初始化项目，由ModelHub XC社区提供模型

Model: promotion/qwen3-8b-htmnpo-skywork-s42
Source: Original Platform

.gitattributes

@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,25 @@
+---
+library_name: transformers
+base_model: Qwen/Qwen3-8B
+tags:
+- ronpo
+- mnpo
+- qwen3
+- preference-optimization
+---
+
+# qwen3-8b-htmnpo-skywork-s42
+
+Research checkpoint for the RONPO AAAI revision experiments.
+
+- Method: HT-MNPO skywork
+- Base model: `Qwen/Qwen3-8B` with non-thinking generation protocol
+- Seed: 42
+- Local source at upload time: `/NHNHOME/WORKSPACE/26msit001_A/BASE/aipr_lab_sjkim_eval/revision_qwen3_8b/full_iter1/train/htmnpo_skywork_s42_nhn-b200-q3-seq-07101430-htmnpo`
+- Uploaded at UTC: 2026-07-11T06:32:58Z
+- Run status metadata: `not found`
+
+Qwen3-8B HT-MNPO single-oracle baseline.
+
+Intended use: reproducibility and evaluation for the RONPO paper. This
+checkpoint is not intended as a production assistant.

all_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 1.0,
+    "total_flos": 0.0,
+    "train_loss": 0.5071899907229698,
+    "train_runtime": 3280.3544,
+    "train_samples": 18640,
+    "train_samples_per_second": 5.682,
+    "train_steps_per_second": 0.045
+}

chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

config.json

@@ -0,0 +1,71 @@
+{
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 12288,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 36,
+  "model_type": "qwen3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "transformers_version": "5.13.0",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

config.yaml

@@ -0,0 +1,52 @@
+model_name_or_path: /NHNHOME/WORKSPACE/26msit001_A/mnpo/revision_qwen3_8b/full_iter1/cache/huggingface/hub/models--Qwen--Qwen3-8B/snapshots/b968826d9c46dd6066d109eabc6255188de91218
+torch_dtype: null
+attn_implementation: eager
+dataset_mixer:
+  /NHNHOME/WORKSPACE/26msit001_A/mnpo/revision_qwen3_8b/full_iter1/precomputed/ht_skywork: 1.0
+dataset_splits:
+  - train
+  - test
+preprocessing_num_workers: 12
+bf16: true
+loss_type: ht_mnpo
+eta: 0.0075
+ratio: 0.3333
+max_history_t: 1
+history_weights: [1.0]
+ht_target_column: ht_target
+ht_target_scale: 1.0
+beta: 10
+learning_rate: 5.0e-7
+lr_scheduler_type: cosine
+warmup_ratio: 0.1
+optim: adamw_torch
+weight_decay: 0.0
+seed: 42
+gradient_accumulation_steps: 16
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+num_train_epochs: 1
+
+per_device_train_batch_size: 1
+per_device_eval_batch_size: 1
+max_length: 2048
+max_prompt_length: 1800
+do_eval: true
+eval_strategy: "steps"
+eval_steps: 500
+logging_steps: 10
+log_level: info
+generate_during_eval: false
+metric_for_best_model: eval_loss
+load_best_model_at_end: false
+save_strategy: "steps"
+save_steps: 1000
+save_total_limit: 1
+save_only_model: true
+save_safetensors: true
+push_to_hub: false
+report_to:
+  - wandb
+output_dir: /NHNHOME/WORKSPACE/26msit001_A/mnpo/revision_qwen3_8b/full_iter1/train/htmnpo_skywork_s42_nhn-b200-q3-seq-07101430-htmnpo
+run_name: rev-q3-htmnpo-skywork-s42-nhn-b200-q3-seq-07101430-htmnpo

