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: 0.0398
-- Rewards/chosen: -30.6169
-- Rewards/rejected: -34.6409
+- Loss: 0.0393
+- Rewards/chosen: -30.8792
+- Rewards/rejected: -34.9336
 - Rewards/accuracies: 1.0
-- Rewards/margins: 4.0239
-- Logps/rejected: -955.4847
-- Logps/chosen: -942.9697
-- Logits/rejected: -1.7039
-- Logits/chosen: -1.7576
+- Rewards/margins: 4.0543
+- Logps/rejected: -961.3383
+- Logps/chosen: -948.7985
+- Logits/rejected: -1.6730
+- Logits/chosen: -1.7277
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -1.7276861667633057,
+    "eval_logits/rejected": -1.6730353832244873,
+    "eval_logps/chosen": -948.7984619140625,
+    "eval_logps/rejected": -961.3382568359375,
+    "eval_loss": 0.03925349563360214,
+    "eval_rewards/accuracies": 1.0,
+    "eval_rewards/chosen": -30.879230499267578,
+    "eval_rewards/margins": 4.054325103759766,
+    "eval_rewards/rejected": -34.933555603027344,
+    "eval_runtime": 96.1529,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 20.8,
+    "eval_steps_per_second": 0.333,
     "train_loss": 0.14747725651603363,
     "train_runtime": 7551.087,
     "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": -2.7366418838500977,
-    "eval_logits/rejected": -2.709143877029419,
-    "eval_logps/chosen": -241.407470703125,
-    "eval_logps/rejected": -256.4153747558594,
-    "eval_loss": 0.4928562045097351,
-    "eval_rewards/accuracies": 0.734375,
-    "eval_rewards/chosen": 21.185983657836914,
-    "eval_rewards/margins": 14.934151649475098,
-    "eval_rewards/rejected": 6.251833915710449,
-    "eval_runtime": 96.1252,
+    "eval_logits/chosen": -1.7276861667633057,
+    "eval_logits/rejected": -1.6730353832244873,
+    "eval_logps/chosen": -948.7984619140625,
+    "eval_logps/rejected": -961.3382568359375,
+    "eval_loss": 0.03925349563360214,
+    "eval_rewards/accuracies": 1.0,
+    "eval_rewards/chosen": -30.879230499267578,
+    "eval_rewards/margins": 4.054325103759766,
+    "eval_rewards/rejected": -34.933555603027344,
+    "eval_runtime": 96.1529,
     "eval_samples": 2000,
-    "eval_samples_per_second": 20.806,
+    "eval_samples_per_second": 20.8,
     "eval_steps_per_second": 0.333
 }