初始化项目，由ModelHub XC社区提供模型

Model: jackf857/llama-3-8b-base-cpo-ultrafeedback-8xh200
Source: Original Platform

all_results.json

@@ -0,0 +1,23 @@
+{
+    "epoch": 0.9989528795811519,
+    "eval_logits/chosen": -0.7640536427497864,
+    "eval_logits/rejected": -0.7481772303581238,
+    "eval_logps/chosen": -382.6802062988281,
+    "eval_logps/rejected": -400.5572814941406,
+    "eval_loss": 5.6526360511779785,
+    "eval_nll_loss": 1.4057114124298096,
+    "eval_rewards/accuracies": 0.6169354915618896,
+    "eval_rewards/chosen": -38.26802062988281,
+    "eval_rewards/margins": 1.7877097129821777,
+    "eval_rewards/rejected": -40.05573272705078,
+    "eval_runtime": 22.8992,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 87.339,
+    "eval_steps_per_second": 2.751,
+    "total_flos": 0.0,
+    "train_loss": 28.95567873588898,
+    "train_runtime": 2691.8939,
+    "train_samples": 61135,
+    "train_samples_per_second": 22.711,
+    "train_steps_per_second": 0.177
+}