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "5.13.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba880fa98ab3af3ac68bb720f6294f057ce623041b0d0d0b8b779c224fbd96f5
+size 16381517208

run_status.json

@@ -0,0 +1,14 @@
+{
+  "method": "ht-mnpo",
+  "loss_type": "ht_mnpo",
+  "objective": "skywork",
+  "eta": 0.0075,
+  "seed": 42,
+  "status": "completed",
+  "returncode": 0,
+  "completed_at": "2026-07-11T06:30:45Z",
+  "dataset": "/NHNHOME/WORKSPACE/26msit001_A/mnpo/revision_qwen3_8b/full_iter1/precomputed/ht_skywork",
+  "output_dir": "/NHNHOME/WORKSPACE/26msit001_A/mnpo/revision_qwen3_8b/full_iter1/train/htmnpo_skywork_s42_nhn-b200-q3-seq-07101430-htmnpo",
+  "train_metrics_exists": true,
+  "trainer_state_exists": true
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
+size 11422650

tokenizer_config.json

@@ -0,0 +1,30 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 2048,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 1.0,
+    "total_flos": 0.0,
+    "train_loss": 0.5071899907229698,
+    "train_runtime": 3280.3544,
+    "train_samples": 18640,
+    "train_samples_per_second": 5.682,
+    "train_steps_per_second": 0.045
+}

