End of training

README.md

@@ -2,9 +2,15 @@
 license: apache-2.0
 base_model: alignment-handbook/zephyr-7b-sft-full
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: zephyr-7b-dpo-full
   results: []
@@ -15,17 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-7b-dpo-full
 
-This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
+This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.2788
-- Rewards/chosen: 31.8547
-- Rewards/rejected: 27.4761
-- Rewards/accuracies: 0.5781
-- Rewards/margins: 4.3786
-- Logps/rejected: -235.1911
-- Logps/chosen: -230.7387
-- Logits/rejected: -2.6619
-- Logits/chosen: -2.6869
+- Loss: 0.2784
+- Rewards/chosen: 31.9054
+- Rewards/rejected: 27.5130
+- Rewards/accuracies: 0.5742
+- Rewards/margins: 4.3924
+- Logps/rejected: -235.1542
+- Logps/chosen: -230.6881
+- Logits/rejected: -2.6646
+- Logits/chosen: -2.6896
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -2.689567804336548,
+    "eval_logits/rejected": -2.664588451385498,
+    "eval_logps/chosen": -230.68809509277344,
+    "eval_logps/rejected": -235.1542205810547,
+    "eval_loss": 0.27835845947265625,
+    "eval_rewards/accuracies": 0.57421875,
+    "eval_rewards/chosen": 31.905372619628906,
+    "eval_rewards/margins": 4.392405033111572,
+    "eval_rewards/rejected": 27.51296615600586,
+    "eval_runtime": 96.4814,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 20.729,
+    "eval_steps_per_second": 0.332,
     "train_loss": 0.31381568898715734,
     "train_runtime": 7749.4814,
     "train_samples": 61134,

config.json

@@ -21,6 +21,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.38.2",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 1.0,
-    "eval_logits/chosen": -2.5251381397247314,
-    "eval_logits/rejected": -2.4836020469665527,
-    "eval_logps/chosen": -266.3936767578125,
-    "eval_logps/rejected": -288.7183532714844,
-    "eval_loss": 0.5105125308036804,
-    "eval_rewards/accuracies": 0.78125,
-    "eval_rewards/chosen": -0.190010204911232,
-    "eval_rewards/margins": 1.112548828125,
-    "eval_rewards/rejected": -1.3025591373443604,
-    "eval_runtime": 96.1755,
+    "eval_logits/chosen": -2.689567804336548,
+    "eval_logits/rejected": -2.664588451385498,
+    "eval_logps/chosen": -230.68809509277344,
+    "eval_logps/rejected": -235.1542205810547,
+    "eval_loss": 0.27835845947265625,
+    "eval_rewards/accuracies": 0.57421875,
+    "eval_rewards/chosen": 31.905372619628906,
+    "eval_rewards/margins": 4.392405033111572,
+    "eval_rewards/rejected": 27.51296615600586,
+    "eval_runtime": 96.4814,
     "eval_samples": 2000,
-    "eval_samples_per_second": 20.795,
-    "eval_steps_per_second": 0.333
+    "eval_samples_per_second": 20.729,
+    "eval_steps_per_second": 0.332
 }