch.backends.cudnn.conv.fp32_precision = 'tf32' or torch.backends.cuda.matmul.fp32_precision = 'ieee'. Old settings, e.g, torch.backends.cuda.matmul.allow_tf32 = True, torch.backends.cudnn.allow_tf32 = True, allowTF32CuDNN() and allowTF32CuBLAS() will be deprecated after Pytorch 2.9. Please see https://pytorch.org/docs/main/notes/cuda.html#tensorfloat-32-tf32-on-ampere-and-later-devices (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:80.) [repeated 4x across cluster]
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [2026-06-07 02:13:05,834 E 2623829 2623869] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) [W607 02:13:07.834305247 Utils.hpp:166] Warning: Environment variable NCCL_BLOCKING_WAIT is deprecated; use TORCH_NCCL_BLOCKING_WAIT instead (function operator()) [repeated 5x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) [W607 02:13:07.833825704 socket.cpp:767] [c10d] The client socket cannot be initialized to connect to [jpbo-041-39-interconnect-1.jupiter.internal]:38597 (errno: 97 - Address family not supported by protocol). [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) [rank0]:[W607 02:13:07.836433304 Utils.hpp:137] Warning: Environment variable TORCH_NCCL_TRACE_BUFFER_SIZE is deprecated; use TORCH_FR_BUFFER_SIZE instead (function operator()) [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) 2026-06-07 02:13:12.381 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1267 - Initializing OpenAIServingChat with custom_chat_template read from: chat_templates/qwen3_thinking_acc.jinja2
(AsyncVLLMInferenceEngine pid=2575093, ip=10.128.32.41) (EngineCore_DP0 pid=2575462)
(AsyncVLLMInferenceEngine pid=2575228, ip=10.128.32.41) (EngineCore_DP0 pid=2575453)
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) (EngineCore_DP0 pid=2575477)
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458)
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:14.347 | INFO  | skyrl_train.trainer:build_models:831 - init policy/ref/critic models done
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:14.366 | INFO  | skyrl_train.fully_async_trainer:_maybe_enable_rollout_fanout:456 - Rollout fan-out ENABLED: replacing single-process generator with RolloutDispatcher (K=4, cpus_per_coordinator=8).
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:14.388 | INFO  | examples.terminal_bench.rollout_coordinator:__init__:387 - [RolloutDispatcher] fan-out path: overriding inference host 127.0.0.1 -> 10.128.32.34 (routable head IP) for coordinator litellm base_url connectivity
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:14.388 | INFO  | examples.terminal_bench.rollout_coordinator:__init__:405 - [RolloutDispatcher] configured num_coordinators=4, cpus_per_coordinator=8
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:14.409 | INFO  | examples.terminal_bench.rollout_coordinator:startup:434 - [RolloutDispatcher] PlacementGroup ready: 4 bundles x 8 CPU (SPREAD)
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) Loading safetensors checkpoint shards: 25% Completed | 1/4 [00:03<00:10, 3.57s/it] [repeated 47x across cluster]
(raylet, ip=10.128.32.39) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e
(AsyncVLLMInferenceEngine pid=2593411, ip=10.128.32.36) (EngineCore_DP0 pid=2593642)
(AsyncVLLMInferenceEngine pid=2593413, ip=10.128.32.36) (EngineCore_DP0 pid=2593666)
(AsyncVLLMInferenceEngine pid=2593412, ip=10.128.32.36) (EngineCore_DP0 pid=2593646)
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 11x across cluster]
(FSDPPolicyWorkerBase pid=2623904, ip=10.128.32.46) [rank3]:[W607 02:13:12.044597514 Utils.hpp:137] Warning: Environment variable TORCH_NCCL_TRACE_BUFFER_SIZE is deprecated; use TORCH_FR_BUFFER_SIZE instead (function operator()) [repeated 8x across cluster]
(FSDPPolicyWorkerBase pid=2623903, ip=10.128.32.46) [2026-06-07 02:13:16,931 E 2623903 2624043] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(AsyncVLLMInferenceEngine pid=290484, ip=10.128.32.47) (EngineCore_DP0 pid=290850)
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) 2026-06-07 02:13:16.277 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1267 - Initializing OpenAIServingChat with custom_chat_template read from: chat_templates/qwen3_thinking_acc.jinja2 [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=290611, ip=10.128.32.47) (EngineCore_DP0 pid=290849)
(AsyncVLLMInferenceEngine pid=290612, ip=10.128.32.47) (EngineCore_DP0 pid=290842)
(AsyncVLLMInferenceEngine pid=290737, ip=10.128.32.47) (EngineCore_DP0 pid=290838)
(AsyncVLLMInferenceEngine pid=2557696, ip=10.128.32.48) (EngineCore_DP0 pid=2558131)
(AsyncVLLMInferenceEngine pid=2557898, ip=10.128.32.48) (EngineCore_DP0 pid=2558139)
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) (EngineCore_DP0 pid=2558151)
(AsyncVLLMInferenceEngine pid=2557899, ip=10.128.32.48) (EngineCore_DP0 pid=2558147)
(AsyncVLLMInferenceEngine pid=2793126, ip=10.128.32.45) (EngineCore_DP0 pid=2793507)
(AsyncVLLMInferenceEngine pid=2793257, ip=10.128.32.45) (EngineCore_DP0 pid=2793503)
(AsyncVLLMInferenceEngine pid=2793258, ip=10.128.32.45) (EngineCore_DP0 pid=2793487)
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492)
(AsyncVLLMInferenceEngine pid=2972527, ip=10.128.32.44) (EngineCore_DP0 pid=2972899)
(AsyncVLLMInferenceEngine pid=2972527, ip=10.128.32.44) (EngineCore_DP0 pid=2972899) Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:13<00:00, 3.30s/it] [repeated 79x across cluster]
(raylet, ip=10.128.32.39) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 5x across cluster]
(AsyncVLLMInferenceEngine pid=2972655, ip=10.128.32.44) (EngineCore_DP0 pid=2972907)
(AsyncVLLMInferenceEngine pid=2972656, ip=10.128.32.44) (EngineCore_DP0 pid=2972886)
(AsyncVLLMInferenceEngine pid=2972657, ip=10.128.32.44) (EngineCore_DP0 pid=2972891)
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.159 | INFO | examples.terminal_bench.rollout_coordinator:_scale_terminal_bench_cfg:121 - [RolloutCoordinator] scaled n_concurrent_trials 675 -> 168 (// 4)
(AsyncVLLMInferenceEngine pid=2538980, ip=10.128.32.40) (EngineCore_DP0 pid=2539337)
(AsyncVLLMInferenceEngine pid=2539109, ip=10.128.32.40) (EngineCore_DP0 pid=2539345)
(AsyncVLLMInferenceEngine pid=2539108, ip=10.128.32.40) (EngineCore_DP0 pid=2539361)
(AsyncVLLMInferenceEngine pid=2539107, ip=10.128.32.40) (EngineCore_DP0 pid=2539341)
(AsyncVLLMInferenceEngine pid=3076043, ip=10.128.32.42) (EngineCore_DP0 pid=3076412)
(AsyncVLLMInferenceEngine pid=3076173, ip=10.128.32.42) (EngineCore_DP0 pid=3076402)
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) (EngineCore_DP0 pid=3076408)
(AsyncVLLMInferenceEngine pid=3076172, ip=10.128.32.42) (EngineCore_DP0 pid=3076398)
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.385 | INFO | examples.terminal_bench.terminal_bench_generator:_configure_harbor_logging:236 - Harbor logging level set to INFO
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.385 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:136 - TerminalBenchGenerator initialized with HarborConfigBuilder. Exposed fields: ['name', 'max_episodes', 'enable_summarize', 'store_all_messages', 'trajectory_config', 'enable_episode_logging', 'record_terminal_session', 'enable_pane_logging', 'strict_json_parser', 'interleaved_thinking', 'extra_body', 'override_timeout_sec', 'override_cpus', 'override_memory_mb', 'override_storage_mb', 'auto_snapshot', 'verifier_override_timeout_sec', 'max_retries', 'min_wait_sec', 'max_wait_sec', 'wait_multiplier', 'exclude_exceptions', 'n_concurrent_trials', 'log_level', 'enable_reward_shaping', 'enable_error_classification', 'mask_exceptions', 'default_error_treatment', 'passthrough_exceptions', 'zero_exceptions']. Retry config: max_retries=3, backoff=60.0-600.0s. Concurrent trials: 168. Reward shaping: enabled=False, shaper=pass_ratio. Error classification: enabled=True
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.387 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:152 - TerminalBenchGenerator initialized with custom chat template read from: chat_templates/qwen3_thinking_acc.jinja2
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.387 | INFO | examples.terminal_bench.rollout_coordinator:__init__:246 - [RolloutCoordinator 0/4] constructed (http=10.128.32.34:8000)
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.390 | INFO | examples.terminal_bench.terminal_bench_generator:_create_orchestrator:292 - QueueOrchestrator created and started with n_concurrent_trials=168, rollback_hook registered for ContextLengthExceededError/AgentTimeoutError
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.390 | INFO | examples.terminal_bench.terminal_bench_generator:startup:249 - TerminalBenchGenerator startup complete. Shared orchestrator ready with n_concurrent_trials=168
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:13:20.390 | INFO | examples.terminal_bench.rollout_coordinator:startup:254 - [RolloutCoordinator 0] startup complete
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:20.391 | INFO  | examples.terminal_bench.rollout_coordinator:startup:469 - [RolloutDispatcher] coordinator 1/4 started
(AsyncVLLMInferenceEngine pid=984055, ip=10.128.32.37) (EngineCore_DP0 pid=984427)
(AsyncVLLMInferenceEngine pid=984182, ip=10.128.32.37) (EngineCore_DP0 pid=984418)
(AsyncVLLMInferenceEngine pid=984183, ip=10.128.32.37) (EngineCore_DP0 pid=984423)
(AsyncVLLMInferenceEngine pid=984310, ip=10.128.32.37) (EngineCore_DP0 pid=984419)
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113)
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) (EngineCore_DP0 pid=2558151) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 16x across cluster]
(FSDPPolicyWorkerBase pid=2701488) [2026-06-07 02:13:17,107 E 2701488 2701649] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14 [repeated 6x across cluster]
(AsyncVLLMInferenceEngine pid=2955860, ip=10.128.32.39) (EngineCore_DP0 pid=2956089)
(AsyncVLLMInferenceEngine pid=2955861, ip=10.128.32.39) (EngineCore_DP0 pid=2956093)
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097)
(AsyncVLLMInferenceEngine pid=2793126, ip=10.128.32.45) 2026-06-07 02:13:21.517 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1267 - Initializing OpenAIServingChat with custom_chat_template read from: chat_templates/qwen3_thinking_acc.jinja2 [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:13<00:00, 3.30s/it] [repeated 46x across cluster]
(raylet, ip=10.128.32.37) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 6x across cluster]
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.158 | INFO | examples.terminal_bench.rollout_coordinator:_scale_terminal_bench_cfg:121 - [RolloutCoordinator] scaled n_concurrent_trials 675 -> 168 (// 4)
