End of training

README.md

@@ -2,9 +2,10 @@
 license: gemma
 library_name: peft
 tags:
+- llama-factory
+- lora
 - trl
 - dpo
-- llama-factory
 - generated_from_trainer
 base_model: google/gemma-2b-it
 model-index:
@@ -17,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # Gemma-2B-It-ORPO
 
-This model is a fine-tuned version of [google/gemma-2b-it](https://huggingface.co/google/gemma-2b-it) on an unknown dataset.
+This model is a fine-tuned version of [google/gemma-2b-it](https://huggingface.co/google/gemma-2b-it) on the dpo_mix_en dataset.
 It achieves the following results on the evaluation set:
 - Loss: 1.3996
 - Rewards/chosen: -0.1332

all_results.json

@@ -0,0 +1,22 @@
+{
+    "epoch": 2.997999555456768,
+    "eval_logits/chosen": -20.655466079711914,
+    "eval_logits/rejected": -20.744272232055664,
+    "eval_logps/chosen": -1.3319367170333862,
+    "eval_logps/rejected": -1.5009632110595703,
+    "eval_loss": 1.3996269702911377,
+    "eval_odds_ratio_loss": 0.67690110206604,
+    "eval_rewards/accuracies": 0.5460000038146973,
+    "eval_rewards/chosen": -0.133193701505661,
+    "eval_rewards/margins": 0.016902634873986244,
+    "eval_rewards/rejected": -0.1500963419675827,
+    "eval_runtime": 79.967,
+    "eval_samples_per_second": 12.505,
+    "eval_sft_loss": 1.3319367170333862,
+    "eval_steps_per_second": 6.253,
+    "total_flos": 5.313908590588723e+17,
+    "train_loss": 1.477055920117832,
+    "train_runtime": 8055.1491,
+    "train_samples_per_second": 3.351,
+    "train_steps_per_second": 0.209
+}

eval_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 2.997999555456768,
+    "eval_logits/chosen": -20.655466079711914,
+    "eval_logits/rejected": -20.744272232055664,
+    "eval_logps/chosen": -1.3319367170333862,
+    "eval_logps/rejected": -1.5009632110595703,
+    "eval_loss": 1.3996269702911377,
+    "eval_odds_ratio_loss": 0.67690110206604,
+    "eval_rewards/accuracies": 0.5460000038146973,
+    "eval_rewards/chosen": -0.133193701505661,
+    "eval_rewards/margins": 0.016902634873986244,
+    "eval_rewards/rejected": -0.1500963419675827,
+    "eval_runtime": 79.967,
+    "eval_samples_per_second": 12.505,
+    "eval_sft_loss": 1.3319367170333862,
+    "eval_steps_per_second": 6.253
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 2.997999555456768,
+    "total_flos": 5.313908590588723e+17,
+    "train_loss": 1.477055920117832,
+    "train_runtime": 8055.1491,
+    "train_samples_per_second": 3.351,
+    "train_steps_per_second": 0.209
+}