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-BDPO_1e-6-1ep_0alp_0.5bdpo_lam_0dpop_lam
 tags:
@@ -12,7 +11,7 @@ licence: license
 
 # Model Card for llama-BDPO_1e-6-1ep_0alp_0.5bdpo_lam_0dpop_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/fuzgcv93) 
+[<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/stdyd2kd) 
 
 
 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": 1.0979223251342773,
-    "eval_logps/chosen": -353.5060119628906,
-    "eval_logps/rejected": -325.21649169921875,
+    "eval_logits/rejected": 1.1020585298538208,
+    "eval_logps/chosen": -354.015869140625,
+    "eval_logps/rejected": -324.1587219238281,
     "eval_loss": NaN,
     "eval_nll_loss": NaN,
-    "eval_rewards/accuracies": 0.6710000038146973,
-    "eval_rewards/chosen": 1.5991406440734863,
-    "eval_rewards/margins": 0.5435751676559448,
-    "eval_rewards/rejected": 1.0555741786956787,
-    "eval_runtime": 63.1799,
-    "eval_samples_per_second": 15.828,
-    "eval_steps_per_second": 3.957
+    "eval_rewards/accuracies": 0.6617063283920288,
+    "eval_rewards/chosen": 1.3769841194152832,
+    "eval_rewards/margins": 0.4590735137462616,
+    "eval_rewards/rejected": 0.9184337854385376,
+    "eval_runtime": 8.4136,
+    "eval_samples_per_second": 118.856,
+    "eval_steps_per_second": 7.488
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 1.0,
     "eval_logits/chosen": NaN,
-    "eval_logits/rejected": 1.0979223251342773,
-    "eval_logps/chosen": -353.5060119628906,
-    "eval_logps/rejected": -325.21649169921875,
+    "eval_logits/rejected": 1.1020585298538208,
+    "eval_logps/chosen": -354.015869140625,
+    "eval_logps/rejected": -324.1587219238281,
     "eval_loss": NaN,
     "eval_nll_loss": NaN,
-    "eval_rewards/accuracies": 0.6710000038146973,
-    "eval_rewards/chosen": 1.5991406440734863,
-    "eval_rewards/margins": 0.5435751676559448,
-    "eval_rewards/rejected": 1.0555741786956787,
-    "eval_runtime": 63.1799,
-    "eval_samples_per_second": 15.828,
-    "eval_steps_per_second": 3.957
+    "eval_rewards/accuracies": 0.6617063283920288,
+    "eval_rewards/chosen": 1.3769841194152832,
+    "eval_rewards/margins": 0.4590735137462616,
+    "eval_rewards/rejected": 0.9184337854385376,
+    "eval_runtime": 8.4136,
+    "eval_samples_per_second": 118.856,
+    "eval_steps_per_second": 7.488
 }