初始化项目，由ModelHub XC社区提供模型

Model: SeongryongJung/Qwen3-4B-Tooluse-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,Tool-use / tooluse,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-tooluse-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_154716-jrdzd11b,wandb
+Data,Train file,datasets/tooluse/train.parquet,script override
+Data,Validation file,datasets/tooluse/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/tooluse/qwen3gen-tooluse-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/tooluse/qwen3gen-tooluse-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-Tooluse-RLSD-TR",
+  "output_dir": "/mnt/mole/SDPO/L2T/hf_upload_tr/Qwen3-4B-Tooluse-RLSD-TR",
+  "experiment": "qwen3gen-tooluse-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8",
+  "best_step": 60,
+  "best_val_mean16": 0.6102941176470589,
+  "best_actor_dir": "/mnt/mole/SDPO/L2T/checkpoints/datasets/tooluse/qwen3gen-tooluse-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8/_actor_archive/global_step_60/actor",
+  "final_step": 100,
+  "final_val_mean16": 0.59375,
+  "final_actor_dir": "/mnt/mole/SDPO/L2T/checkpoints/datasets/tooluse/qwen3gen-tooluse-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 sha256:95370ee52efbdf669b5fc10e15ecc460a787c3cc495ea0344c4857fb23494324
+size 143231

results/training_scores.csv

@@ -0,0 +1,91 @@
+step,critic_score_mean
+1,0.25390625
+2,0.18359375
+3,0.19921875
+4,0.28515625
+5,0.2890625
+6,0.31640625
+7,0.47265625
+8,0.3359375
+9,0.296875
+11,0.3984375
+12,0.26953125
+13,0.39453125
+14,0.30859375
+15,0.30078125
+16,0.21875
+17,0.40234375
+18,0.38671875
+19,0.44921875
+21,0.3125
+22,0.359375
+23,0.49609375
+24,0.26953125
+25,0.34375
+26,0.4140625
+27,0.2421875
+28,0.5
+29,0.1796875
+31,0.3828125
+32,0.44140625
+33,0.234375
+34,0.27734375
+35,0.33203125
+36,0.43359375
+37,0.3046875
+38,0.19140625
+39,0.44921875
+41,0.24609375
+42,0.33984375
+43,0.296875
+44,0.296875
+45,0.421875
+46,0.25
+47,0.37109375
+48,0.265625
+49,0.3515625
+51,0.30078125
+52,0.25
+53,0.4140625
+54,0.3984375
+55,0.26953125
+56,0.30859375
+57,0.30859375
+58,0.4140625
+59,0.37890625
+61,0.2890625
+62,0.4921875
+63,0.4296875
+64,0.2890625
+65,0.4140625
+66,0.28125
+67,0.29296875
+68,0.28125
+69,0.29296875
+71,0.484375
+72,0.34375
+73,0.28515625
+74,0.3046875
+75,0.44921875
+76,0.29296875
+77,0.375
+78,0.25
+79,0.421875
+81,0.4453125
+82,0.38671875
+83,0.48046875
+84,0.30078125
+85,0.38671875
+86,0.4453125
+87,0.33984375
+88,0.31640625
+89,0.43359375
+91,0.40625
+92,0.3828125
+93,0.3125
+94,0.3671875
+95,0.4140625
+96,0.35546875
+97,0.35546875
+98,0.35546875
+99,0.44140625

results/validation_mean16.csv

@@ -0,0 +1,11 @@
+step,val_mean16
+10,0.5762867647058824
+20,0.5863970588235294
+30,0.5753676470588235
+40,0.5707720588235294
+50,0.5533088235294118
+60,0.6102941176470589
+70,0.5946691176470589
+80,0.5928308823529411
+90,0.53125
+100,0.59375