初始化项目，由ModelHub XC社区提供模型

Model: CriteriaPO/qwen2.5-3b-dpo-coarse
Source: Original Platform

all_results.json

@@ -0,0 +1,22 @@
+{
+    "epoch": 1.0,
+    "eval_logits/chosen": -2.75,
+    "eval_logits/rejected": -2.46875,
+    "eval_logps/chosen": -680.0,
+    "eval_logps/rejected": -1048.0,
+    "eval_loss": 0.23313412070274353,
+    "eval_rewards/accuracies": 0.8960731029510498,
+    "eval_rewards/chosen": -5.125,
+    "eval_rewards/margins": 3.78125,
+    "eval_rewards/rejected": -8.9375,
+    "eval_runtime": 3316.1686,
+    "eval_samples": 95033,
+    "eval_samples_per_second": 28.495,
+    "eval_steps_per_second": 0.445,
+    "total_flos": 0.0,
+    "train_loss": 0.0,
+    "train_runtime": 43.6247,
+    "train_samples": 1701200,
+    "train_samples_per_second": 38996.291,
+    "train_steps_per_second": 304.667
+}