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue. [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) 2026-06-07 02:13:26.137 | INFO | skyrl_train.inference_engines.vllm.vllm_engine:_create_engine:1267 - Initializing OpenAIServingChat with custom_chat_template read from: chat_templates/qwen3_thinking_acc.jinja2 [repeated 20x across cluster]
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.376 | INFO | examples.terminal_bench.terminal_bench_generator:_configure_harbor_logging:236 - Harbor logging level set to INFO
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.376 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:136 - TerminalBenchGenerator initialized with HarborConfigBuilder. Exposed fields: ['name', 'max_episodes', 'enable_summarize', 'store_all_messages', 'trajectory_config', 'enable_episode_logging', 'record_terminal_session', 'enable_pane_logging', 'strict_json_parser', 'interleaved_thinking', 'extra_body', 'override_timeout_sec', 'override_cpus', 'override_memory_mb', 'override_storage_mb', 'auto_snapshot', 'verifier_override_timeout_sec', 'max_retries', 'min_wait_sec', 'max_wait_sec', 'wait_multiplier', 'exclude_exceptions', 'n_concurrent_trials', 'log_level', 'enable_reward_shaping', 'enable_error_classification', 'mask_exceptions', 'default_error_treatment', 'passthrough_exceptions', 'zero_exceptions']. Retry config: max_retries=3, backoff=60.0-600.0s. Concurrent trials: 168. Reward shaping: enabled=False, shaper=pass_ratio. Error classification: enabled=True
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.376 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:152 - TerminalBenchGenerator initialized with custom chat template read from: chat_templates/qwen3_thinking_acc.jinja2
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.376 | INFO | examples.terminal_bench.rollout_coordinator:__init__:246 - [RolloutCoordinator 1/4] constructed (http=10.128.32.34:8000)
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.379 | INFO | examples.terminal_bench.terminal_bench_generator:_create_orchestrator:292 - QueueOrchestrator created and started with n_concurrent_trials=168, rollback_hook registered for ContextLengthExceededError/AgentTimeoutError
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.379 | INFO | examples.terminal_bench.terminal_bench_generator:startup:249 - TerminalBenchGenerator startup complete. Shared orchestrator ready with n_concurrent_trials=168
(RolloutCoordinator pid=984614, ip=10.128.32.37) 2026-06-07 02:13:28.379 | INFO | examples.terminal_bench.rollout_coordinator:startup:254 - [RolloutCoordinator 1] startup complete
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:28.381 | INFO  | examples.terminal_bench.rollout_coordinator:startup:469 - [RolloutDispatcher] coordinator 2/4 started
(raylet, ip=10.128.32.43) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e
(raylet, ip=10.128.32.43) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.536 | INFO | examples.terminal_bench.rollout_coordinator:_scale_terminal_bench_cfg:121 - [RolloutCoordinator] scaled n_concurrent_trials 675 -> 168 (// 4)
(RolloutCoordinator pid=984614, ip=10.128.32.37) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
(raylet, ip=10.128.32.43) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 4x across cluster]
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:36.763 | INFO  | examples.terminal_bench.rollout_coordinator:startup:469 - [RolloutDispatcher] coordinator 3/4 started
(RolloutCoordinator pid=2660382, ip=10.128.32.43) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.758 | INFO | examples.terminal_bench.terminal_bench_generator:_configure_harbor_logging:236 - Harbor logging level set to INFO
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.758 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:136 - TerminalBenchGenerator initialized with HarborConfigBuilder. Exposed fields: ['name', 'max_episodes', 'enable_summarize', 'store_all_messages', 'trajectory_config', 'enable_episode_logging', 'record_terminal_session', 'enable_pane_logging', 'strict_json_parser', 'interleaved_thinking', 'extra_body', 'override_timeout_sec', 'override_cpus', 'override_memory_mb', 'override_storage_mb', 'auto_snapshot', 'verifier_override_timeout_sec', 'max_retries', 'min_wait_sec', 'max_wait_sec', 'wait_multiplier', 'exclude_exceptions', 'n_concurrent_trials', 'log_level', 'enable_reward_shaping', 'enable_error_classification', 'mask_exceptions', 'default_error_treatment', 'passthrough_exceptions', 'zero_exceptions']. Retry config: max_retries=3, backoff=60.0-600.0s. Concurrent trials: 168. Reward shaping: enabled=False, shaper=pass_ratio. Error classification: enabled=True
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.758 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:152 - TerminalBenchGenerator initialized with custom chat template read from: chat_templates/qwen3_thinking_acc.jinja2
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.758 | INFO | examples.terminal_bench.rollout_coordinator:__init__:246 - [RolloutCoordinator 2/4] constructed (http=10.128.32.34:8000)
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.761 | INFO | examples.terminal_bench.terminal_bench_generator:_create_orchestrator:292 - QueueOrchestrator created and started with n_concurrent_trials=168, rollback_hook registered for ContextLengthExceededError/AgentTimeoutError
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.761 | INFO | examples.terminal_bench.terminal_bench_generator:startup:249 - TerminalBenchGenerator startup complete. Shared orchestrator ready with n_concurrent_trials=168
(RolloutCoordinator pid=2660382, ip=10.128.32.43) 2026-06-07 02:13:36.761 | INFO | examples.terminal_bench.rollout_coordinator:startup:254 - [RolloutCoordinator 2] startup complete
(RolloutCoordinator pid=2956288, ip=10.128.32.39) [2026-06-07 02:13:45,184 E 2956288 2956328] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(raylet, ip=10.128.32.36) [proxychains] DLL init: proxychains-ng 4.17-git-9-g78ead0e [repeated 6x across cluster]
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.382 | INFO | examples.terminal_bench.rollout_coordinator:_scale_terminal_bench_cfg:121 - [RolloutCoordinator] scaled n_concurrent_trials 675 -> 168 (// 4)
(RolloutCoordinator pid=2593907, ip=10.128.32.36) The tokenizer you are loading from '/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.604 | INFO | examples.terminal_bench.terminal_bench_generator:_configure_harbor_logging:236 - Harbor logging level set to INFO
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.604 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:136 - TerminalBenchGenerator initialized with HarborConfigBuilder. Exposed fields: ['name', 'max_episodes', 'enable_summarize', 'store_all_messages', 'trajectory_config', 'enable_episode_logging', 'record_terminal_session', 'enable_pane_logging', 'strict_json_parser', 'interleaved_thinking', 'extra_body', 'override_timeout_sec', 'override_cpus', 'override_memory_mb', 'override_storage_mb', 'auto_snapshot', 'verifier_override_timeout_sec', 'max_retries', 'min_wait_sec', 'max_wait_sec', 'wait_multiplier', 'exclude_exceptions', 'n_concurrent_trials', 'log_level', 'enable_reward_shaping', 'enable_error_classification', 'mask_exceptions', 'default_error_treatment', 'passthrough_exceptions', 'zero_exceptions']. Retry config: max_retries=3, backoff=60.0-600.0s. Concurrent trials: 168. Reward shaping: enabled=False, shaper=pass_ratio. Error classification: enabled=True
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.604 | INFO | examples.terminal_bench.terminal_bench_generator:__init__:152 - TerminalBenchGenerator initialized with custom chat template read from: chat_templates/qwen3_thinking_acc.jinja2
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.604 | INFO | examples.terminal_bench.rollout_coordinator:__init__:246 - [RolloutCoordinator 3/4] constructed (http=10.128.32.34:8000)
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.607 | INFO | examples.terminal_bench.terminal_bench_generator:_create_orchestrator:292 - QueueOrchestrator created and started with n_concurrent_trials=168, rollback_hook registered for ContextLengthExceededError/AgentTimeoutError
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.607 | INFO | examples.terminal_bench.terminal_bench_generator:startup:249 - TerminalBenchGenerator startup complete. Shared orchestrator ready with n_concurrent_trials=168
(RolloutCoordinator pid=2593907, ip=10.128.32.36) 2026-06-07 02:13:45.607 | INFO | examples.terminal_bench.rollout_coordinator:startup:254 - [RolloutCoordinator 3] startup complete
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.608 | INFO  | examples.terminal_bench.rollout_coordinator:startup:469 - [RolloutDispatcher] coordinator 4/4 started
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.608 | INFO  | examples.terminal_bench.rollout_coordinator:startup:477 - [RolloutDispatcher] 4 coordinators started
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.608 | INFO  | skyrl_train.fully_async_trainer:train:485 - Generator startup complete
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.608 | INFO  | skyrl_train.fully_async_trainer:_train_loop:510 - Started: 'load_checkpoints'
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.617 | INFO  | skyrl_train.trainer:load_checkpoints:1664 - Loading checkpoint from: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints/global_step_28
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.617 | INFO  | skyrl_train.trainer:load_checkpoints:1670 - Resuming from global_step: 28
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.646 | INFO  | skyrl_train.trainer:load_checkpoints:1686 - Successfully loaded trainer state
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.660 | INFO  | skyrl_train.trainer:load_checkpoints:1696 - Successfully loaded dataloader state
(skyrl_entrypoint pid=2701319) 2026-06-07 02:13:45.660 | INFO  | skyrl_train.trainer:load_checkpoints:1705 - Loading policy checkpoint from /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints/global_step_28/policy
(RolloutCoordinator pid=984614, ip=10.128.32.37) [2026-06-07 02:13:53,284 E 984614 984654] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(RolloutCoordinator pid=2660382, ip=10.128.32.43) [2026-06-07 02:14:01,250 E 2660382 2660422] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/distributed_c10d.py:4876: UserWarning: barrier(): using the device under current context. You can specify `device_id` in `init_process_group` to mute this warning.
