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-ultrabin-high-bleu
   results: []
@@ -15,17 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-7b-dpo-full-ultrabin-high-bleu
 
-This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on an unknown 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: 0.6329
-- Rewards/chosen: -0.2639
-- Rewards/rejected: -0.4887
-- Rewards/accuracies: 0.6875
-- Rewards/margins: 0.2248
-- Logps/rejected: -311.5294
-- Logps/chosen: -289.0153
-- Logits/rejected: -2.2959
-- Logits/chosen: -2.3456
+- Loss: 0.6282
+- Rewards/chosen: -0.2999
+- Rewards/rejected: -0.5424
+- Rewards/accuracies: 0.7031
+- Rewards/margins: 0.2425
+- Logps/rejected: -316.9008
+- Logps/chosen: -292.6180
+- Logits/rejected: -2.2318
+- Logits/chosen: -2.2813
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.9965156794425087,
+    "eval_logits/chosen": -2.281348943710327,
+    "eval_logits/rejected": -2.2318127155303955,
+    "eval_logps/chosen": -292.6180114746094,
+    "eval_logps/rejected": -316.9007568359375,
+    "eval_loss": 0.6281932592391968,
+    "eval_rewards/accuracies": 0.703125,
+    "eval_rewards/chosen": -0.2998806834220886,
+    "eval_rewards/margins": 0.2425060272216797,
+    "eval_rewards/rejected": -0.5423867106437683,
+    "eval_runtime": 104.9019,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 19.065,
+    "eval_steps_per_second": 0.305,
     "total_flos": 0.0,
     "train_loss": 0.6509068312344851,
     "train_runtime": 3880.8302,

config.json

@@ -22,6 +22,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.44.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 0.9965156794425087,
+    "eval_logits/chosen": -2.281348943710327,
+    "eval_logits/rejected": -2.2318127155303955,
+    "eval_logps/chosen": -292.6180114746094,
+    "eval_logps/rejected": -316.9007568359375,
+    "eval_loss": 0.6281932592391968,
+    "eval_rewards/accuracies": 0.703125,
+    "eval_rewards/chosen": -0.2998806834220886,
+    "eval_rewards/margins": 0.2425060272216797,
+    "eval_rewards/rejected": -0.5423867106437683,
+    "eval_runtime": 104.9019,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 19.065,
+    "eval_steps_per_second": 0.305
+}