trainer_state.json

@@ -0,0 +1,344 @@
+{
+  "best_global_step": 146,
+  "best_metric": 0.09535243362188339,
+  "best_model_checkpoint": "/NHNHOME/WORKSPACE/26msit001_A/mnpo/revision_qwen3_8b/full_iter1/train/htmnpo_skywork_s42_nhn-b200-q3-seq-07101430-htmnpo/checkpoint-146",
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 146,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.06866952789699571,
+      "grad_norm": 658.2609913586693,
+      "ht_mnpo/logit": -0.07977066189050674,
+      "ht_mnpo/residual": -0.08727066218852997,
+      "ht_mnpo/residual_abs": 0.5136820673942566,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 3e-07,
+      "logits/chosen": -0.81640625,
+      "logits/rejected": -0.81640625,
+      "logps/chosen": -15.9375,
+      "logps/rejected": -15.875,
+      "loss": 1.032188057899475,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": -154.9237518310547,
+      "rewards/margins": -0.7977066040039062,
+      "rewards/rejected": -154.1260528564453,
+      "step": 10
+    },
+    {
+      "epoch": 0.13733905579399142,
+      "grad_norm": 486.85604769071176,
+      "ht_mnpo/logit": 0.07697143405675888,
+      "ht_mnpo/residual": 0.06947144120931625,
+      "ht_mnpo/residual_abs": 0.46500977873802185,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 4.988506452457066e-07,
+      "logits/chosen": -0.80859375,
+      "logits/rejected": -0.83984375,
+      "logps/chosen": -15.875,
+      "logps/rejected": -15.875,
+      "loss": 0.5586695075035095,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": -153.60064697265625,
+      "rewards/margins": 0.76971435546875,
+      "rewards/rejected": -154.370361328125,
+      "step": 20
+    },
+    {
+      "epoch": 0.20600858369098712,
+      "grad_norm": 279.6489736984974,
+      "ht_mnpo/logit": 0.08701781928539276,
+      "ht_mnpo/residual": 0.07951782643795013,
+      "ht_mnpo/residual_abs": 0.48369044065475464,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 4.860414659112948e-07,
+      "logits/chosen": -0.8359375,
+      "logits/rejected": -0.8359375,
+      "logps/chosen": -15.75,
+      "logps/rejected": -15.75,
+      "loss": 0.6932045221328735,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": -152.07644653320312,
+      "rewards/margins": 0.87017822265625,
+      "rewards/rejected": -152.94662475585938,
+      "step": 30
+    },
+    {
+      "epoch": 0.27467811158798283,
+      "grad_norm": 405.5407274927664,
+      "ht_mnpo/logit": 0.11707458645105362,
+      "ht_mnpo/residual": 0.1095745787024498,
+      "ht_mnpo/residual_abs": 0.40353816747665405,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 4.597220474379125e-07,
+      "logits/chosen": -0.82421875,
+      "logits/rejected": -0.82421875,
+      "logps/chosen": -15.5,
+      "logps/rejected": -15.5,
+      "loss": 0.6164440512657166,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -149.45849609375,
+      "rewards/margins": 1.170745849609375,
+      "rewards/rejected": -150.62924194335938,
+      "step": 40
+    },
+    {
+      "epoch": 0.34334763948497854,
+      "grad_norm": 173.47778729238564,
+      "ht_mnpo/logit": 0.09282531589269638,
+      "ht_mnpo/residual": 0.08532532304525375,
+      "ht_mnpo/residual_abs": 0.3633372485637665,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 4.2139882699413613e-07,
+      "logits/chosen": -0.89453125,
+      "logits/rejected": -0.921875,
+      "logps/chosen": -14.9375,
+      "logps/rejected": -14.9375,
+      "loss": 0.38052481412887573,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -144.1920166015625,
+      "rewards/margins": 0.928253173828125,
+      "rewards/rejected": -145.12026977539062,
+      "step": 50
+    },
+    {
+      "epoch": 0.41201716738197425,
+      "grad_norm": 220.4191286143336,
+      "ht_mnpo/logit": 0.15768662095069885,
+      "ht_mnpo/residual": 0.15018662810325623,
+      "ht_mnpo/residual_abs": 0.4016141891479492,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 3.7326529999303633e-07,
+      "logits/chosen": -0.89453125,
+      "logits/rejected": -0.91796875,
+      "logps/chosen": -14.5625,
+      "logps/rejected": -14.625,
+      "loss": 0.8031173944473267,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": -140.22152709960938,
+      "rewards/margins": 1.5768661499023438,
+      "rewards/rejected": -141.7983856201172,
+      "step": 60
+    },
+    {
+      "epoch": 0.48068669527896996,
+      "grad_norm": 226.8002386346793,
+      "ht_mnpo/logit": 0.15667113661766052,
+      "ht_mnpo/residual": 0.1491711437702179,
+      "ht_mnpo/residual_abs": 0.4558327794075012,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 3.1807647161082797e-07,
+      "logits/chosen": -0.890625,
+      "logits/rejected": -0.8828125,
+      "logps/chosen": -14.3125,
+      "logps/rejected": -14.3125,
+      "loss": 1.1153709888458252,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -136.7957305908203,
+      "rewards/margins": 1.56671142578125,
+      "rewards/rejected": -138.36244201660156,
+      "step": 70
+    },
+    {
+      "epoch": 0.5493562231759657,
+      "grad_norm": 188.6296821878058,
+      "ht_mnpo/logit": 0.040885161608457565,
+      "ht_mnpo/residual": 0.03338516503572464,
+      "ht_mnpo/residual_abs": 0.35882776975631714,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 2.5899116932879534e-07,
+      "logits/chosen": -0.83203125,
+      "logits/rejected": -0.87109375,
+      "logps/chosen": -14.1875,
+      "logps/rejected": -14.1875,
+      "loss": 0.46190714836120605,
+      "rewards/accuracies": 0.4937500059604645,
