End of training

README.md

@@ -3,9 +3,15 @@ library_name: transformers
 license: apache-2.0
 base_model: mistralai/Mistral-7B-Instruct-v0.2
 tags:
+- alignment-handbook
 - trl
 - simpo
 - generated_from_trainer
+- trl
+- simpo
+- generated_from_trainer
+datasets:
+- princeton-nlp/mistral-instruct-ultrafeedback
 model-index:
 - name: mistral-7b-instruct-v0.2-simpo-full
   results: []
@@ -16,7 +22,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # mistral-7b-instruct-v0.2-simpo-full
 
-This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) on an unknown dataset.
+This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) on the princeton-nlp/mistral-instruct-ultrafeedback dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.8195
+- Rewards/chosen: -1.125
+- Rewards/rejected: -1.1797
+- Rewards/accuracies: 0.5532
+- Rewards/margins: 0.0525
+- Logps/rejected: -0.4727
+- Logps/chosen: -0.4512
+- Logits/rejected: -2.375
+- Logits/chosen: -2.375
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.998660594695955,
+    "eval_logits/chosen": -2.375,
+    "eval_logits/rejected": -2.375,
+    "eval_logps/chosen": -0.451171875,
+    "eval_logps/rejected": -0.47265625,
+    "eval_loss": 0.8194748759269714,
+    "eval_rewards/accuracies": 0.5531914830207825,
+    "eval_rewards/chosen": -1.125,
+    "eval_rewards/margins": 0.052490234375,
+    "eval_rewards/rejected": -1.1796875,
+    "eval_runtime": 63.5524,
+    "eval_samples": 2994,
+    "eval_samples_per_second": 47.111,
+    "eval_steps_per_second": 1.479,
     "total_flos": 0.0,
     "train_loss": 0.8197165558778166,
     "train_runtime": 5383.1312,

config.json

@@ -22,6 +22,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.46.3",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 0.998660594695955,
-    "eval_logits/chosen": -2.515625,
-    "eval_logits/rejected": -2.515625,
-    "eval_logps/chosen": -1.21875,
-    "eval_logps/rejected": -1.3359375,
-    "eval_loss": 0.7703349590301514,
-    "eval_rewards/accuracies": 0.6329787373542786,
-    "eval_rewards/chosen": -3.03125,
-    "eval_rewards/margins": 0.298828125,
-    "eval_rewards/rejected": -3.34375,
-    "eval_runtime": 63.5185,
+    "eval_logits/chosen": -2.375,
+    "eval_logits/rejected": -2.375,
+    "eval_logps/chosen": -0.451171875,
+    "eval_logps/rejected": -0.47265625,
+    "eval_loss": 0.8194748759269714,
+    "eval_rewards/accuracies": 0.5531914830207825,
+    "eval_rewards/chosen": -1.125,
+    "eval_rewards/margins": 0.052490234375,
+    "eval_rewards/rejected": -1.1796875,
+    "eval_runtime": 63.5524,
     "eval_samples": 2994,
-    "eval_samples_per_second": 47.136,
-    "eval_steps_per_second": 1.48
+    "eval_samples_per_second": 47.111,
+    "eval_steps_per_second": 1.479
 }