End of training

README.md

@@ -2,9 +2,9 @@
 base_model: princeton-nlp/Llama-3-Base-8B-SFT
 library_name: peft
 tags:
+- alignment-handbook
 - trl
 - dpo
-- alignment-handbook
 - generated_from_trainer
 model-index:
 - name: llama3-wpo-lora
@@ -18,16 +18,16 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [princeton-nlp/Llama-3-Base-8B-SFT](https://huggingface.co/princeton-nlp/Llama-3-Base-8B-SFT) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.5175
-- Rewards/chosen: -0.0517
-- Rewards/rejected: -0.9092
-- Rewards/accuracies: 0.7280
-- Rewards/margins: 0.8575
-- Logps/rejected: -285.7691
-- Logps/chosen: -293.0645
+- Loss: 0.5172
+- Rewards/chosen: -0.0494
+- Rewards/rejected: -0.9056
+- Rewards/accuracies: 0.7300
+- Rewards/margins: 0.8562
+- Logps/rejected: -285.7321
+- Logps/chosen: -293.0410
 - Logps/ref Response: -0.5364
-- Logits/rejected: -0.3072
-- Logits/chosen: -0.3443
+- Logits/rejected: -0.3074
+- Logits/chosen: -0.3445
 
 ## Model description
 

all_results.json

@@ -1,19 +1,19 @@
 {
     "epoch": 0.9997382884061764,
-    "eval_logits/chosen": -0.38019832968711853,
-    "eval_logits/rejected": -0.3463886082172394,
-    "eval_logps/chosen": -292.6591796875,
+    "eval_logits/chosen": -0.34449103474617004,
+    "eval_logits/rejected": -0.3073786199092865,
+    "eval_logps/chosen": -293.0410461425781,
     "eval_logps/ref_response": -0.5363935232162476,
-    "eval_logps/rejected": -285.1923828125,
-    "eval_loss": 0.5184887647628784,
-    "eval_rewards/accuracies": 0.7279999852180481,
-    "eval_rewards/chosen": -0.011204298585653305,
-    "eval_rewards/margins": 0.8403744101524353,
-    "eval_rewards/rejected": -0.8515786528587341,
-    "eval_runtime": 348.8884,
+    "eval_logps/rejected": -285.7321472167969,
+    "eval_loss": 0.5171638131141663,
+    "eval_rewards/accuracies": 0.7300000190734863,
+    "eval_rewards/chosen": -0.04938787221908569,
+    "eval_rewards/margins": 0.8561664819717407,
+    "eval_rewards/rejected": -0.9055544137954712,
+    "eval_runtime": 348.587,
     "eval_samples": 2000,
-    "eval_samples_per_second": 5.732,
-    "eval_steps_per_second": 0.358,
+    "eval_samples_per_second": 5.737,
+    "eval_steps_per_second": 0.359,
     "total_flos": 0.0,
     "train_loss": 0.5383632463934533,
     "train_runtime": 19109.9128,

eval_results.json

@@ -1,17 +1,17 @@
 {
     "epoch": 0.9997382884061764,
-    "eval_logits/chosen": -0.38019832968711853,
-    "eval_logits/rejected": -0.3463886082172394,
-    "eval_logps/chosen": -292.6591796875,
+    "eval_logits/chosen": -0.34449103474617004,
+    "eval_logits/rejected": -0.3073786199092865,
+    "eval_logps/chosen": -293.0410461425781,
     "eval_logps/ref_response": -0.5363935232162476,
-    "eval_logps/rejected": -285.1923828125,
-    "eval_loss": 0.5184887647628784,
-    "eval_rewards/accuracies": 0.7279999852180481,
-    "eval_rewards/chosen": -0.011204298585653305,
-    "eval_rewards/margins": 0.8403744101524353,
-    "eval_rewards/rejected": -0.8515786528587341,
-    "eval_runtime": 348.8884,
+    "eval_logps/rejected": -285.7321472167969,
+    "eval_loss": 0.5171638131141663,
+    "eval_rewards/accuracies": 0.7300000190734863,
+    "eval_rewards/chosen": -0.04938787221908569,
+    "eval_rewards/margins": 0.8561664819717407,
+    "eval_rewards/rejected": -0.9055544137954712,
+    "eval_runtime": 348.587,
     "eval_samples": 2000,
-    "eval_samples_per_second": 5.732,
-    "eval_steps_per_second": 0.358
+    "eval_samples_per_second": 5.737,
+    "eval_steps_per_second": 0.359
 }