初始化项目，由ModelHub XC社区提供模型

Model: longtermrisk/Qwen3-4B-Instruct-2507-ftjob-e3f6e890af59
Source: Original Platform

checkpoint-3/trainer_state.json

@@ -0,0 +1,79 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 3,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "grad_norm": 30.078697204589844,
+      "learning_rate": 0.0,
+      "logits/chosen": -3.45875883102417,
+      "logits/rejected": -3.423557996749878,
+      "logps/chosen": -848.84765625,
+      "logps/rejected": -816.5891723632812,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 30.074949264526367,
+      "learning_rate": 2e-05,
+      "logits/chosen": -3.5248451232910156,
+      "logits/rejected": -3.506687879562378,
+      "logps/chosen": -848.84765625,
+      "logps/rejected": -816.5891723632812,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 2
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 18.459684371948242,
+      "learning_rate": 4e-05,
+      "logits/chosen": -3.505913257598877,
+      "logits/rejected": -3.469709873199463,
+      "logps/chosen": -845.7001953125,
+      "logps/rejected": -820.7750854492188,
+      "loss": 0.415,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": 0.3147410452365875,
+      "rewards/margins": 0.7333309650421143,
+      "rewards/rejected": -0.41858991980552673,
+      "step": 3
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 3,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 5000,
+  "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": 2,
+  "trial_name": null,
+  "trial_params": null
+}