+      "rewards/chosen": -136.6307373046875,
+      "rewards/margins": 0.40885162353515625,
+      "rewards/rejected": -137.03958129882812,
+      "step": 80
+    },
+    {
+      "epoch": 0.6180257510729614,
+      "grad_norm": 159.63381058216007,
+      "ht_mnpo/logit": 0.2624649107456207,
+      "ht_mnpo/residual": 0.2549648880958557,
+      "ht_mnpo/residual_abs": 0.4282897412776947,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 1.993912420112756e-07,
+      "logits/chosen": -0.83984375,
+      "logits/rejected": -0.84375,
+      "logps/chosen": -14.0625,
+      "logps/rejected": -14.125,
+      "loss": 0.8312055468559265,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -134.35748291015625,
+      "rewards/margins": 2.6246490478515625,
+      "rewards/rejected": -136.9821319580078,
+      "step": 90
+    },
+    {
+      "epoch": 0.6866952789699571,
+      "grad_norm": 217.2628805931693,
+      "ht_mnpo/logit": 0.059761811047792435,
+      "ht_mnpo/residual": 0.05226181820034981,
+      "ht_mnpo/residual_abs": 0.303397536277771,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 1.426879939713322e-07,
+      "logits/chosen": -0.86328125,
+      "logits/rejected": -0.875,
+      "logps/chosen": -13.875,
+      "logps/rejected": -13.875,
+      "loss": 0.22050431370735168,
+      "rewards/accuracies": 0.518750011920929,
+      "rewards/chosen": -134.1347198486328,
+      "rewards/margins": 0.5976181030273438,
+      "rewards/rejected": -134.73233032226562,
+      "step": 100
+    },
+    {
+      "epoch": 0.7553648068669528,
+      "grad_norm": 307.6598272099915,
+      "ht_mnpo/logit": 0.08397064357995987,
+      "ht_mnpo/residual": 0.07647065818309784,
+      "ht_mnpo/residual_abs": 0.3651581406593323,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 9.212693310317479e-08,
+      "logits/chosen": -0.83203125,
+      "logits/rejected": -0.84765625,
+      "logps/chosen": -13.9375,
+      "logps/rejected": -13.9375,
+      "loss": 0.46880072355270386,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": -133.74819946289062,
+      "rewards/margins": 0.8397064208984375,
+      "rewards/rejected": -134.58790588378906,
+      "step": 110
+    },
+    {
+      "epoch": 0.8240343347639485,
+      "grad_norm": 101.02657333091942,
+      "ht_mnpo/logit": 0.02447204664349556,
+      "ht_mnpo/residual": 0.016972053796052933,
+      "ht_mnpo/residual_abs": 0.24492737650871277,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 5.060200865767605e-08,
+      "logits/chosen": -0.80859375,
+      "logits/rejected": -0.8359375,
+      "logps/chosen": -13.8125,
+      "logps/rejected": -13.875,
+      "loss": 0.1213923841714859,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -133.80438232421875,
+      "rewards/margins": 0.244720458984375,
+      "rewards/rejected": -134.04910278320312,
+      "step": 120
+    },
+    {
+      "epoch": 0.8927038626609443,
+      "grad_norm": 62.457306149658216,
+      "ht_mnpo/logit": 0.18063659965991974,
+      "ht_mnpo/residual": 0.1731366068124771,
+      "ht_mnpo/residual_abs": 0.3555470108985901,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 2.0489971081290193e-08,
+      "logits/chosen": -0.78125,
+      "logits/rejected": -0.80859375,
+      "logps/chosen": -13.875,
+      "logps/rejected": -13.9375,
+      "loss": 0.5861908197402954,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -133.495361328125,
+      "rewards/margins": 1.806365966796875,
+      "rewards/rejected": -135.30172729492188,
+      "step": 130
+    },
+    {
+      "epoch": 0.9613733905579399,
+      "grad_norm": 59.652930086354274,
+      "ht_mnpo/logit": 0.009463501162827015,
+      "ht_mnpo/residual": 0.001963504124432802,
+      "ht_mnpo/residual_abs": 0.3243860602378845,
+      "ht_mnpo/target": 0.007499999366700649,
+      "ht_mnpo/target_abs": 0.007499999366700649,
+      "learning_rate": 3.5143346177878565e-09,
+      "logits/chosen": -0.79296875,
+      "logits/rejected": -0.83203125,
+      "logps/chosen": -13.875,
+      "logps/rejected": -13.9375,
+      "loss": 0.5267720818519592,
+      "rewards/accuracies": 0.518750011920929,
+      "rewards/chosen": -134.2991943359375,
+      "rewards/margins": 0.094635009765625,
+      "rewards/rejected": -134.39382934570312,
+      "step": 140
+    },
+    {
+      "epoch": 1.0,
+      "eval_ht_mnpo/logit": 0.0946219190955162,
+      "eval_ht_mnpo/residual": 0.08712193369865417,
+      "eval_ht_mnpo/residual_abs": 0.23038248717784882,
+      "eval_ht_mnpo/target": 0.007499998901039362,
+      "eval_ht_mnpo/target_abs": 0.007499998901039362,
+      "eval_logits/chosen": -0.78515625,
+      "eval_logits/rejected": -0.76953125,
+      "eval_logps/chosen": -13.8125,
+      "eval_logps/rejected": -13.875,
+      "eval_loss": 0.09535243362188339,
+      "eval_rewards/accuracies": 0.6578947305679321,
+      "eval_rewards/chosen": -134.2300262451172,
+      "eval_rewards/margins": 0.9462206959724426,
+      "eval_rewards/rejected": -135.17623901367188,
+      "eval_runtime": 70.1838,
+      "eval_samples_per_second": 8.606,
+      "eval_steps_per_second": 1.083,
+      "step": 146
+    },
+    {
+      "epoch": 1.0,
+      "step": 146,
+      "total_flos": 0.0,
+      "train_loss": 0.5071899907229698,
+      "train_runtime": 3280.3544,
+      "train_samples_per_second": 5.682,
+      "train_steps_per_second": 0.045
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 146,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2c2e246504705393c26a1d7d6037bc4db16586f5468f14d8745f3b477fa8aee
+size 8401