End of training

README.md

@@ -2,9 +2,15 @@
 license: llama2
 base_model: Minbyul/meditron-7b-wo-kqa_golden-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-kqa_golden
   results: []
@@ -15,7 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # meditron-7b-dpo-full-sft-wo-kqa_golden
 
-This model is a fine-tuned version of [Minbyul/meditron-7b-wo-kqa_golden-sft](https://huggingface.co/Minbyul/meditron-7b-wo-kqa_golden-sft) on an unknown dataset.
+This model is a fine-tuned version of [Minbyul/meditron-7b-wo-kqa_golden-sft](https://huggingface.co/Minbyul/meditron-7b-wo-kqa_golden-sft) on the HuggingFaceH4/ultrafeedback_binarized dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.6415
+- Rewards/chosen: -0.0188
+- Rewards/rejected: -0.1421
+- Rewards/accuracies: 0.7750
+- Rewards/margins: 0.1233
+- Logps/rejected: -693.1373
+- Logps/chosen: -157.5604
+- Logits/rejected: -1.0479
+- Logits/chosen: -1.3209
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -1.3208973407745361,
+    "eval_logits/rejected": -1.0479389429092407,
+    "eval_logps/chosen": -157.5604248046875,
+    "eval_logps/rejected": -693.1373291015625,
+    "eval_loss": 0.6415364146232605,
+    "eval_rewards/accuracies": 0.7749999761581421,
+    "eval_rewards/chosen": -0.018806004896759987,
+    "eval_rewards/margins": 0.12330160290002823,
+    "eval_rewards/rejected": -0.14210760593414307,
+    "eval_runtime": 16.2949,
+    "eval_samples": 144,
+    "eval_samples_per_second": 8.837,
+    "eval_steps_per_second": 0.307,
     "train_loss": 0.5870200258786561,
     "train_runtime": 812.4392,
     "train_samples": 3886,

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": 1.0,
+    "eval_logits/chosen": -1.3208973407745361,
+    "eval_logits/rejected": -1.0479389429092407,
+    "eval_logps/chosen": -157.5604248046875,
+    "eval_logps/rejected": -693.1373291015625,
+    "eval_loss": 0.6415364146232605,
+    "eval_rewards/accuracies": 0.7749999761581421,
+    "eval_rewards/chosen": -0.018806004896759987,
+    "eval_rewards/margins": 0.12330160290002823,
+    "eval_rewards/rejected": -0.14210760593414307,
+    "eval_runtime": 16.2949,
+    "eval_samples": 144,
+    "eval_samples_per_second": 8.837,
+    "eval_steps_per_second": 0.307
+}