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) warnings.warn( # warn only once
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:03.113 | INFO  | skyrl_train.trainer:load_checkpoints:1715 - Successfully loaded policy checkpoint
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:03.114 | INFO  | skyrl_train.trainer:load_checkpoints:1731 - Successfully loaded complete checkpoint state from global_step_28
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:03.114 | INFO  | skyrl_train.fully_async_trainer:_train_loop:512 - Resumed training from global_step 28
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:03.131 | INFO  | skyrl_train.utils.data_tracker:load_state:97 - Loaded data tracker state: epoch=0, consumed_in_epoch=1792, total_consumed=1792
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:03.132 | INFO  | skyrl_train.fully_async_trainer:_train_loop:510 - Finished: 'load_checkpoints', time cost: 17.52s
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:03.132 | INFO  | skyrl_train.fully_async_trainer:_train_loop:544 - Started: 'init_weight_sync_state'
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash:
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) No module named 'vllm._version'
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) from .version import __version__, __version_tuple__ # isort:skip
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) 2026-06-07 02:14:08.866 | INFO  | logging:info:2216 - [weight-sync] Using master_addr=10.128.32.46, master_port=59681
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) 2026-06-07 02:14:08.865 | INFO  | logging:info:2216 - [weight-sync] get_node_ip_address()=10.128.32.46 [repeated 3x across cluster]
(FSDPPolicyWorkerBase pid=2701490) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/__init__.py:7: RuntimeWarning: Failed to read commit hash: [repeated 7x across cluster]
(FSDPPolicyWorkerBase pid=2701490) No module named 'vllm._version' [repeated 7x across cluster]
(FSDPPolicyWorkerBase pid=2701490) from .version import __version__, __version_tuple__ # isort:skip [repeated 7x across cluster]
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank0]:[W607 02:14:08.715725173 socket.cpp:767] [c10d] The client socket cannot be initialized to connect to [jpbo-041-46-interconnect-1.jupiter.internal]:59681 (errno: 97 - Address family not supported by protocol).
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:08.938 | INFO  | skyrl_train.fully_async_trainer:_train_loop:544 - Finished: 'init_weight_sync_state', time cost: 5.81s
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:08.938 | ERROR  | skyrl_train.fully_async_trainer:train:493 - Train loop failed at global_step 28: init_weight_sync_state failed at Ray boundary: RayTaskError(RuntimeError)(RuntimeError('tp_size() failed at Ray boundary: ActorDiedError(RayTaskError(\'__init__\', \'Traceback (most recent call last):\\n File "python/ray/_raylet.pyx", line 1722, in ray._raylet.execute_task\\n File "python/ray/_raylet.pyx", line 1659, in ray._raylet.execute_task.function_executor\\n File "python/ray/_raylet.pyx", line 4342, in ray._raylet.CoreWorker.run_async_func_or_coro_in_event_loop\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 449, in result\\n return self.__get_result()\\n ^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result\\n raise self._exception\\n File "python/ray/_raylet.pyx", line 4329, in async_func\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/async_compat.py", line 52, in wrapper\\n return func(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/function_manager.py", line 693, in actor_method_executor\\n return method(__ray_actor, *args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1139, in __init__\\n super().__init__(*args, **kwargs)\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 603, in __init__\\n self.llm = self._create_engine(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1216, in _create_engine\\n engine = vllm.AsyncLLMEngine.from_engine_args(engine_args, stat_loggers=stat_loggers)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 251, in from_engine_args\\n return cls(\\n ^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 148, in __init__\\n self.engine_core = EngineCoreClient.make_async_mp_client(\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 124, in make_async_mp_client\\n return AsyncMPClient(*client_args)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 835, in __init__\\n super().__init__(\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 490, in __init__\\n with launch_core_engines(vllm_config, executor_class, log_stats) as (\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/contextlib.py", line 144, in __ex
(skyrl_entrypoint pid=2701319) Traceback (most recent call last):
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/fully_async_trainer.py", line 491, in train
(skyrl_entrypoint pid=2701319) await self._train_loop()
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/fully_async_trainer.py", line 545, in _train_loop
(skyrl_entrypoint pid=2701319) self.init_weight_sync_state()
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/trainer.py", line 851, in init_weight_sync_state
(skyrl_entrypoint pid=2701319) raise RuntimeError(
(skyrl_entrypoint pid=2701319) RuntimeError: init_weight_sync_state failed at Ray boundary: RayTaskError(RuntimeError)(RuntimeError('tp_size() failed at Ray boundary: ActorDiedError(RayTaskError(\'__init__\', \'Traceback (most recent call last):\\n File "python/ray/_raylet.pyx", line 1722, in ray._raylet.execute_task\\n File "python/ray/_raylet.pyx", line 1659, in ray._raylet.execute_task.function_executor\\n File "python/ray/_raylet.pyx", line 4342, in ray._raylet.CoreWorker.run_async_func_or_coro_in_event_loop\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 449, in result\\n return self.__get_result()\\n ^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result\\n raise self._exception\\n File "python/ray/_raylet.pyx", line 4329, in async_func\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/async_compat.py", line 52, in wrapper\\n return func(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/function_manager.py", line 693, in actor_method_executor\\n return method(__ray_actor, *args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1139, in __init__\\n super().__init__(*args, **kwargs)\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 603, in __init__\\n self.llm = self._create_engine(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1216, in _create_engine\\n engine = vllm.AsyncLLMEngine.from_engine_args(engine_args, stat_loggers=stat_loggers)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 251, in from_engine_args\\n return cls(\\n ^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 148, in __init__\\n self.engine_core = EngineCoreClient.make_async_mp_client(\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 124, in make_async_mp_client\\n return AsyncMPClient(*client_args)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 835, in __init__\\n super().__init__(\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 490, in __init__\\n with launch_core_engines(vllm_config, executor_class, log_stats) as (\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/contextlib.py", line 144, in __exit__\\n next(self.gen)\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/utils.py",
(RolloutCoordinator pid=2593907, ip=10.128.32.36) [2026-06-07 02:14:09,830 E 2593907 2593949] core_worker_process.cc:842: Failed to establish connection to the metrics exporter agent. Metrics will not be exported. Exporter agent status: RpcError: Running out of retries to initialize the metrics agent. rpc_code: 14
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:09.964 | INFO  | skyrl_train.inference_engines.inference_engine_client_http_endpoint:shutdown_server:203 - Server shut down after 2 seconds
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:09.964 | INFO  | skyrl_train.trainer:_guarded_sync:226 - HTTP endpoint shutdown complete
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:09.968 | INFO  | skyrl_train.trainer:_guarded_async:215 - Generator shutdown complete
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:09.980 | WARNING  | skyrl_train.trainer:_guarded_async:219 - Inference engine teardown error (non-fatal): The actor died because of an error raised in its creation task, ray::AsyncVLLMInferenceEngine.__init__() (pid=2593284, ip=10.128.32.36, actor_id=a9ee7a598f6c230f2b02b9e202000000, repr=<skyrl_train.inference_engines.vllm.vllm_engine.AsyncVLLMInferenceEngine object at 0x400c0f140ec0>)
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 449, in result
(skyrl_entrypoint pid=2701319) return self.__get_result()
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result
(skyrl_entrypoint pid=2701319) raise self._exception
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1139, in __init__
(skyrl_entrypoint pid=2701319) super().__init__(*args, **kwargs)
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 603, in __init__
(skyrl_entrypoint pid=2701319) self.llm = self._create_engine(*args, **kwargs)
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1216, in _create_engine
(skyrl_entrypoint pid=2701319) engine = vllm.AsyncLLMEngine.from_engine_args(engine_args, stat_loggers=stat_loggers)
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 251, in from_engine_args
(skyrl_entrypoint pid=2701319) return cls(
(skyrl_entrypoint pid=2701319) ^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 148, in __init__
(skyrl_entrypoint pid=2701319) self.engine_core = EngineCoreClient.make_async_mp_client(
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 124, in make_async_mp_client
(skyrl_entrypoint pid=2701319) return AsyncMPClient(*client_args)
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 835, in __init__
(skyrl_entrypoint pid=2701319) super().__init__(
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 490, in __init__
(skyrl_entrypoint pid=2701319) with launch_core_engines(vllm_config, executor_class, log_stats) as (
(skyrl_entrypoint pid=2701319) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/contextlib.py", line 144, in __exit__
(skyrl_entrypoint pid=2701319) next(self.gen)
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 936, in launch_core_engines
(skyrl_entrypoint pid=2701319) wait_for_engine_startup(
(skyrl_entrypoint pid=2701319) File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 995, in wait_for_engine_startup
(skyrl_entrypoint pid=2701319) raise RuntimeError(
(skyrl_entrypoint pid=2701319) RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}
(skyrl_entrypoint pid=2701319) 2026-06-07 02:14:09.981 | INFO  | skyrl_train.trainer:_kill_ray_actors:248 - Killing policy_model actors...
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:14:09.965 | INFO | examples.terminal_bench.terminal_bench_generator:shutdown:363 - Shutting down shared QueueOrchestrator...
