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: 0.7391
-- Rewards/chosen: -0.0991
-- Rewards/rejected: -0.2198
-- Rewards/accuracies: 0.5486
-- Rewards/margins: 0.1207
-- Logps/rejected: -36.2805
-- Logps/chosen: -32.5848
-- Logits/rejected: 98.6399
-- Logits/chosen: 98.6637
+- Loss: 0.7418
+- Rewards/chosen: -0.0873
+- Rewards/rejected: -0.1994
+- Rewards/accuracies: 0.5511
+- Rewards/margins: 0.1121
+- Logps/rejected: -36.2514
+- Logps/chosen: -32.5679
+- Logits/rejected: 98.6136
+- Logits/chosen: 98.6370
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": 98.63704681396484,
+    "eval_logits/rejected": 98.61363220214844,
+    "eval_logps/chosen": -32.56790542602539,
+    "eval_logps/rejected": -36.25141525268555,
+    "eval_loss": 0.7417604923248291,
+    "eval_rewards/accuracies": 0.5510797500610352,
+    "eval_rewards/chosen": -0.08731049299240112,
+    "eval_rewards/margins": 0.11211014539003372,
+    "eval_rewards/rejected": -0.19942064583301544,
+    "eval_runtime": 103.7829,
+    "eval_samples": 343,
+    "eval_samples_per_second": 3.305,
+    "eval_steps_per_second": 0.414,
     "train_loss": 0.5948986065852178,
     "train_runtime": 2557.7017,
     "train_samples": 3079,

eval_results.json

@@ -1,15 +1,15 @@
 {
     "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_logits/chosen": 98.63704681396484,
+    "eval_logits/rejected": 98.61363220214844,
+    "eval_logps/chosen": -32.56790542602539,
+    "eval_logps/rejected": -36.25141525268555,
+    "eval_loss": 0.7417604923248291,
+    "eval_rewards/accuracies": 0.5510797500610352,
+    "eval_rewards/chosen": -0.08731049299240112,
+    "eval_rewards/margins": 0.11211014539003372,
+    "eval_rewards/rejected": -0.19942064583301544,
+    "eval_runtime": 103.7829,
     "eval_samples": 343,
     "eval_samples_per_second": 3.305,
     "eval_steps_per_second": 0.414