Model save

README.md

@@ -1,6 +1,5 @@
 ---
 base_model: meta-llama/Llama-3.2-1B-Instruct
-datasets: trl-lib/ultrafeedback_binarized
 library_name: transformers
 model_name: llama-DPOP_1e-6-1ep_0alp_0.5bdpo_lam_5dpop_lam
 tags:
@@ -12,7 +11,7 @@ licence: license
 
 # Model Card for llama-DPOP_1e-6-1ep_0alp_0.5bdpo_lam_5dpop_lam
 
-This model is a fine-tuned version of [meta-llama/Llama-3.2-1B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct) on the [trl-lib/ultrafeedback_binarized](https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized) dataset.
+This model is a fine-tuned version of [meta-llama/Llama-3.2-1B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct).
 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/kfh83259) 
+[<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/hcc4n724) 
 
 
 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

@@ -1,16 +1,16 @@
 {
     "epoch": 1.0,
     "eval_logits/chosen": NaN,
-    "eval_logits/rejected": 1.1184862852096558,
-    "eval_logps/chosen": -352.75799560546875,
-    "eval_logps/rejected": -324.30499267578125,
+    "eval_logits/rejected": 1.129960298538208,
+    "eval_logps/chosen": -353.19049072265625,
+    "eval_logps/rejected": -322.8888854980469,
     "eval_loss": NaN,
     "eval_nll_loss": NaN,
-    "eval_rewards/accuracies": 0.6639999747276306,
-    "eval_rewards/chosen": 1.667421817779541,
-    "eval_rewards/margins": 0.5185019373893738,
-    "eval_rewards/rejected": 1.1488417387008667,
-    "eval_runtime": 61.789,
-    "eval_samples_per_second": 16.184,
-    "eval_steps_per_second": 4.046
+    "eval_rewards/accuracies": 0.6557539701461792,
+    "eval_rewards/chosen": 1.4410961866378784,
+    "eval_rewards/margins": 0.3948955237865448,
+    "eval_rewards/rejected": 1.0456969738006592,
+    "eval_runtime": 8.4152,
+    "eval_samples_per_second": 118.832,
+    "eval_steps_per_second": 7.486
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 1.0,
     "eval_logits/chosen": NaN,
-    "eval_logits/rejected": 1.1184862852096558,
-    "eval_logps/chosen": -352.75799560546875,
-    "eval_logps/rejected": -324.30499267578125,
+    "eval_logits/rejected": 1.129960298538208,
+    "eval_logps/chosen": -353.19049072265625,
+    "eval_logps/rejected": -322.8888854980469,
     "eval_loss": NaN,
     "eval_nll_loss": NaN,
-    "eval_rewards/accuracies": 0.6639999747276306,
-    "eval_rewards/chosen": 1.667421817779541,
-    "eval_rewards/margins": 0.5185019373893738,
-    "eval_rewards/rejected": 1.1488417387008667,
-    "eval_runtime": 61.789,
-    "eval_samples_per_second": 16.184,
-    "eval_steps_per_second": 4.046
+    "eval_rewards/accuracies": 0.6557539701461792,
+    "eval_rewards/chosen": 1.4410961866378784,
+    "eval_rewards/margins": 0.3948955237865448,
+    "eval_rewards/rejected": 1.0456969738006592,
+    "eval_runtime": 8.4152,
+    "eval_samples_per_second": 118.832,
+    "eval_steps_per_second": 7.486
 }