Model save

README.md

@@ -27,7 +27,7 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yuchenl4/lmpref/runs/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs256_lr5e-06_0try1Np798OezumxB5XeTF3RtHQdWE8kF58TTgI9hJ74c9OvNz)
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yuchenl4/lmpref/runs/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs256_lr5e-06_0try1BkYpU9eF8GSJTrDxIJzrzwSUTgp6NKZMTBRNCcZDnwFFf)
 
 This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).
 

all_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 0.9971988795518207,
     "total_flos": 0.0,
-    "train_loss": 0.44990385248419945,
-    "train_runtime": 40900.2747,
+    "train_loss": 0.5297158916344803,
+    "train_runtime": 36666.0128,
     "train_samples": 45608,
-    "train_samples_per_second": 1.115,
-    "train_steps_per_second": 0.004
+    "train_samples_per_second": 1.244,
+    "train_steps_per_second": 0.005
 }

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 0.9971988795518207,
     "total_flos": 0.0,
-    "train_loss": 0.44990385248419945,
-    "train_runtime": 40900.2747,
+    "train_loss": 0.5297158916344803,
+    "train_runtime": 36666.0128,
     "train_samples": 45608,
-    "train_samples_per_second": 1.115,
-    "train_steps_per_second": 0.004
+    "train_samples_per_second": 1.244,
+    "train_steps_per_second": 0.005
 }