End of training

README.md

@@ -3,21 +3,31 @@ library_name: transformers
 license: other
 base_model: trl-lib/qwen1.5-0.5b-sft
 tags:
+- alignment-handbook
 - trl
 - simpo
-- alignment-handbook
 - generated_from_trainer
 model-index:
-- name: simpo-exps_qwen05b
+- name: trl-lib/qwen1.5-0.5b-sft
   results: []
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-# simpo-exps_qwen05b
+# trl-lib/qwen1.5-0.5b-sft
 
 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.
+It achieves the following results on the evaluation set:
+- Loss: 5.8364
+- Rewards/chosen: -16.6906
+- Rewards/rejected: -17.6492
+- Rewards/accuracies: 0.5242
+- Rewards/margins: 0.9586
+- Logps/rejected: -1.7649
+- Logps/chosen: -1.6691
+- Logits/rejected: -0.3821
+- Logits/chosen: -0.4446
 
 ## Model description
 

all_results.json

@@ -1,18 +1,18 @@
 {
     "epoch": 0.005211431160959583,
-    "eval_logits/chosen": -0.3587622344493866,
-    "eval_logits/rejected": -0.32808682322502136,
-    "eval_logps/chosen": -1.7873238325119019,
-    "eval_logps/rejected": -1.8731167316436768,
-    "eval_loss": 6.489209175109863,
-    "eval_rewards/accuracies": 0.5237388610839844,
-    "eval_rewards/chosen": -17.87323760986328,
-    "eval_rewards/margins": 0.8579314351081848,
-    "eval_rewards/rejected": -18.731168746948242,
-    "eval_runtime": 60.3757,
-    "eval_samples": 1345,
-    "eval_samples_per_second": 22.277,
-    "eval_steps_per_second": 5.582,
+    "eval_logits/chosen": -0.44456326961517334,
+    "eval_logits/rejected": -0.3820805549621582,
+    "eval_logps/chosen": -1.669058084487915,
+    "eval_logps/rejected": -1.7649223804473877,
+    "eval_loss": 5.83641242980957,
+    "eval_rewards/accuracies": 0.524183988571167,
+    "eval_rewards/chosen": -16.690580368041992,
+    "eval_rewards/margins": 0.9586421847343445,
+    "eval_rewards/rejected": -17.64922523498535,
+    "eval_runtime": 292.3296,
+    "eval_samples": 5595,
+    "eval_samples_per_second": 19.139,
+    "eval_steps_per_second": 4.786,
     "total_flos": 0.0,
     "train_loss": 3.4047957710597827,
     "train_runtime": 14.6858,

config.json

@@ -24,7 +24,7 @@
   "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.48.3",
-  "use_cache": false,
+  "use_cache": true,
   "use_sliding_window": false,
   "vocab_size": 151648
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 0.005887272119083459,
-    "eval_logits/chosen": -0.3587622344493866,
-    "eval_logits/rejected": -0.32808682322502136,
-    "eval_logps/chosen": -1.7873238325119019,
-    "eval_logps/rejected": -1.8731167316436768,
-    "eval_loss": 6.489209175109863,
-    "eval_rewards/accuracies": 0.5237388610839844,
-    "eval_rewards/chosen": -17.87323760986328,
-    "eval_rewards/margins": 0.8579314351081848,
-    "eval_rewards/rejected": -18.731168746948242,
-    "eval_runtime": 60.3757,
-    "eval_samples": 1345,
-    "eval_samples_per_second": 22.277,
-    "eval_steps_per_second": 5.582
+    "epoch": 0.005211431160959583,
+    "eval_logits/chosen": -0.44456326961517334,
+    "eval_logits/rejected": -0.3820805549621582,
+    "eval_logps/chosen": -1.669058084487915,
+    "eval_logps/rejected": -1.7649223804473877,
+    "eval_loss": 5.83641242980957,
+    "eval_rewards/accuracies": 0.524183988571167,
+    "eval_rewards/chosen": -16.690580368041992,
+    "eval_rewards/margins": 0.9586421847343445,
+    "eval_rewards/rejected": -17.64922523498535,
+    "eval_runtime": 292.3296,
+    "eval_samples": 5595,
+    "eval_samples_per_second": 19.139,
+    "eval_steps_per_second": 4.786
 }