End of training

README.md

@@ -2,10 +2,10 @@
 license: apache-2.0
 library_name: peft
 tags:
-- trl
-- dpo
 - llama-factory
 - lora
+- trl
+- dpo
 - generated_from_trainer
 base_model: mistralai/Mistral-7B-Instruct-v0.2
 model-index:
@@ -18,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # Mistral-7B-Instruct-v0.2-ORPO
 
-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 dpo_mix_en dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.8975
 - Rewards/chosen: -0.0835

all_results.json

@@ -1,22 +1,22 @@
 {
-    "epoch": 2.986666666666667,
-    "eval_logits/chosen": -2.677677869796753,
-    "eval_logits/rejected": -2.6818528175354004,
-    "eval_logps/chosen": -1.0619781017303467,
-    "eval_logps/rejected": -1.1974726915359497,
-    "eval_loss": 1.1249287128448486,
-    "eval_odds_ratio_loss": 0.62950599193573,
-    "eval_rewards/accuracies": 0.4399999976158142,
-    "eval_rewards/chosen": -0.10619781166315079,
-    "eval_rewards/margins": 0.013549459166824818,
-    "eval_rewards/rejected": -0.11974727362394333,
-    "eval_runtime": 57.437,
-    "eval_samples_per_second": 1.741,
-    "eval_sft_loss": 1.0619781017303467,
-    "eval_steps_per_second": 0.435,
-    "total_flos": 2.303993975490478e+17,
-    "train_loss": 1.1586682626179285,
-    "train_runtime": 5352.1206,
-    "train_samples_per_second": 0.504,
-    "train_steps_per_second": 0.031
+    "epoch": 2.997999555456768,
+    "eval_logits/chosen": -2.8460819721221924,
+    "eval_logits/rejected": -2.8720545768737793,
+    "eval_logps/chosen": -0.8352006673812866,
+    "eval_logps/rejected": -1.0736578702926636,
+    "eval_loss": 0.8975116014480591,
+    "eval_odds_ratio_loss": 0.6231085658073425,
+    "eval_rewards/accuracies": 0.5899999737739563,
+    "eval_rewards/chosen": -0.0835200697183609,
+    "eval_rewards/margins": 0.023845719173550606,
+    "eval_rewards/rejected": -0.10736579447984695,
+    "eval_runtime": 189.1831,
+    "eval_samples_per_second": 5.286,
+    "eval_sft_loss": 0.8352006673812866,
+    "eval_steps_per_second": 2.643,
+    "total_flos": 1.985199772705751e+18,
+    "train_loss": 0.9645641791862949,
+    "train_runtime": 17170.3123,
+    "train_samples_per_second": 1.572,
+    "train_steps_per_second": 0.098
 }

eval_results.json

@@ -1,17 +1,17 @@
 {
-    "epoch": 2.986666666666667,
-    "eval_logits/chosen": -2.677677869796753,
-    "eval_logits/rejected": -2.6818528175354004,
-    "eval_logps/chosen": -1.0619781017303467,
-    "eval_logps/rejected": -1.1974726915359497,
-    "eval_loss": 1.1249287128448486,
-    "eval_odds_ratio_loss": 0.62950599193573,
-    "eval_rewards/accuracies": 0.4399999976158142,
-    "eval_rewards/chosen": -0.10619781166315079,
-    "eval_rewards/margins": 0.013549459166824818,
-    "eval_rewards/rejected": -0.11974727362394333,
-    "eval_runtime": 57.437,
-    "eval_samples_per_second": 1.741,
-    "eval_sft_loss": 1.0619781017303467,
-    "eval_steps_per_second": 0.435
+    "epoch": 2.997999555456768,
+    "eval_logits/chosen": -2.8460819721221924,
+    "eval_logits/rejected": -2.8720545768737793,
+    "eval_logps/chosen": -0.8352006673812866,
+    "eval_logps/rejected": -1.0736578702926636,
+    "eval_loss": 0.8975116014480591,
+    "eval_odds_ratio_loss": 0.6231085658073425,
+    "eval_rewards/accuracies": 0.5899999737739563,
+    "eval_rewards/chosen": -0.0835200697183609,
+    "eval_rewards/margins": 0.023845719173550606,
+    "eval_rewards/rejected": -0.10736579447984695,
+    "eval_runtime": 189.1831,
+    "eval_samples_per_second": 5.286,
+    "eval_sft_loss": 0.8352006673812866,
+    "eval_steps_per_second": 2.643
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 2.986666666666667,
-    "total_flos": 2.303993975490478e+17,
-    "train_loss": 1.1586682626179285,
-    "train_runtime": 5352.1206,
-    "train_samples_per_second": 0.504,
-    "train_steps_per_second": 0.031
+    "epoch": 2.997999555456768,
+    "total_flos": 1.985199772705751e+18,
+    "train_loss": 0.9645641791862949,
+    "train_runtime": 17170.3123,
+    "train_samples_per_second": 1.572,
+    "train_steps_per_second": 0.098
 }