End of training

README.md

@@ -3,10 +3,16 @@ library_name: transformers
 license: apache-2.0
 base_model: alignment-handbook/zephyr-7b-sft-full
 tags:
+- alignment-handbook
+- trl
+- dpo
+- generated_from_trainer
 - trl
 - dpo
 - alignment-handbook
 - generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: zephyr-7b-dpo-full
   results: []
@@ -17,17 +23,17 @@ 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: 0.3161
-- Rewards/chosen: -4.1174
-- Rewards/rejected: -8.3387
-- Rewards/accuracies: 0.8212
-- Rewards/margins: 4.2213
-- Logps/rejected: -1123.9625
-- Logps/chosen: -702.7068
-- Logits/rejected: 0.5558
-- Logits/chosen: -0.4246
+- Loss: 0.3160
+- Rewards/chosen: -4.1121
+- Rewards/rejected: -8.3353
+- Rewards/accuracies: 0.8201
+- Rewards/margins: 4.2232
+- Logps/rejected: -1123.6224
+- Logps/chosen: -702.1752
+- Logits/rejected: 0.5403
+- Logits/chosen: -0.4404
 
 ## Model description
 

all_results.json

@@ -1,17 +1,17 @@
 {
     "epoch": 0.9997120644975526,
-    "eval_logits/chosen": -0.29694831371307373,
-    "eval_logits/rejected": 0.6135479807853699,
-    "eval_logps/chosen": -754.9666748046875,
-    "eval_logps/rejected": -1189.8031005859375,
-    "eval_loss": 0.3148016333580017,
-    "eval_rewards/accuracies": 0.8178251385688782,
-    "eval_rewards/chosen": -4.639986038208008,
-    "eval_rewards/margins": 4.357123374938965,
-    "eval_rewards/rejected": -8.997109413146973,
-    "eval_runtime": 645.0187,
+    "eval_logits/chosen": -0.44044333696365356,
+    "eval_logits/rejected": 0.5402784943580627,
+    "eval_logps/chosen": -702.1751708984375,
+    "eval_logps/rejected": -1123.6224365234375,
+    "eval_loss": 0.3159700930118561,
+    "eval_rewards/accuracies": 0.820067286491394,
+    "eval_rewards/chosen": -4.112070560455322,
+    "eval_rewards/margins": 4.223231792449951,
+    "eval_rewards/rejected": -8.335301399230957,
+    "eval_runtime": 644.1576,
     "eval_samples": 7126,
-    "eval_samples_per_second": 11.048,
+    "eval_samples_per_second": 11.063,
     "eval_steps_per_second": 0.346,
     "total_flos": 0.0,
     "train_loss": 0.42924998652550483,

config.json

@@ -22,6 +22,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.44.1",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 0.9997120644975526,
-    "eval_logits/chosen": -0.29694831371307373,
-    "eval_logits/rejected": 0.6135479807853699,
-    "eval_logps/chosen": -754.9666748046875,
-    "eval_logps/rejected": -1189.8031005859375,
-    "eval_loss": 0.3148016333580017,
-    "eval_rewards/accuracies": 0.8178251385688782,
-    "eval_rewards/chosen": -4.639986038208008,
-    "eval_rewards/margins": 4.357123374938965,
-    "eval_rewards/rejected": -8.997109413146973,
-    "eval_runtime": 645.0187,
+    "eval_logits/chosen": -0.44044333696365356,
+    "eval_logits/rejected": 0.5402784943580627,
+    "eval_logps/chosen": -702.1751708984375,
+    "eval_logps/rejected": -1123.6224365234375,
+    "eval_loss": 0.3159700930118561,
+    "eval_rewards/accuracies": 0.820067286491394,
+    "eval_rewards/chosen": -4.112070560455322,
+    "eval_rewards/margins": 4.223231792449951,
+    "eval_rewards/rejected": -8.335301399230957,
+    "eval_runtime": 644.1576,
     "eval_samples": 7126,
-    "eval_samples_per_second": 11.048,
+    "eval_samples_per_second": 11.063,
     "eval_steps_per_second": 0.346
 }

runs/Aug25_02-05-30_ip-10-0-9-154.ec2.internal/events.out.tfevents.1724586855.ip-10-0-9-154.ec2.internal.57471.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cb2be816a85681f807d39cfb9c335289e7b968edc0c80bde20d9f847775f79e
+size 828