End of training

README.md

@@ -1,11 +1,13 @@
 ---
-license: apache-2.0
 library_name: peft
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
 base_model: norallm/normistral-7b-warm
+datasets:
+- hugodk-sch/aftonposten_title_prefs
 model-index:
 - name: ap-normistral-7b-align-scan
   results: []
@@ -16,17 +18,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # ap-normistral-7b-align-scan
 
-This model is a fine-tuned version of [norallm/normistral-7b-warm](https://huggingface.co/norallm/normistral-7b-warm) on the None dataset.
+This model is a fine-tuned version of [data/ap-normistral-7b-sft-qlora](https://huggingface.co/data/ap-normistral-7b-sft-qlora) on the hugodk-sch/aftonposten_title_prefs dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.1081
-- Rewards/chosen: -0.0561
-- Rewards/rejected: -0.0772
-- Rewards/accuracies: 0.4776
-- Rewards/margins: 0.0212
-- Logps/rejected: -36.0524
-- Logps/chosen: -32.5055
-- Logits/rejected: 98.7108
-- Logits/chosen: 98.7301
+- Loss: 1.1077
+- Rewards/chosen: -0.0230
+- Rewards/rejected: -0.0718
+- Rewards/accuracies: 0.4988
+- Rewards/margins: 0.0488
+- Logps/rejected: -36.0463
+- Logps/chosen: -32.4687
+- Logits/rejected: 98.7000
+- Logits/chosen: 98.7211
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": 98.72113800048828,
+    "eval_logits/rejected": 98.69998931884766,
+    "eval_logps/chosen": -32.468719482421875,
+    "eval_logps/rejected": -36.04634094238281,
+    "eval_loss": 1.107681393623352,
+    "eval_rewards/accuracies": 0.4987541437149048,
+    "eval_rewards/chosen": -0.022987432777881622,
+    "eval_rewards/margins": 0.04884451627731323,
+    "eval_rewards/rejected": -0.07183194905519485,
+    "eval_runtime": 103.7689,
+    "eval_samples": 343,
+    "eval_samples_per_second": 3.305,
+    "eval_steps_per_second": 0.414,
     "train_loss": 0.7878170496457583,
     "train_runtime": 2553.4781,
     "train_samples": 3079,

eval_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": 97.65992736816406,
-    "eval_logits/rejected": 97.62660217285156,
-    "eval_logps/chosen": -33.290000915527344,
-    "eval_logps/rejected": -37.37261199951172,
-    "eval_loss": 0.994398832321167,
-    "eval_rewards/accuracies": 0.6004983186721802,
-    "eval_rewards/chosen": -0.008468217216432095,
-    "eval_rewards/margins": 0.005592645611613989,
-    "eval_rewards/rejected": -0.014060864225029945,
-    "eval_runtime": 103.8235,
+    "eval_logits/chosen": 98.72113800048828,
+    "eval_logits/rejected": 98.69998931884766,
+    "eval_logps/chosen": -32.468719482421875,
+    "eval_logps/rejected": -36.04634094238281,
+    "eval_loss": 1.107681393623352,
+    "eval_rewards/accuracies": 0.4987541437149048,
+    "eval_rewards/chosen": -0.022987432777881622,
+    "eval_rewards/margins": 0.04884451627731323,
+    "eval_rewards/rejected": -0.07183194905519485,
+    "eval_runtime": 103.7689,
     "eval_samples": 343,
-    "eval_samples_per_second": 3.304,
+    "eval_samples_per_second": 3.305,
     "eval_steps_per_second": 0.414
 }