初始化项目，由ModelHub XC社区提供模型

Model: SeongryongJung/Qwen3-4B-Material-GRPO-TR
Source: Original Platform

results/hyperparameters.csv

@@ -0,0 +1,33 @@
+section,parameter,value,source
+Run identity,Base model,Qwen/Qwen3-4B,queue/script override
+Run identity,Dataset,Material / SciKnowEval material,run_qwen3_generalization.sh
+Run identity,Method,GRPO,run_qwen3_generalization.sh
+Run identity,Config,baseline_grpo,run_qwen3_generalization.sh
+Run identity,Experiment,qwen3gen-material-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8,run_qwen3_generalization.sh
+Run identity,W&B run,run-20260702_125526-lnzvk3fv,wandb
+Data,Train file,datasets/sciknoweval/material/train.parquet,script override
+Data,Validation file,datasets/sciknoweval/material/test.parquet,script override
+Data,Train batch size,32,queue/script override
+Data,Train max samples,3200,queue/script override
+Schedule,Total training steps,100,queue/script override
+Schedule,Validation before train,False,queue/script override
+Schedule,Save frequency,10,queue/script override
+Schedule,Validation frequency,10,queue/script override
+Sequence,Max prompt length,2048,queue/script override
+Sequence,Max response length,8192,queue/script override
+Sequence,Max model length,10240,queue/script override
+Rollout,Train rollout n,8,queue/script override
+Rollout,Validation rollout n,16,queue/script override
+Rollout,vLLM GPU memory utilization,0.8,queue/script override
+Optimization,Learning rate,1e-6,GRPO method override
+Optimization,Weight decay,0.01,script override
+PPO/GRPO,PPO mini batch size,8,queue/script override
+PPO/GRPO,Normalize GRPO advantages by std,False,baseline_grpo.yaml / script override
+Rollout correction,Importance sampling mode,token,script override
+Rollout correction,IS threshold,2.0,script override
+Checkpoint/Logging,Checkpoint root,checkpoints/datasets/sciknoweval/material/qwen3gen-material-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8,script override
+Checkpoint/Logging,Latest checkpointed iteration,100,latest_checkpointed_iteration.txt
+Checkpoint/Logging,External actor archive,checkpoints/datasets/sciknoweval/material/qwen3gen-material-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/_actor_archive,preserve_actor_checkpoints.py
+Checkpoint/Logging,Logger,"console, wandb",ppo_trainer.yaml
+PPO/GRPO,Policy loss mode,vanilla,method override
+PPO/GRPO,Actor KL loss coef,0.0,method override

results/summary.json

@@ -0,0 +1,27 @@
+{
+  "repo_id": "SeongryongJung/Qwen3-4B-Material-GRPO-TR",
+  "output_dir": "/mnt/mole/SDPO/L2T/hf_upload_tr/Qwen3-4B-Material-GRPO-TR",
+  "experiment": "qwen3gen-material-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8",
+  "best_step": 60,
+  "best_val_mean16": 0.7659574468085106,
+  "best_actor_dir": "/mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/material/qwen3gen-material-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/_actor_archive/global_step_60/actor",
+  "final_step": 100,
+  "final_val_mean16": 0.7626329787234043,
+  "final_actor_dir": "/mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/material/qwen3gen-material-GRPO-Qwen-Qwen3-4B-mbs8-train32-rollout8-lr1e-6-vllm0.8/global_step_100/actor",
+  "train_rows": 90,
+  "val_rows": 10,
+  "hf_model_files": [
+    "added_tokens.json",
+    "chat_template.jinja",
+    "config.json",
+    "generation_config.json",
+    "merges.txt",
+    "model.safetensors.index.json",
+    "special_tokens_map.json",
+    "tokenizer.json",
+    "tokenizer_config.json",
+    "vocab.json",
+    "model-00001-of-00002.safetensors",
+    "model-00002-of-00002.safetensors"
+  ]
+}

results/training_score.png

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b7a0f03d7ae334cc4f31318502f2c08d5134f1349afab19faf7f278572303e2
+size 150554

results/training_scores.csv

@@ -0,0 +1,91 @@
+step,critic_score_mean
+1,0.5625
+2,0.53515625
+3,0.62109375
+4,0.578125
+5,0.56640625
+6,0.640625
+7,0.6484375
+8,0.58984375
+9,0.6328125
+11,0.4921875
+12,0.58984375
+13,0.54296875
+14,0.76953125
+15,0.66015625
+16,0.6953125
+17,0.5703125
+18,0.5546875
+19,0.63671875
+21,0.546875
+22,0.75
+23,0.80859375
+24,0.89453125
+25,0.6796875
+26,0.65234375
+27,0.8203125
+28,0.6171875
+29,0.68359375
+31,0.63671875
+32,0.6640625
+33,0.8203125
+34,0.7890625
+35,0.7734375
+36,0.671875
+37,0.86328125
+38,0.6171875
+39,0.77734375
+41,0.5859375
+42,0.68359375
+43,0.5859375
+44,0.73828125
+45,0.78515625
+46,0.78125
+47,0.69921875
+48,0.83984375
+49,0.640625
+51,0.703125
+52,0.76953125
+53,0.6328125
+54,0.5546875
+55,0.703125
+56,0.81640625
+57,0.7578125
+58,0.828125
+59,0.65234375
+61,0.81640625
+62,0.8125
+63,0.71875
+64,0.74609375
+65,0.60546875
+66,0.70703125
+67,0.78515625
+68,0.81640625
+69,0.83203125
+71,0.74609375
+72,0.7734375
+73,0.70703125
+74,0.83203125
+75,0.69140625
+76,0.734375
+77,0.5546875
+78,0.609375
+79,0.7734375
+81,0.71875
+82,0.7265625
+83,0.6875
+84,0.7421875
+85,0.7421875
+86,0.84765625
+87,0.625
+88,0.6640625
+89,0.74609375
+91,0.80859375
+92,0.7109375
+93,0.78515625
+94,0.73828125
+95,0.76171875
+96,0.58984375
+97,0.85546875
+98,0.8828125
+99,0.71875

results/validation_mean16.csv

@@ -0,0 +1,11 @@
+step,val_mean16
+10,0.6688829787234043
+20,0.6954787234042553
+30,0.7167553191489362
+40,0.7393617021276596
+50,0.754654255319149
+60,0.7659574468085106
+70,0.75
+80,0.7506648936170213
+90,0.7566489361702128
+100,0.7626329787234043