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:14:09.965 | INFO | examples.terminal_bench.terminal_bench_generator:shutdown:365 - QueueOrchestrator shutdown complete
(RolloutCoordinator pid=2956288, ip=10.128.32.39) 2026-06-07 02:14:09.966 | INFO | examples.terminal_bench.rollout_coordinator:shutdown:258 - [RolloutCoordinator 0] shutdown complete
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py:470: UserWarning: No model update group to destroy
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) warnings.warn("No model update group to destroy")
2026-06-07 02:14:10.027 | ERROR | __main__:main:134 - Training failed: ray::skyrl_entrypoint() (pid=2701319, ip=10.128.32.34)
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/examples/terminal_bench/entrypoints/main_tbench.py", line 105, in skyrl_entrypoint
exp.run()
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/entrypoints/main_base.py", line 483, in run
asyncio.run(trainer.train())
File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/asyncio/runners.py", line 195, in run
return runner.run(main)
^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/asyncio/runners.py", line 118, in run
return self._loop.run_until_complete(task)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/asyncio/base_events.py", line 691, in run_until_complete
return future.result()
^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/fully_async_trainer.py", line 491, in train
await self._train_loop()
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/fully_async_trainer.py", line 545, in _train_loop
self.init_weight_sync_state()
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/trainer.py", line 851, in init_weight_sync_state
raise RuntimeError(
RuntimeError: init_weight_sync_state failed at Ray boundary: RayTaskError(RuntimeError)(RuntimeError('tp_size() failed at Ray boundary: ActorDiedError(RayTaskError(\'__init__\', \'Traceback (most recent call last):\\n File "python/ray/_raylet.pyx", line 1722, in ray._raylet.execute_task\\n File "python/ray/_raylet.pyx", line 1659, in ray._raylet.execute_task.function_executor\\n File "python/ray/_raylet.pyx", line 4342, in ray._raylet.CoreWorker.run_async_func_or_coro_in_event_loop\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 449, in result\\n return self.__get_result()\\n ^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result\\n raise self._exception\\n File "python/ray/_raylet.pyx", line 4329, in async_func\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/async_compat.py", line 52, in wrapper\\n return func(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/function_manager.py", line 693, in actor_method_executor\\n return method(__ray_actor, *args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1139, in __init__\\n super().__init__(*args, **kwargs)\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 603, in __init__\\n self.llm = self._create_engine(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1216, in _create_engine\\n engine = vllm.AsyncLLMEngine.from_engine_args(engine_args, stat_loggers=stat_loggers)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 251, in from_engine_args\\n return cls(\\n ^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 148, in __init__\\n self.engine_core = EngineCoreClient.make_async_mp_client(\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 124, in make_async_mp_client\\n return AsyncMPClient(*client_args)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 835, in __init__\\n super().__init__(\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 490, in __init__\\n with launch_core_engines(vllm_config, executor_class, log_stats) as (\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/contextlib.py", line 144, in __exit__\\n next(self.gen)\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 936, in launch_core_engines\\n wait_for_engine_st
2026-06-07 02:14:10.027 | INFO | __main__:main:137 - Shutting down Ray on head node...
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) /e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py:470: UserWarning: No model update group to destroy [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) warnings.warn("No model update group to destroy") [repeated 27x across cluster]
(skyrl_entrypoint pid=2701319) [fd-monitor] Started monitoring (every 120s)
(skyrl_entrypoint pid=2701319) [fd-monitor] [02:11:44] OK: 50 / 131,072 FDs open (0.0% of soft limit, hard limit: 131,072)
(skyrl_entrypoint pid=2701319) [fd-monitor] [02:11:44] OK: RSS 1.38 GiB | node mem 126.9/858.0 GiB used (14.8%), avail 731.1 GiB
(skyrl_entrypoint pid=2701319) ⚙️ Running in WANDB offline mode
(skyrl_entrypoint pid=2701319) INFO 06-07 02:12:04 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s
(pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:26 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s
(pid=2666657, ip=10.128.32.38) INFO 06-07 02:12:30 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) WARNING 06-07 02:12:31 [arg_utils.py:1256] The global random seed is set to 42. Since VLLM_ENABLE_V1_MULTIPROCESSING is set to False, this may affect the random state of the Python process that launched vLLM.
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:31 [model.py:529] Resolved architecture: Qwen3ForCausalLM
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:31 [model.py:1549] Using max model len 32768
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:31 [arg_utils.py:1469] Using ray runtime env (env vars redacted): {'env_vars': {'NCCL_CUMEM_ENABLE': '***', 'NCCL_DEBUG': '***', 'NCCL_SOCKET_FAMILY': '***', 'NCCL_SOCKET_IFNAME': '***', 'RAY_ADDRESS': '***', 'VLLM_ALLOW_INSECURE_SERIALIZATION': '***', 'VLLM_ALLOW_RUNTIME_LORA_UPDATING': '***', 'VLLM_DISABLE_COMPILE_CACHE': '***', 'WANDB_API_KEY': '***'}}
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:31 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=65536.
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:31 [vllm.py:690] Asynchronous scheduling is enabled.
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) WARNING 06-07 02:12:31 [vllm.py:728] Enforce eager set, overriding optimization level to -O0
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:31 [vllm.py:846] Cudagraph is disabled under eager mode
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) WARNING 06-07 02:12:31 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) WARNING 06-07 02:12:31 [system_utils.py:140] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: In a Ray actor and can only be spawned
(pid=2659764, ip=10.128.32.43) INFO 06-07 02:12:35 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) WARNING 06-07 02:12:34 [arg_utils.py:1256] The global random seed is set to 55. Since VLLM_ENABLE_V1_MULTIPROCESSING is set to False, this may affect the random state of the Python process that launched vLLM. [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) INFO 06-07 02:12:34 [model.py:529] Resolved architecture: Qwen3ForCausalLM [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) INFO 06-07 02:12:34 [model.py:1549] Using max model len 32768 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) INFO 06-07 02:12:34 [arg_utils.py:1469] Using ray runtime env (env vars redacted): {'env_vars': {'NCCL_CUMEM_ENABLE': '***', 'NCCL_DEBUG': '***', 'NCCL_SOCKET_FAMILY': '***', 'NCCL_SOCKET_IFNAME': '***', 'RAY_ADDRESS': '***', 'VLLM_ALLOW_INSECURE_SERIALIZATION': '***', 'VLLM_ALLOW_RUNTIME_LORA_UPDATING': '***', 'VLLM_DISABLE_COMPILE_CACHE': '***', 'WANDB_API_KEY': '***'}} [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) INFO 06-07 02:12:34 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=65536. [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) INFO 06-07 02:12:34 [vllm.py:690] Asynchronous scheduling is enabled. [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) WARNING 06-07 02:12:34 [vllm.py:728] Enforce eager set, overriding optimization level to -O0 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) INFO 06-07 02:12:34 [vllm.py:846] Cudagraph is disabled under eager mode [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) WARNING 06-07 02:12:34 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666658, ip=10.128.32.38) WARNING 06-07 02:12:34 [system_utils.py:140] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: In a Ray actor and can only be spawned [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) INFO 06-07 02:12:39 [core.py:97] Initializing a V1 LLM engine (vdev) with config: model='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', speculative_config=None, tokenizer='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=42, served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'compile_sizes': [], 'compile_ranges_split_points': [65536], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []}
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) INFO 06-07 02:12:39 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) WARNING 06-07 02:12:43 [arg_utils.py:1256] The global random seed is set to 81. Since VLLM_ENABLE_V1_MULTIPROCESSING is set to False, this may affect the random state of the Python process that launched vLLM. [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) INFO 06-07 02:12:43 [model.py:529] Resolved architecture: Qwen3ForCausalLM [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) INFO 06-07 02:12:43 [model.py:1549] Using max model len 32768 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) INFO 06-07 02:12:43 [arg_utils.py:1469] Using ray runtime env (env vars redacted): {'env_vars': {'NCCL_CUMEM_ENABLE': '***', 'NCCL_DEBUG': '***', 'NCCL_SOCKET_FAMILY': '***', 'NCCL_SOCKET_IFNAME': '***', 'RAY_ADDRESS': '***', 'VLLM_ALLOW_INSECURE_SERIALIZATION': '***', 'VLLM_ALLOW_RUNTIME_LORA_UPDATING': '***', 'VLLM_DISABLE_COMPILE_CACHE': '***', 'WANDB_API_KEY': '***'}} [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) INFO 06-07 02:12:43 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=65536. [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) INFO 06-07 02:12:43 [vllm.py:690] Asynchronous scheduling is enabled. [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) WARNING 06-07 02:12:43 [vllm.py:728] Enforce eager set, overriding optimization level to -O0 [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) INFO 06-07 02:12:43 [vllm.py:846] Cudagraph is disabled under eager mode [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) WARNING 06-07 02:12:43 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1 [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=2659897, ip=10.128.32.43) WARNING 06-07 02:12:43 [system_utils.py:140] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: In a Ray actor and can only be spawned [repeated 3x across cluster]
(pid=2623829, ip=10.128.32.46) ⚙️ Running in WANDB offline mode
(AsyncVLLMInferenceEngine pid=2666657, ip=10.128.32.38) (EngineCore_DP0 pid=2666896) INFO 06-07 02:12:44 [core.py:97] Initializing a V1 LLM engine (vdev) with config: model='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', speculative_config=None, tokenizer='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=56, served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'compile_sizes': [], 'compile_ranges_split_points': [65536], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []} [repeated 3x across cluster]
(pid=2972657, ip=10.128.32.44) INFO 06-07 02:12:44 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 39x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) WARNING 06-07 02:12:48 [arg_utils.py:1256] The global random seed is set to 71. Since VLLM_ENABLE_V1_MULTIPROCESSING is set to False, this may affect the random state of the Python process that launched vLLM. [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) INFO 06-07 02:12:48 [model.py:529] Resolved architecture: Qwen3ForCausalLM [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) INFO 06-07 02:12:48 [model.py:1549] Using max model len 32768 [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) INFO 06-07 02:12:48 [arg_utils.py:1469] Using ray runtime env (env vars redacted): {'env_vars': {'NCCL_CUMEM_ENABLE': '***', 'NCCL_DEBUG': '***', 'NCCL_SOCKET_FAMILY': '***', 'NCCL_SOCKET_IFNAME': '***', 'RAY_ADDRESS': '***', 'VLLM_ALLOW_INSECURE_SERIALIZATION': '***', 'VLLM_ALLOW_RUNTIME_LORA_UPDATING': '***', 'VLLM_DISABLE_COMPILE_CACHE': '***', 'WANDB_API_KEY': '***'}} [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) INFO 06-07 02:12:48 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=65536. [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) INFO 06-07 02:12:48 [vllm.py:690] Asynchronous scheduling is enabled. [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) WARNING 06-07 02:12:48 [vllm.py:728] Enforce eager set, overriding optimization level to -O0 [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=3076171, ip=10.128.32.42) INFO 06-07 02:12:48 [vllm.py:846] Cudagraph is disabled under eager mode [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) WARNING 06-07 02:12:48 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1 [repeated 18x across cluster]
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) WARNING 06-07 02:12:48 [system_utils.py:140] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: In a Ray actor and can only be spawned [repeated 18x across cluster]
(AsyncVLLMInferenceEngine pid=2666656, ip=10.128.32.38) (EngineCore_DP0 pid=2666891) INFO 06-07 02:12:50 [worker_base.py:289] Injected <class 'skyrl_train.inference_engines.vllm.vllm_engine.WorkerWrap'> into <class 'vllm.v1.worker.gpu_worker.Worker'> for extended collective_rpc calls ['_apply_fp8_weight_loader_patches', '_is_fp8_model', '_quantize_weights_for_fp8', '_restore_param_subclasses', '_undo_param_subclasses', 'begin_weight_update', 'destroy_weights_update_group', 'end_weight_update', 'init_weight_update_communicator', 'load_weights', 'read_named_weights', 'set_numa_affinity', 'test_rpc']
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) INFO 06-07 02:12:50 [parallel_state.py:1234] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://10.128.32.38:51279 backend=nccl
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) INFO 06-07 02:12:50 [parallel_state.py:1445] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank N/A
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:12:48 [core.py:97] Initializing a V1 LLM engine (vdev) with config: model='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', speculative_config=None, tokenizer='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=66, served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'compile_sizes': [], 'compile_ranges_split_points': [65536], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []}
(AsyncVLLMInferenceEngine pid=2659899, ip=10.128.32.43) (EngineCore_DP0 pid=2660131) INFO 06-07 02:12:51 [core.py:97] Initializing a V1 LLM engine (vdev) with config: model='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', speculative_config=None, tokenizer='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=80, served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'compile_sizes': [], 'compile_ranges_split_points': [65536], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []}
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) INFO 06-07 02:12:52 [gpu_model_runner.py:4125] Starting to load model /e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6...
