End of training

README.md

@@ -3,9 +3,15 @@ library_name: transformers
 license: other
 base_model: trl-lib/qwen1.5-0.5b-sft
 tags:
+- alignment-handbook
 - trl
 - simpo
 - generated_from_trainer
+- trl
+- simpo
+- generated_from_trainer
+datasets:
+- yakazimir/ultrafeedback_binarized
 model-index:
 - name: qwen_fUNL_entropy
   results: []
@@ -16,7 +22,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # qwen_fUNL_entropy
 
-This model is a fine-tuned version of [trl-lib/qwen1.5-0.5b-sft](https://huggingface.co/trl-lib/qwen1.5-0.5b-sft) on an unknown dataset.
+This model is a fine-tuned version of [trl-lib/qwen1.5-0.5b-sft](https://huggingface.co/trl-lib/qwen1.5-0.5b-sft) on the yakazimir/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.0000
 - Rewards/chosen: -42.7794

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 2.999297541394882,
+    "eval_logits/chosen": 7.539320468902588,
+    "eval_logits/rejected": 7.256679534912109,
+    "eval_logps/chosen": -42.77936935424805,
+    "eval_logps/rejected": -43.914947509765625,
+    "eval_loss": 1.999994623247403e-07,
+    "eval_rewards/accuracies": 0.5667656064033508,
+    "eval_rewards/chosen": -42.77936935424805,
+    "eval_rewards/margins": 1.1355794668197632,
+    "eval_rewards/rejected": -43.914947509765625,
+    "eval_runtime": 40.6405,
+    "eval_samples": 1345,
+    "eval_samples_per_second": 33.095,
+    "eval_steps_per_second": 8.292,
     "total_flos": 0.0,
     "train_loss": 0.010826070491851839,
     "train_runtime": 30412.7115,

config.json

@@ -22,7 +22,7 @@
   "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.44.2",
-  "use_cache": false,
+  "use_cache": true,
   "use_sliding_window": false,
   "vocab_size": 151648
 }

eval_results.json

@@ -1,17 +1,16 @@
 {
     "epoch": 2.999297541394882,
-    "eval_logits/chosen": NaN,
-    "eval_logits/rejected": NaN,
-    "eval_logps/chosen": NaN,
-    "eval_logps/rejected": NaN,
-    "eval_loss": NaN,
-    "eval_rewards/accuracies": 0.0,
-    "eval_rewards/chosen": NaN,
-    "eval_rewards/margins": NaN,
-    "eval_rewards/rejected": NaN,
-    "eval_runtime": 34.2578,
+    "eval_logits/chosen": 7.539320468902588,
+    "eval_logits/rejected": 7.256679534912109,
+    "eval_logps/chosen": -42.77936935424805,
+    "eval_logps/rejected": -43.914947509765625,
+    "eval_loss": 1.999994623247403e-07,
+    "eval_rewards/accuracies": 0.5667656064033508,
+    "eval_rewards/chosen": -42.77936935424805,
+    "eval_rewards/margins": 1.1355794668197632,
+    "eval_rewards/rejected": -43.914947509765625,
+    "eval_runtime": 40.6405,
     "eval_samples": 1345,
-    "eval_samples_per_second": 39.261,
-    "eval_semantic_entropy": NaN,
-    "eval_steps_per_second": 9.837
+    "eval_samples_per_second": 33.095,
+    "eval_steps_per_second": 8.292
 }