End of training

README.md

@@ -1,9 +1,15 @@
 ---
 base_model: Minbyul/llama2-7b-wo-healthsearch_qa-sft
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: llama2-7b-dpo-full-sft-wo-healthsearch_qa
   results: []
@@ -14,7 +20,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # llama2-7b-dpo-full-sft-wo-healthsearch_qa
 
-This model is a fine-tuned version of [Minbyul/llama2-7b-wo-healthsearch_qa-sft](https://huggingface.co/Minbyul/llama2-7b-wo-healthsearch_qa-sft) on an unknown dataset.
+This model is a fine-tuned version of [Minbyul/llama2-7b-wo-healthsearch_qa-sft](https://huggingface.co/Minbyul/llama2-7b-wo-healthsearch_qa-sft) on the HuggingFaceH4/ultrafeedback_binarized dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.6914
+- Rewards/chosen: -0.0016
+- Rewards/rejected: -0.0049
+- Rewards/accuracies: 0.6105
+- Rewards/margins: 0.0033
+- Logps/rejected: -671.7160
+- Logps/chosen: -334.7261
+- Logits/rejected: -0.1596
+- Logits/chosen: -0.4926
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -0.4925585687160492,
+    "eval_logits/rejected": -0.15963555872440338,
+    "eval_logps/chosen": -334.72607421875,
+    "eval_logps/rejected": -671.7160034179688,
+    "eval_loss": 0.6914218664169312,
+    "eval_rewards/accuracies": 0.6104797720909119,
+    "eval_rewards/chosen": -0.001606568112038076,
+    "eval_rewards/margins": 0.003300695214420557,
+    "eval_rewards/rejected": -0.004907263442873955,
+    "eval_runtime": 630.223,
+    "eval_samples": 6317,
+    "eval_samples_per_second": 10.023,
+    "eval_steps_per_second": 0.314,
     "train_loss": 0.6694252616480777,
     "train_runtime": 502.7814,
     "train_samples": 2413,

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": 32000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 1.0,
+    "eval_logits/chosen": -0.4925585687160492,
+    "eval_logits/rejected": -0.15963555872440338,
+    "eval_logps/chosen": -334.72607421875,
+    "eval_logps/rejected": -671.7160034179688,
+    "eval_loss": 0.6914218664169312,
+    "eval_rewards/accuracies": 0.6104797720909119,
+    "eval_rewards/chosen": -0.001606568112038076,
+    "eval_rewards/margins": 0.003300695214420557,
+    "eval_rewards/rejected": -0.004907263442873955,
+    "eval_runtime": 630.223,
+    "eval_samples": 6317,
+    "eval_samples_per_second": 10.023,
+    "eval_steps_per_second": 0.314
+}