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) INFO 06-07 02:12:51 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 8x across cluster]
(pid=2623902, ip=10.128.32.46) ⚙️ Running in WANDB offline mode
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) WARNING 06-07 02:12:53 [arg_utils.py:1256] The global random seed is set to 86. Since VLLM_ENABLE_V1_MULTIPROCESSING is set to False, this may affect the random state of the Python process that launched vLLM. [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) INFO 06-07 02:12:53 [model.py:529] Resolved architecture: Qwen3ForCausalLM [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) INFO 06-07 02:12:53 [model.py:1549] Using max model len 32768 [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) INFO 06-07 02:12:53 [arg_utils.py:1469] Using ray runtime env (env vars redacted): {'env_vars': {'NCCL_CUMEM_ENABLE': '***', 'NCCL_DEBUG': '***', 'NCCL_SOCKET_FAMILY': '***', 'NCCL_SOCKET_IFNAME': '***', 'RAY_ADDRESS': '***', 'VLLM_ALLOW_INSECURE_SERIALIZATION': '***', 'VLLM_ALLOW_RUNTIME_LORA_UPDATING': '***', 'VLLM_DISABLE_COMPILE_CACHE': '***', 'WANDB_API_KEY': '***'}} [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) INFO 06-07 02:12:53 [scheduler.py:224] Chunked prefill is enabled with max_num_batched_tokens=65536. [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) INFO 06-07 02:12:53 [vllm.py:690] Asynchronous scheduling is enabled. [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) WARNING 06-07 02:12:53 [vllm.py:728] Enforce eager set, overriding optimization level to -O0 [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) INFO 06-07 02:12:53 [vllm.py:846] Cudagraph is disabled under eager mode [repeated 13x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) WARNING 06-07 02:12:53 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1 [repeated 22x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) WARNING 06-07 02:12:53 [system_utils.py:140] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: In a Ray actor and can only be spawned [repeated 22x across cluster]
(AsyncVLLMInferenceEngine pid=2666524, ip=10.128.32.38) (EngineCore_DP0 pid=2666870) INFO 06-07 02:12:54 [cuda.py:367] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) INFO 06-07 02:12:55 [worker_base.py:289] Injected <class 'skyrl_train.inference_engines.vllm.vllm_engine.WorkerWrap'> into <class 'vllm.v1.worker.gpu_worker.Worker'> for extended collective_rpc calls ['_apply_fp8_weight_loader_patches', '_is_fp8_model', '_quantize_weights_for_fp8', '_restore_param_subclasses', '_undo_param_subclasses', 'begin_weight_update', 'destroy_weights_update_group', 'end_weight_update', 'init_weight_update_communicator', 'load_weights', 'read_named_weights', 'set_numa_affinity', 'test_rpc'] [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) INFO 06-07 02:12:55 [parallel_state.py:1234] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://10.128.32.43:50889 backend=nccl [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) INFO 06-07 02:12:55 [parallel_state.py:1445] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank N/A [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) INFO 06-07 02:12:52 [core.py:97] Initializing a V1 LLM engine (vdev) with config: model='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', speculative_config=None, tokenizer='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=81, served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'compile_sizes': [], 'compile_ranges_split_points': [65536], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []} [repeated 2x across cluster]
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) INFO 06-07 02:12:55 [gpu_model_runner.py:4125] Starting to load model /e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6... [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2593411, ip=10.128.32.36) INFO 06-07 02:12:57 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 12x across cluster]
(pid=2701488) ⚙️ Running in WANDB offline mode [repeated 6x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) WARNING 06-07 02:12:54 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) WARNING 06-07 02:12:54 [system_utils.py:140] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: In a Ray actor and can only be spawned
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) (EngineCore_DP0 pid=2660148) INFO 06-07 02:12:56 [cuda.py:367] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION']. [repeated 7x across cluster]
(AsyncVLLMInferenceEngine pid=2593412, ip=10.128.32.36) (EngineCore_DP0 pid=2593646) INFO 06-07 02:13:00 [worker_base.py:289] Injected <class 'skyrl_train.inference_engines.vllm.vllm_engine.WorkerWrap'> into <class 'vllm.v1.worker.gpu_worker.Worker'> for extended collective_rpc calls ['_apply_fp8_weight_loader_patches', '_is_fp8_model', '_quantize_weights_for_fp8', '_restore_param_subclasses', '_undo_param_subclasses', 'begin_weight_update', 'destroy_weights_update_group', 'end_weight_update', 'init_weight_update_communicator', 'load_weights', 'read_named_weights', 'set_numa_affinity', 'test_rpc'] [repeated 2x across cluster]
(AsyncVLLMInferenceEngine pid=2575228, ip=10.128.32.41) (EngineCore_DP0 pid=2575453) INFO 06-07 02:13:01 [parallel_state.py:1234] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://10.128.32.41:58449 backend=nccl
(AsyncVLLMInferenceEngine pid=2575228, ip=10.128.32.41) (EngineCore_DP0 pid=2575453) INFO 06-07 02:13:01 [parallel_state.py:1445] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank N/A
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:00 [core.py:97] Initializing a V1 LLM engine (vdev) with config: model='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', speculative_config=None, tokenizer='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=75, served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'compile_sizes': [], 'compile_ranges_split_points': [65536], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []} [repeated 36x across cluster]
(AsyncVLLMInferenceEngine pid=2575228, ip=10.128.32.41) (EngineCore_DP0 pid=2575453) INFO 06-07 02:13:02 [gpu_model_runner.py:4125] Starting to load model /e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6...
(AsyncVLLMInferenceEngine pid=2955861, ip=10.128.32.39) INFO 06-07 02:13:01 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 25x across cluster]
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] EngineCore failed to start.
