End of training

README.md

@@ -2,9 +2,15 @@
 license: llama2
 base_model: Minbyul/meditron-7b-wo-live_qa-sft
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: meditron-7b-dpo-full-sft-wo-live_qa
   results: []
@@ -15,7 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # meditron-7b-dpo-full-sft-wo-live_qa
 
-This model is a fine-tuned version of [Minbyul/meditron-7b-wo-live_qa-sft](https://huggingface.co/Minbyul/meditron-7b-wo-live_qa-sft) on an unknown dataset.
+This model is a fine-tuned version of [Minbyul/meditron-7b-wo-live_qa-sft](https://huggingface.co/Minbyul/meditron-7b-wo-live_qa-sft) on the HuggingFaceH4/ultrafeedback_binarized dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.6677
+- Rewards/chosen: -0.0447
+- Rewards/rejected: -0.1284
+- Rewards/accuracies: 0.875
+- Rewards/margins: 0.0838
+- Logps/rejected: -656.5873
+- Logps/chosen: -92.6797
+- Logits/rejected: -1.2739
+- Logits/chosen: -1.2205
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.99,
+    "eval_logits/chosen": -1.2204631567001343,
+    "eval_logits/rejected": -1.2738878726959229,
+    "eval_logps/chosen": -92.67973327636719,
+    "eval_logps/rejected": -656.5872802734375,
+    "eval_loss": 0.6677024960517883,
+    "eval_rewards/accuracies": 0.875,
+    "eval_rewards/chosen": -0.04467713087797165,
+    "eval_rewards/margins": 0.0837688222527504,
+    "eval_rewards/rejected": -0.12844595313072205,
+    "eval_runtime": 9.5521,
+    "eval_samples": 56,
+    "eval_samples_per_second": 5.863,
+    "eval_steps_per_second": 0.209,
     "train_loss": 0.5556549103029312,
     "train_runtime": 1091.242,
     "train_samples": 3974,

config.json

@@ -23,6 +23,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.39.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32017
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 0.99,
+    "eval_logits/chosen": -1.2204631567001343,
+    "eval_logits/rejected": -1.2738878726959229,
+    "eval_logps/chosen": -92.67973327636719,
+    "eval_logps/rejected": -656.5872802734375,
+    "eval_loss": 0.6677024960517883,
+    "eval_rewards/accuracies": 0.875,
+    "eval_rewards/chosen": -0.04467713087797165,
+    "eval_rewards/margins": 0.0837688222527504,
+    "eval_rewards/rejected": -0.12844595313072205,
+    "eval_runtime": 9.5521,
+    "eval_samples": 56,
+    "eval_samples_per_second": 5.863,
+    "eval_steps_per_second": 0.209
+}