Model save

README.md

@@ -1,6 +1,5 @@
 ---
 base_model: JayHyeon/Qwen2.5-0.5B-SFT-2e-5-2ep
-datasets: trl-lib/ultrafeedback_binarized
 library_name: transformers
 model_name: Qwen_0.5-VIPO_5e-7-3ep_0alp_0lam
 tags:
@@ -12,7 +11,7 @@ licence: license
 
 # Model Card for Qwen_0.5-VIPO_5e-7-3ep_0alp_0lam
 
-This model is a fine-tuned version of [JayHyeon/Qwen2.5-0.5B-SFT-2e-5-2ep](https://huggingface.co/JayHyeon/Qwen2.5-0.5B-SFT-2e-5-2ep) on the [trl-lib/ultrafeedback_binarized](https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized) dataset.
+This model is a fine-tuned version of [JayHyeon/Qwen2.5-0.5B-SFT-2e-5-2ep](https://huggingface.co/JayHyeon/Qwen2.5-0.5B-SFT-2e-5-2ep).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start
@@ -28,7 +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/bonin147/huggingface/runs/uf2of27b)
+[<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/bonin147/huggingface/runs/0x5zonoo)
 
 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

@@ -1,21 +1,21 @@
 {
-    "epoch": 2.717948717948718,
-    "eval_log_ratio_diff/mean": 0.00030517578125,
-    "eval_logits/chosen": -1.34375,
-    "eval_logits/rejected": -1.3046875,
-    "eval_logps/chosen": -1.046875,
-    "eval_logps/rejected": -1.265625,
-    "eval_loss": 7.732848167419434,
-    "eval_nll_loss": 1.046875,
+    "epoch": 2.999098751126561,
+    "eval_log_ratio_diff/mean": 0.625,
+    "eval_logits/chosen": -2.640625,
+    "eval_logits/rejected": -2.578125,
+    "eval_logps/chosen": -2.046875,
+    "eval_logps/rejected": -2.890625,
+    "eval_loss": 5.383897304534912,
+    "eval_nll_loss": 2.046875,
     "eval_ref_probs/chosen": 0.39426562190055847,
     "eval_ref_probs/rejected": 0.3336612582206726,
-    "eval_rejected_term/max": 0.00020611572836060077,
-    "eval_rejected_term/min": 0.00020611572836060077,
-    "eval_rewards/accuracies": 0.31200000643730164,
-    "eval_rewards/chosen": 3.170967102050781e-05,
-    "eval_rewards/margins": 3.0517578125e-05,
-    "eval_rewards/rejected": 1.3560056686401367e-06,
-    "eval_runtime": 25.3241,
-    "eval_samples_per_second": 39.488,
-    "eval_steps_per_second": 9.872
+    "eval_rejected_term/max": -0.09389671683311462,
+    "eval_rejected_term/min": -0.09389671683311462,
+    "eval_rewards/accuracies": 0.7360000014305115,
+    "eval_rewards/chosen": -0.099609375,
+    "eval_rewards/margins": 0.0625,
+    "eval_rewards/rejected": -0.162109375,
+    "eval_runtime": 25.2148,
+    "eval_samples_per_second": 39.659,
+    "eval_steps_per_second": 9.915
 }

eval_results.json

@@ -1,21 +1,21 @@
 {
-    "epoch": 2.717948717948718,
-    "eval_log_ratio_diff/mean": 0.00030517578125,
-    "eval_logits/chosen": -1.34375,
-    "eval_logits/rejected": -1.3046875,
-    "eval_logps/chosen": -1.046875,
-    "eval_logps/rejected": -1.265625,
-    "eval_loss": 7.732848167419434,
-    "eval_nll_loss": 1.046875,
+    "epoch": 2.999098751126561,
+    "eval_log_ratio_diff/mean": 0.625,
+    "eval_logits/chosen": -2.640625,
+    "eval_logits/rejected": -2.578125,
+    "eval_logps/chosen": -2.046875,
+    "eval_logps/rejected": -2.890625,
+    "eval_loss": 5.383897304534912,
+    "eval_nll_loss": 2.046875,
     "eval_ref_probs/chosen": 0.39426562190055847,
     "eval_ref_probs/rejected": 0.3336612582206726,
-    "eval_rejected_term/max": 0.00020611572836060077,
-    "eval_rejected_term/min": 0.00020611572836060077,
-    "eval_rewards/accuracies": 0.31200000643730164,
-    "eval_rewards/chosen": 3.170967102050781e-05,
-    "eval_rewards/margins": 3.0517578125e-05,
-    "eval_rewards/rejected": 1.3560056686401367e-06,
-    "eval_runtime": 25.3241,
-    "eval_samples_per_second": 39.488,
-    "eval_steps_per_second": 9.872
+    "eval_rejected_term/max": -0.09389671683311462,
+    "eval_rejected_term/min": -0.09389671683311462,
+    "eval_rewards/accuracies": 0.7360000014305115,
+    "eval_rewards/chosen": -0.099609375,
+    "eval_rewards/margins": 0.0625,
+    "eval_rewards/rejected": -0.162109375,
+    "eval_runtime": 25.2148,
+    "eval_samples_per_second": 39.659,
+    "eval_steps_per_second": 9.915
 }