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] Traceback (most recent call last):
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 996, in run_engine_core
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] engine_core = EngineCoreProc(*args, engine_index=dp_rank, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 740, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] super().__init__(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 106, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] self.model_executor = executor_class(vllm_config)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/executor/abstract.py", line 103, in __init__
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] self._init_executor()
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/executor/uniproc_executor.py", line 47, in _init_executor
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] self.driver_worker.init_device()
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/worker/worker_base.py", line 332, in init_device
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] self.worker.init_device() # type: ignore
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/worker/gpu_worker.py", line 232, in init_device
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] init_worker_distributed_environment(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/worker/gpu_worker.py", line 1040, in init_worker_distributed_environment
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] init_distributed_environment(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/distributed/parallel_state.py", line 1256, in init_distributed_environment
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] torch.distributed.init_process_group(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/c10d_logger.py", line 81, in wrapper
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] return func(*args, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/c10d_logger.py", line 95, in wrapper
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] func_return = func(*args, **kwargs)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/distributed_c10d.py", line 1762, in init_process_group
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] store, rank, world_size = next(rendezvous_iterator)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/rendezvous.py", line 230, in _tcp_rendezvous_handler
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] store = _create_c10d_store(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^^^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/distributed/rendezvous.py", line 198, in _create_c10d_store
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] return TCPStore(
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] ^^^^^^^^^
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] torch.distributed.DistNetworkError: The server socket has failed to listen on any local network address. port: 37945, useIpv6: false, code: -98, name: EADDRINUSE, message: address already in use
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] Exception raised from makeWithPort at /pytorch/torch/csrc/distributed/c10d/TCPStoreLibUvBackend.cpp:307 (most recent call first):
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0xb0 (0x4000730cc700 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libc10.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #1: <unknown function> + 0x5f29220 (0x400053269220 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #2: <unknown function> + 0x5f4326c (0x40005328326c in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #3: <unknown function> + 0x5f49074 (0x400053289074 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #4: <unknown function> + 0x5f49138 (0x400053289138 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #5: <unknown function> + 0x5f2ccc4 (0x40005326ccc4 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #6: c10d::TCPStore::TCPStore(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, c10d::TCPStoreOptions const&) + 0x104 (0x400053271564 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_cpu.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #7: <unknown function> + 0x109a094 (0x40004d01a094 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #8: <unknown function> + 0x113236c (0x40004d0b236c in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #9: <unknown function> + 0x5d6d60 (0x40004c556d60 in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #10: <unknown function> + 0x1b7a38 (0xaaaabc307a38 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #11: _PyObject_MakeTpCall + 0x98 (0xaaaabc2b5db8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #12: <unknown function> + 0x169f50 (0xaaaabc2b9f50 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #13: <unknown function> + 0x1682e4 (0xaaaabc2b82e4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #14: <unknown function> + 0x1e0ce8 (0xaaaabc330ce8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #15: <unknown function> + 0x1d7ddc (0xaaaabc327ddc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #16: <unknown function> + 0x646b0c (0x40004c5c6b0c in /e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/torch/lib/libtorch_python.so)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #17: _PyObject_MakeTpCall + 0x98 (0xaaaabc2b5db8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #18: _PyEval_EvalFrameDefault + 0x280c (0xaaaabc3bae54 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #19: <unknown function> + 0x1808c0 (0xaaaabc2d08c0 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #20: <unknown function> + 0x182bf8 (0xaaaabc2d2bf8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #21: <unknown function> + 0x25fd30 (0xaaaabc3afd30 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #22: <unknown function> + 0x1b7d20 (0xaaaabc307d20 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #23: PyObject_Vectorcall + 0x54 (0xaaaabc2b60e4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #24: _PyEval_EvalFrameDefault + 0x280c (0xaaaabc3bae54 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #25: _PyObject_FastCallDictTstate + 0x80 (0xaaaabc2b7f20 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #26: _PyObject_Call_Prepend + 0x140 (0xaaaabc2b81ec in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #27: <unknown function> + 0x1e0d80 (0xaaaabc330d80 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #28: <unknown function> + 0x1d7ddc (0xaaaabc327ddc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #29: _PyObject_MakeTpCall + 0x98 (0xaaaabc2b5db8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #30: _PyEval_EvalFrameDefault + 0x280c (0xaaaabc3bae54 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #31: _PyObject_FastCallDictTstate + 0x10c (0xaaaabc2b7fac in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #32: _PyObject_Call_Prepend + 0x140 (0xaaaabc2b81ec in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #33: <unknown function> + 0x1e0d80 (0xaaaabc330d80 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #34: <unknown function> + 0x1d7ddc (0xaaaabc327ddc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #35: _PyObject_Call + 0x68 (0xaaaabc2b8488 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #36: _PyEval_EvalFrameDefault + 0x52ac (0xaaaabc3bd8f4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #37: PyEval_EvalCode + 0xb4 (0xaaaabc3c2eb4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #38: <unknown function> + 0x2ccdcc (0xaaaabc41cdcc in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #39: <unknown function> + 0x2ccef4 (0xaaaabc41cef4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #40: PyRun_StringFlags + 0x90 (0xaaaabc421050 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #41: PyRun_SimpleStringFlags + 0x58 (0xaaaabc4210f8 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #42: Py_RunMain + 0x2c8 (0xaaaabc449190 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #43: Py_BytesMain + 0x64 (0xaaaabc449fb4 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #44: <unknown function> + 0x27540 (0x40003b5c7540 in /lib64/libc.so.6)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #45: __libc_start_main + 0x98 (0x40003b5c7618 in /lib64/libc.so.6)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006] frame #46: <unknown function> + 0x10e0c0 (0xaaaabc25e0c0 in VLLM::EngineCore)
(AsyncVLLMInferenceEngine pid=2593284, ip=10.128.32.36) (EngineCore_DP0 pid=2593652) ERROR 06-07 02:13:02 [core.py:1006]
(AsyncVLLMInferenceEngine pid=2593411, ip=10.128.32.36) (EngineCore_DP0 pid=2593642) INFO 06-07 02:13:02 [gpu_model_runner.py:4125] Starting to load model /e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6...
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) INFO 06-07 02:13:03 [cuda.py:367] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION']. [repeated 12x across cluster]
(AsyncVLLMInferenceEngine pid=2972655, ip=10.128.32.44) (EngineCore_DP0 pid=2972907) INFO 06-07 02:13:05 [worker_base.py:289] Injected <class 'skyrl_train.inference_engines.vllm.vllm_engine.WorkerWrap'> into <class 'vllm.v1.worker.gpu_worker.Worker'> for extended collective_rpc calls ['_apply_fp8_weight_loader_patches', '_is_fp8_model', '_quantize_weights_for_fp8', '_restore_param_subclasses', '_undo_param_subclasses', 'begin_weight_update', 'destroy_weights_update_group', 'end_weight_update', 'init_weight_update_communicator', 'load_weights', 'read_named_weights', 'set_numa_affinity', 'test_rpc'] [repeated 33x across cluster]
(AsyncVLLMInferenceEngine pid=984310, ip=10.128.32.37) (EngineCore_DP0 pid=984419) INFO 06-07 02:13:06 [parallel_state.py:1234] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://10.128.32.37:34951 backend=nccl [repeated 35x across cluster]
(AsyncVLLMInferenceEngine pid=984310, ip=10.128.32.37) (EngineCore_DP0 pid=984419) INFO 06-07 02:13:06 [parallel_state.py:1445] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank N/A [repeated 34x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) INFO 06-07 02:13:03 [core.py:97] Initializing a V1 LLM engine (vdev) with config: model='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', speculative_config=None, tokenizer='/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=87, served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'compile_sizes': [], 'compile_ranges_split_points': [65536], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []} [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) INFO 06-07 02:13:02 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 3x across cluster]
(AsyncVLLMInferenceEngine pid=984310, ip=10.128.32.37) (EngineCore_DP0 pid=984419) INFO 06-07 02:13:06 [gpu_model_runner.py:4125] Starting to load model /e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6... [repeated 33x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:08 [default_loader.py:293] Loading weights took 13.01 seconds
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:08 [gpu_model_runner.py:4222] Model loading took 15.27 GiB memory and 13.474666 seconds
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:11 [gpu_worker.py:373] Available KV cache memory: 49.32 GiB
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:11 [kv_cache_utils.py:1307] GPU KV cache size: 359,104 tokens
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:11 [kv_cache_utils.py:1312] Maximum concurrency for 32,768 tokens per request: 10.96x
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113) INFO 06-07 02:13:08 [cuda.py:367] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION']. [repeated 27x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113) INFO 06-07 02:13:07 [worker_base.py:289] Injected <class 'skyrl_train.inference_engines.vllm.vllm_engine.WorkerWrap'> into <class 'vllm.v1.worker.gpu_worker.Worker'> for extended collective_rpc calls ['_apply_fp8_weight_loader_patches', '_is_fp8_model', '_quantize_weights_for_fp8', '_restore_param_subclasses', '_undo_param_subclasses', 'begin_weight_update', 'destroy_weights_update_group', 'end_weight_update', 'init_weight_update_communicator', 'load_weights', 'read_named_weights', 'set_numa_affinity', 'test_rpc'] [repeated 5x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:11 [kernel_warmup.py:44] Skipping FlashInfer autotune because it is disabled.
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113) INFO 06-07 02:13:07 [parallel_state.py:1234] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://10.128.32.39:38597 backend=nccl [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113) INFO 06-07 02:13:07 [parallel_state.py:1445] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank N/A [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:11 [core.py:278] init engine (profile, create kv cache, warmup model) took 3.23 seconds
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) WARNING 06-07 02:13:12 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:12 [vllm.py:690] Asynchronous scheduling is enabled.
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) WARNING 06-07 02:13:12 [vllm.py:735] Inductor compilation was disabled by user settings, optimizations settings that are only active during inductor compilation will be ignored.
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) (EngineCore_DP0 pid=2660123) INFO 06-07 02:13:12 [vllm.py:846] Cudagraph is disabled under eager mode
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) NCCL version 2.27.7+cuda13.0
(AsyncVLLMInferenceEngine pid=2659764, ip=10.128.32.43) WARNING 06-07 02:13:12 [model.py:1350] Default vLLM sampling parameters have been overridden by the model's `generation_config.json`: `{'temperature': 0.6, 'top_k': 20, 'top_p': 0.95}`. If this is not intended, please relaunch vLLM instance with `--generation-config vllm`.
