End of training

README.md

@@ -2,9 +2,15 @@
 license: apache-2.0
 base_model: BioMistral/BioMistral-7B
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: biomistral-7b-dpo-full-wo-live_qa-ep3
   results: []
@@ -15,17 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # biomistral-7b-dpo-full-wo-live_qa-ep3
 
-This model is a fine-tuned version of [BioMistral/BioMistral-7B](https://huggingface.co/BioMistral/BioMistral-7B) on an unknown dataset.
+This model is a fine-tuned version of [BioMistral/BioMistral-7B](https://huggingface.co/BioMistral/BioMistral-7B) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
-- Logits/chosen: -1.3915
-- Logits/rejected: 1.0056
-- Logps/chosen: -385.5208
-- Logps/rejected: -374.2370
-- Loss: 0.5345
+- Loss: 0.5347
+- Rewards/chosen: -0.8760
+- Rewards/rejected: -1.7973
 - Rewards/accuracies: 0.6528
-- Rewards/chosen: -0.8742
-- Rewards/margins: 0.9209
-- Rewards/rejected: -1.7951
+- Rewards/margins: 0.9213
+- Logps/rejected: -374.4584
+- Logps/chosen: -385.7039
+- Logits/rejected: 1.0068
+- Logits/chosen: -1.3925
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -1.3924791812896729,
+    "eval_logits/rejected": 1.0068244934082031,
+    "eval_logps/chosen": -385.703857421875,
+    "eval_logps/rejected": -374.45843505859375,
+    "eval_loss": 0.5347221493721008,
+    "eval_rewards/accuracies": 0.6527777910232544,
+    "eval_rewards/chosen": -0.8760086297988892,
+    "eval_rewards/margins": 0.9213289618492126,
+    "eval_rewards/rejected": -1.7973376512527466,
+    "eval_runtime": 61.4466,
+    "eval_samples": 568,
+    "eval_samples_per_second": 9.244,
+    "eval_steps_per_second": 0.293,
     "train_loss": 0.004050073670405967,
     "train_runtime": 90.6565,
     "train_samples": 19572,

config.json

@@ -21,6 +21,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.39.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 3.0,
-    "eval_logits/chosen": 1.0175929069519043,
-    "eval_logits/rejected": 0.28610384464263916,
-    "eval_logps/chosen": -333.84112548828125,
-    "eval_logps/rejected": -77.75727844238281,
-    "eval_loss": 0.8236874341964722,
-    "eval_rewards/accuracies": 0.875,
-    "eval_rewards/chosen": -0.4579945206642151,
-    "eval_rewards/margins": 0.2422550916671753,
-    "eval_rewards/rejected": -0.7002495527267456,
-    "eval_runtime": 4.3045,
-    "eval_samples": 103,
-    "eval_samples_per_second": 23.929,
-    "eval_steps_per_second": 0.465
+    "epoch": 1.0,
+    "eval_logits/chosen": -1.3924791812896729,
+    "eval_logits/rejected": 1.0068244934082031,
+    "eval_logps/chosen": -385.703857421875,
+    "eval_logps/rejected": -374.45843505859375,
+    "eval_loss": 0.5347221493721008,
+    "eval_rewards/accuracies": 0.6527777910232544,
+    "eval_rewards/chosen": -0.8760086297988892,
+    "eval_rewards/margins": 0.9213289618492126,
+    "eval_rewards/rejected": -1.7973376512527466,
+    "eval_runtime": 61.4466,
+    "eval_samples": 568,
+    "eval_samples_per_second": 9.244,
+    "eval_steps_per_second": 0.293
 }