初始化项目，由ModelHub XC社区提供模型

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

results/hyperparameters.csv

@@ -0,0 +1,38 @@
+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,RLSD_TR,run_qwen3_generalization.sh
+Run identity,Config,rlsd,run_qwen3_generalization.sh
+Run identity,Experiment,qwen3gen-material-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8,run_qwen3_generalization.sh
+Run identity,W&B run,run-20260702_135403-jd6eugui,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,RLSD_TR 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-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-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-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8/_actor_archive,preserve_actor_checkpoints.py
+Checkpoint/Logging,Logger,"console, wandb",ppo_trainer.yaml
+RLSD_TR,Policy loss mode,rlsd,method override
+RLSD_TR,Teacher regularization,trust-region,method override
+RLSD_TR,Trust-region mix / teacher update rate,0.1,queue/script override
+RLSD_TR,Token reweight lambda,0.5,queue/script override
+RLSD_TR,Token reweight eps_w,0.2,queue/script override
+RLSD_TR,Token reweight decay steps,0,queue/script override
+RLSD_TR,Fused kernels,False,method override

results/summary.json

@@ -0,0 +1,27 @@
+{
+  "repo_id": "SeongryongJung/Qwen3-4B-Material-RLSD-TR",
+  "output_dir": "/mnt/mole/SDPO/L2T/hf_upload_tr/Qwen3-4B-Material-RLSD-TR",
+  "experiment": "qwen3gen-material-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8",
+  "best_step": 100,
+  "best_val_mean16": 0.7686170212765957,
+  "best_actor_dir": "/mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/material/qwen3gen-material-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8/global_step_100/actor",
+  "final_step": 100,
+  "final_val_mean16": 0.7686170212765957,
+  "final_actor_dir": "/mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/material/qwen3gen-material-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-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 sha256:7903d18b1763e5c4a0dc70ce433e1486643cbf173f4eec1daa493a56601706ff
+size 133928

results/training_scores.csv

@@ -0,0 +1,91 @@
+step,critic_score_mean
+1,0.55078125
+2,0.5078125
+3,0.6171875
+4,0.62109375
+5,0.55078125
+6,0.625
+7,0.66796875
+8,0.62109375
+9,0.62109375
+11,0.48046875
+12,0.61328125
+13,0.51953125
+14,0.7890625
+15,0.671875
+16,0.70703125
+17,0.578125
+18,0.609375
+19,0.65625
+21,0.59375
+22,0.7109375
+23,0.78125
+24,0.875
+25,0.64453125
+26,0.66796875
+27,0.7890625
+28,0.63671875
+29,0.6640625
+31,0.6328125
+32,0.6875
+33,0.74609375
+34,0.8125
+35,0.765625
+36,0.66796875
+37,0.8359375
+38,0.5859375
+39,0.75390625
+41,0.5859375
+42,0.69140625
+43,0.546875
+44,0.734375
+45,0.8125
+46,0.765625
+47,0.6953125
+48,0.8203125
+49,0.6171875
+51,0.671875
+52,0.7421875
+53,0.65234375
+54,0.5703125
+55,0.73828125
+56,0.7890625
+57,0.7578125
+58,0.8359375
+59,0.70703125
+61,0.83984375
+62,0.82421875
+63,0.68359375
+64,0.7890625
+65,0.62890625
+66,0.72265625
+67,0.80078125
+68,0.8203125
+69,0.828125
+71,0.78515625
+72,0.7890625
+73,0.6875
+74,0.84375
+75,0.66015625
+76,0.69140625
+77,0.5
+78,0.6328125
+79,0.80078125
+81,0.69921875
+82,0.7109375
+83,0.6953125
+84,0.84375
+85,0.71875
+86,0.81640625
+87,0.62109375
+88,0.6640625
+89,0.73828125
+91,0.75
+92,0.69140625
+93,0.7734375
+94,0.75
+95,0.71875
+96,0.6484375
+97,0.87890625
+98,0.921875
+99,0.7578125

results/validation_mean16.csv

@@ -0,0 +1,11 @@
+step,val_mean16
+10,0.6529255319148937
+20,0.7014627659574468
+30,0.7253989361702128
+40,0.7420212765957447
+50,0.7473404255319149
+60,0.7606382978723404
+70,0.7666223404255319
+80,0.7579787234042553
+90,0.7639627659574468
+100,0.7686170212765957