(AsyncVLLMInferenceEngine pid=2955732, ip=10.128.32.39) (EngineCore_DP0 pid=2956113) INFO 06-07 02:13:08 [gpu_model_runner.py:4125] Starting to load model /e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6... [repeated 4x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:12 [default_loader.py:293] Loading weights took 7.01 seconds [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:12 [gpu_model_runner.py:4222] Model loading took 15.27 GiB memory and 7.348132 seconds [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:15 [gpu_worker.py:373] Available KV cache memory: 49.32 GiB [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:15 [kv_cache_utils.py:1307] GPU KV cache size: 359,104 tokens [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:15 [kv_cache_utils.py:1312] Maximum concurrency for 32,768 tokens per request: 10.96x [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) (EngineCore_DP0 pid=2575477) INFO 06-07 02:13:15 [kernel_warmup.py:44] Skipping FlashInfer autotune because it is disabled. [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:15 [core.py:278] init engine (profile, create kv cache, warmup model) took 2.79 seconds [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) WARNING 06-07 02:13:16 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1 [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:16 [vllm.py:690] Asynchronous scheduling is enabled. [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) WARNING 06-07 02:13:16 [vllm.py:735] Inductor compilation was disabled by user settings, optimizations settings that are only active during inductor compilation will be ignored. [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575227, ip=10.128.32.41) (EngineCore_DP0 pid=2575458) INFO 06-07 02:13:16 [vllm.py:846] Cudagraph is disabled under eager mode [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2575229, ip=10.128.32.41) WARNING 06-07 02:13:16 [model.py:1350] Default vLLM sampling parameters have been overridden by the model's `generation_config.json`: `{'temperature': 0.6, 'top_k': 20, 'top_p': 0.95}`. If this is not intended, please relaunch vLLM instance with `--generation-config vllm`. [repeated 11x across cluster]
(AsyncVLLMInferenceEngine pid=2972657, ip=10.128.32.44) (EngineCore_DP0 pid=2972891) INFO 06-07 02:13:20 [default_loader.py:293] Loading weights took 13.29 seconds [repeated 19x across cluster]
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) INFO 06-07 02:13:18 [gpu_model_runner.py:4222] Model loading took 15.27 GiB memory and 14.779174 seconds [repeated 15x across cluster]
(RolloutCoordinator pid=2956288, ip=10.128.32.39) [fd-monitor] Started monitoring (every 120s)
(RolloutCoordinator pid=2956288, ip=10.128.32.39) [fd-monitor] [02:13:20] OK: 46 / 131,072 FDs open (0.0% of soft limit, hard limit: 131,072)
(RolloutCoordinator pid=2956288, ip=10.128.32.39) [fd-monitor] [02:13:20] OK: RSS 0.78 GiB | node mem 203.7/858.0 GiB used (23.7%), avail 654.3 GiB
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) INFO 06-07 02:13:20 [gpu_worker.py:373] Available KV cache memory: 49.32 GiB [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) INFO 06-07 02:13:20 [kv_cache_utils.py:1307] GPU KV cache size: 359,104 tokens [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) INFO 06-07 02:13:20 [kv_cache_utils.py:1312] Maximum concurrency for 32,768 tokens per request: 10.96x [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) INFO 06-07 02:13:20 [kernel_warmup.py:44] Skipping FlashInfer autotune because it is disabled. [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) INFO 06-07 02:13:20 [core.py:278] init engine (profile, create kv cache, warmup model) took 2.81 seconds [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2793256, ip=10.128.32.45) (EngineCore_DP0 pid=2793492) WARNING 06-07 02:13:21 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1 [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) (EngineCore_DP0 pid=2558151) INFO 06-07 02:13:21 [vllm.py:690] Asynchronous scheduling is enabled. [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) (EngineCore_DP0 pid=2558151) WARNING 06-07 02:13:21 [vllm.py:735] Inductor compilation was disabled by user settings, optimizations settings that are only active during inductor compilation will be ignored. [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2557897, ip=10.128.32.48) (EngineCore_DP0 pid=2558151) INFO 06-07 02:13:21 [vllm.py:846] Cudagraph is disabled under eager mode [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2793126, ip=10.128.32.45) WARNING 06-07 02:13:21 [model.py:1350] Default vLLM sampling parameters have been overridden by the model's `generation_config.json`: `{'temperature': 0.6, 'top_k': 20, 'top_p': 0.95}`. If this is not intended, please relaunch vLLM instance with `--generation-config vllm`. [repeated 15x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) INFO 06-07 02:13:22 [default_loader.py:293] Loading weights took 13.19 seconds [repeated 16x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) INFO 06-07 02:13:22 [gpu_model_runner.py:4222] Model loading took 15.27 GiB memory and 14.293649 seconds [repeated 20x across cluster]
(RolloutCoordinator pid=984614, ip=10.128.32.37) [fd-monitor] Started monitoring (every 120s)
(RolloutCoordinator pid=984614, ip=10.128.32.37) [fd-monitor] [02:13:28] OK: 46 / 131,072 FDs open (0.0% of soft limit, hard limit: 131,072)
(RolloutCoordinator pid=984614, ip=10.128.32.37) [fd-monitor] [02:13:28] OK: RSS 0.78 GiB | node mem 389.3/858.0 GiB used (45.4%), avail 468.6 GiB
(AsyncVLLMInferenceEngine pid=2955861, ip=10.128.32.39) (EngineCore_DP0 pid=2956093) INFO 06-07 02:13:25 [gpu_worker.py:373] Available KV cache memory: 49.32 GiB [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955861, ip=10.128.32.39) (EngineCore_DP0 pid=2956093) INFO 06-07 02:13:25 [kv_cache_utils.py:1307] GPU KV cache size: 359,104 tokens [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955861, ip=10.128.32.39) (EngineCore_DP0 pid=2956093) INFO 06-07 02:13:25 [kv_cache_utils.py:1312] Maximum concurrency for 32,768 tokens per request: 10.96x [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) INFO 06-07 02:13:25 [kernel_warmup.py:44] Skipping FlashInfer autotune because it is disabled. [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) INFO 06-07 02:13:25 [core.py:278] init engine (profile, create kv cache, warmup model) took 2.80 seconds [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) WARNING 06-07 02:13:25 [serial_utils.py:57] Allowing insecure serialization using pickle due to VLLM_ALLOW_INSECURE_SERIALIZATION=1 [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) INFO 06-07 02:13:25 [vllm.py:690] Asynchronous scheduling is enabled. [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) WARNING 06-07 02:13:25 [vllm.py:735] Inductor compilation was disabled by user settings, optimizations settings that are only active during inductor compilation will be ignored. [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) (EngineCore_DP0 pid=2956097) INFO 06-07 02:13:25 [vllm.py:846] Cudagraph is disabled under eager mode [repeated 20x across cluster]
(AsyncVLLMInferenceEngine pid=2955859, ip=10.128.32.39) WARNING 06-07 02:13:26 [model.py:1350] Default vLLM sampling parameters have been overridden by the model's `generation_config.json`: `{'temperature': 0.6, 'top_k': 20, 'top_p': 0.95}`. If this is not intended, please relaunch vLLM instance with `--generation-config vllm`. [repeated 20x across cluster]
(RolloutCoordinator pid=2660382, ip=10.128.32.43) [fd-monitor] Started monitoring (every 120s)
(RolloutCoordinator pid=2660382, ip=10.128.32.43) [fd-monitor] [02:13:36] OK: 46 / 131,072 FDs open (0.0% of soft limit, hard limit: 131,072)
(RolloutCoordinator pid=2660382, ip=10.128.32.43) [fd-monitor] [02:13:36] OK: RSS 0.78 GiB | node mem 390.8/858.0 GiB used (45.5%), avail 467.2 GiB
(skyrl_entrypoint pid=2701319) [fd-monitor] [02:13:44] OK: 101 / 131,072 FDs open (0.1% of soft limit, hard limit: 131,072)
(skyrl_entrypoint pid=2701319) [fd-monitor] [02:13:44] OK: RSS 1.64 GiB | node mem 213.7/858.0 GiB used (24.9%), avail 644.2 GiB
(RolloutCoordinator pid=2593907, ip=10.128.32.36) [fd-monitor] Started monitoring (every 120s)
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank-0]: Loading model from /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints/global_step_28/policy/model_world_size_8_rank_0.pt
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank-0]: Loading extra_state from /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints/global_step_28/policy/extra_state_world_size_8_rank_0.pt
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank-0]: Loading optim from /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints/global_step_28/policy/optim_world_size_8_rank_0.pt
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank-0]: Successfully loaded model state dict
(RolloutCoordinator pid=2593907, ip=10.128.32.36) [fd-monitor] [02:13:45] OK: 46 / 131,072 FDs open (0.0% of soft limit, hard limit: 131,072)
(RolloutCoordinator pid=2593907, ip=10.128.32.36) [fd-monitor] [02:13:45] OK: RSS 0.78 GiB | node mem 324.8/858.0 GiB used (37.9%), avail 533.2 GiB
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank-0]: Successfully loaded optimizer state
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank-0]: Successfully loaded scheduler state
(FSDPPolicyWorkerBase pid=2623829, ip=10.128.32.46) [rank-0]: Checkpoint loaded successfully from /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints/global_step_28/policy
(FSDPPolicyWorkerBase pid=2701489) INFO 06-07 02:14:06 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s
(AsyncVLLMInferenceEngine pid=2659898, ip=10.128.32.43) ERROR 06-07 02:14:10 [core_client.py:616] Engine core proc EngineCore_DP0 died unexpectedly, shutting down client.
(FSDPPolicyWorkerBase pid=2623904, ip=10.128.32.46) INFO 06-07 02:14:07 [pynccl.py:178] pynccl trace buffer enabled: size=10000 dump_dir=/e/data1/datasets/playground/ot-baf/experiments/_nccl_dumps flush_interval=0 s [repeated 7x across cluster]
Error executing job with overrides: ['+terminal_bench_config=terminal_bench', 'trainer.strategy=fsdp2', 'trainer.algorithm.advantage_estimator=rloo_n', 'trainer.algorithm.use_kl_loss=false', 'trainer.algorithm.kl_loss_coef=0.0', 'trainer.algorithm.eps_clip_low=0.2', 'trainer.algorithm.eps_clip_high=0.05', 'trainer.algorithm.loss_reduction=sequence_mean', 'trainer.epochs=2', 'trainer.max_steps=80', 'trainer.update_epochs_per_batch=1', 'trainer.train_batch_size=64', 'trainer.policy_mini_batch_size=64', 'trainer.eval_batch_size=64', 'trainer.micro_forward_batch_size_per_gpu=4', 'trainer.micro_train_batch_size_per_gpu=1', 'trainer.max_prompt_length=999999', 'trainer.eval_interval=999999', 'trainer.eval_before_train=false', 'trainer.ckpt_interval=2', 'trainer.resume_mode=latest', 'trainer.hf_save_interval=5', '++trainer.hf_hub_repo_id=laion/ablation-pymethods2test-seqmean-arm0', '++trainer.hf_hub_private=false', '++trainer.hf_hub_revision=main', '++trainer.enable_db_registration=false', 'trainer.project_name=OpenThoughts-Agent', 'trainer.log_level=INFO', 'trainer.tracker_commit_each_step=true', 'trainer.logger=console', 'trainer.run_name=ablation-pymethods2test-seqmean-arm0', 'trainer.ckpt_path=/e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/checkpoints', 'trainer.export_path=/e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/exports', 'trainer.policy.optimizer_config.lr=8e-6', 'trainer.policy.optimizer_config.weight_decay=0.0', 'trainer.policy.optimizer_config.adam_betas=[0.9,0.999]', 'trainer.policy.optimizer_config.max_grad_norm=0.9', 'trainer.policy.fsdp_config.cpu_offload=false', 'trainer.policy.fsdp_config.reshard_after_forward=true', 'trainer.policy.fsdp_config.fsdp_size=4', 'trainer.policy.model.path=/e/home/jusers/feuer1/jupiter/.cache/huggingface/hub/models--laion--GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink/snapshots/0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', 'trainer.ref.fsdp_config.cpu_offload=false', 'trainer.ref.fsdp_config.reshard_after_forward=true', 'trainer.ref.fsdp_config.fsdp_size=4', 'trainer.placement.colocate_all=false', 'trainer.placement.policy_num_nodes=2', 'trainer.placement.ref_num_nodes=2', 'trainer.placement.policy_num_gpus_per_node=4', 'trainer.placement.ref_num_gpus_per_node=4', 'trainer.fully_async.max_staleness_steps=16', 'trainer.fully_async.num_parallel_generation_workers=338', 'generator.backend=vllm', 'generator.timeout_multiplier=1.0', 'generator.model_dtype=bfloat16', 'generator.inference_engine_tensor_parallel_size=1', 'generator.num_inference_engines=48', 'generator.n_samples_per_prompt=8', 'generator.eval_n_samples_per_prompt=8', 'generator.gpu_memory_utilization=0.75', 'generator.max_num_seqs=24', 'generator.max_num_batched_tokens=65536', 'generator.enable_prefix_caching=true', 'generator.enable_chunked_prefill=true', 'generator.run_engines_locally=true', 'generator.weight_sync_backend=nccl', 'generator.async_engine=true', 'generator.batched=false', 'generator.enable_http_endpoint=true', 'generator.enable_ray_prometheus_stats=false', 'generator.vllm_stats_interval=1', 'generator.append_eos_token_after_stop_str_in_multi_turn=true', 'generator.max_turns=999999', 'generator.sampling_params.max_generate_length=4096', 'generator.sampling_params.temperature=0.7', 'generator.sampling_params.top_p=0.95', 'generator.sampling_params.top_k=20', '++generator.engine_init_kwargs.max_model_len=32768', '++generator.engine_init_kwargs.custom_chat_template_chat_completion_path=chat_templates/qwen3_thinking_acc.jinja2', '++generator.engine_init_kwargs.served_model_name=0e3bff0c4e51f6b9ec0713b98b9eec36efb91cc6', 'data.train_data=["/e/scratch/jureap59/feuer1/tasks/exp_rpt_pymethods2test-large"]', 'data.val_data=[]', '+terminal_bench_config.trials_dir=/e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/trace_jobs', '+terminal_bench_config.harbor.name=terminus-2', '+terminal_bench_config.harb
Traceback (most recent call last):
File "<frozen runpy>", line 198, in _run_module_as_main
File "<frozen runpy>", line 88, in _run_code
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/examples/terminal_bench/entrypoints/main_tbench.py", line 142, in <module>
main()
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/main.py", line 94, in decorated_main
_run_hydra(
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra
_run_app(
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/_internal/utils.py", line 457, in _run_app
run_and_report(
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/_internal/utils.py", line 223, in run_and_report
raise ex
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/_internal/utils.py", line 220, in run_and_report
return func()
^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/_internal/utils.py", line 458, in <lambda>
lambda: hydra.run(
^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/_internal/hydra.py", line 132, in run
_ = ret.return_value
^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/core/utils.py", line 260, in return_value
raise self._return_value
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/hydra/core/utils.py", line 186, in run_job
ret.return_value = task_function(task_cfg)
^^^^^^^^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/examples/terminal_bench/entrypoints/main_tbench.py", line 132, in main
ray.get(skyrl_entrypoint.remote(cfg))
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/auto_init_hook.py", line 22, in auto_init_wrapper
return fn(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/client_mode_hook.py", line 104, in wrapper
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/worker.py", line 2967, in get
values, debugger_breakpoint = worker.get_objects(
^^^^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/worker.py", line 1015, in get_objects
raise value.as_instanceof_cause()
ray.exceptions.RayTaskError(RuntimeError): ray::skyrl_entrypoint() (pid=2701319, ip=10.128.32.34)
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/examples/terminal_bench/entrypoints/main_tbench.py", line 105, in skyrl_entrypoint
exp.run()
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/entrypoints/main_base.py", line 483, in run
asyncio.run(trainer.train())
File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/asyncio/runners.py", line 195, in run
return runner.run(main)
^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/asyncio/runners.py", line 118, in run
return self._loop.run_until_complete(task)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/asyncio/base_events.py", line 691, in run_until_complete
return future.result()
^^^^^^^^^^^^^^^
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/fully_async_trainer.py", line 491, in train
await self._train_loop()
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/fully_async_trainer.py", line 545, in _train_loop
self.init_weight_sync_state()
File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/trainer.py", line 851, in init_weight_sync_state
raise RuntimeError(
RuntimeError: init_weight_sync_state failed at Ray boundary: RayTaskError(RuntimeError)(RuntimeError('tp_size() failed at Ray boundary: ActorDiedError(RayTaskError(\'__init__\', \'Traceback (most recent call last):\\n File "python/ray/_raylet.pyx", line 1722, in ray._raylet.execute_task\\n File "python/ray/_raylet.pyx", line 1659, in ray._raylet.execute_task.function_executor\\n File "python/ray/_raylet.pyx", line 4342, in ray._raylet.CoreWorker.run_async_func_or_coro_in_event_loop\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 449, in result\\n return self.__get_result()\\n ^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result\\n raise self._exception\\n File "python/ray/_raylet.pyx", line 4329, in async_func\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/async_compat.py", line 52, in wrapper\\n return func(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/_private/function_manager.py", line 693, in actor_method_executor\\n return method(__ray_actor, *args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1139, in __init__\\n super().__init__(*args, **kwargs)\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 603, in __init__\\n self.llm = self._create_engine(*args, **kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span\\n return method(self, *_args, **_kwargs)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/SkyRL/skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py", line 1216, in _create_engine\\n engine = vllm.AsyncLLMEngine.from_engine_args(engine_args, stat_loggers=stat_loggers)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 251, in from_engine_args\\n return cls(\\n ^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 148, in __init__\\n self.engine_core = EngineCoreClient.make_async_mp_client(\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 124, in make_async_mp_client\\n return AsyncMPClient(*client_args)\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 835, in __init__\\n super().__init__(\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 490, in __init__\\n with launch_core_engines(vllm_config, executor_class, log_stats) as (\\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n File "/e/scratch/jureap59/feuer1/miniforge3/lib/python3.12/contextlib.py", line 144, in __exit__\\n next(self.gen)\\n File "/e/scratch/jureap59/feuer1/OpenThoughts-Agent/envs/rl/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 936, in launch_core_engines\\n wait_for_engine_st
Stopping Ray cluster...
Ray cluster stopped
[RLJobRunner] Crash detected (exit!=0) — preserving Ray logs to /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0/ray_logs BEFORE trace upload (so a wall-clock kill can't lose crash evidence)...
Collecting Ray logs from worker jpbo-041-36...
Collecting Ray logs from worker jpbo-041-37...
Collecting Ray logs from worker jpbo-041-38...
Collecting Ray logs from worker jpbo-041-39...
Collecting Ray logs from worker jpbo-041-40...
Collecting Ray logs from worker jpbo-041-41...
Collecting Ray logs from worker jpbo-041-42...
Collecting Ray logs from worker jpbo-041-43...
Collecting Ray logs from worker jpbo-041-44...
Collecting Ray logs from worker jpbo-041-45...
Collecting Ray logs from worker jpbo-041-46...
Collecting Ray logs from worker jpbo-041-47...
Collecting Ray logs from worker jpbo-041-48...
[RLJobRunner] Crash-time Ray log preservation complete.
[RLJobRunner] Launching trace upload (training exit code: 1):
repo_id: DCAgent/ablation-pymethods2test-seqmean-arm0
job_dir: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ablation-pymethods2test-seqmean-arm0
episodes: last
log: /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/logs/ablation-pymethods2test-seqmean-arm0_trace_upload.log
[RLJobRunner] Waiting for trace upload to complete...
[RLJobRunner] Trace upload failed with exit code 1.
Preserving Ray logs to /e/data1/datasets/playground/ot-baf/ablation-pymethods2test-seqmean-arm0/ray_logs/
Collecting Ray logs from worker jpbo-041-36...
Collecting Ray logs from worker jpbo-041-37...
Collecting Ray logs from worker jpbo-041-38...
Collecting Ray logs from worker jpbo-041-39...
Collecting Ray logs from worker jpbo-041-40...
Collecting Ray logs from worker jpbo-041-41...
Collecting Ray logs from worker jpbo-041-42...
Collecting Ray logs from worker jpbo-041-43...
Collecting Ray logs from worker jpbo-041-44...
Collecting Ray logs from worker jpbo-041-45...
Collecting Ray logs from worker jpbo-041-46...
Collecting Ray logs from worker jpbo-041-47...
Collecting Ray logs from worker jpbo-041-48...
Ray log preservation complete
Reference in New Issue Copy Permalink