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 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: 0.2490
-- Rewards/chosen: -4.6026
-- Rewards/rejected: -9.7319
-- Rewards/accuracies: 0.8463
-- Rewards/margins: 5.1294
-- Logps/rejected: -1276.6984
-- Logps/chosen: -751.7161
-- Logits/rejected: -1.1260
-- Logits/chosen: -1.5108
+- Loss: 0.2487
+- Rewards/chosen: -4.9500
+- Rewards/rejected: -10.6858
+- Rewards/accuracies: 0.8533
+- Rewards/margins: 5.7358
+- Logps/rejected: -1372.0834
+- Logps/chosen: -786.4601
+- Logits/rejected: -1.1251
+- Logits/chosen: -1.5758
 
 ## Model description
 

all_results.json

@@ -1,17 +1,17 @@
 {
     "epoch": 0.9980806142034548,
-    "eval_logits/chosen": -1.7157304286956787,
-    "eval_logits/rejected": -1.2709392309188843,
-    "eval_logps/chosen": -759.6055297851562,
-    "eval_logps/rejected": -1354.8134765625,
-    "eval_loss": 0.24732786417007446,
-    "eval_rewards/accuracies": 0.8524844646453857,
-    "eval_rewards/chosen": -4.681485176086426,
-    "eval_rewards/margins": 5.831614017486572,
-    "eval_rewards/rejected": -10.513099670410156,
-    "eval_runtime": 461.8444,
+    "eval_logits/chosen": -1.575805425643921,
+    "eval_logits/rejected": -1.1250745058059692,
+    "eval_logps/chosen": -786.4600830078125,
+    "eval_logps/rejected": -1372.0833740234375,
+    "eval_loss": 0.2487475574016571,
+    "eval_rewards/accuracies": 0.85326087474823,
+    "eval_rewards/chosen": -4.950030326843262,
+    "eval_rewards/margins": 5.735767841339111,
+    "eval_rewards/rejected": -10.685796737670898,
+    "eval_runtime": 461.9104,
     "eval_samples": 5126,
-    "eval_samples_per_second": 11.099,
+    "eval_samples_per_second": 11.097,
     "eval_steps_per_second": 0.349,
     "total_flos": 0.0,
     "train_loss": 0.32177775089557353,

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": 0.9980806142034548,
-    "eval_logits/chosen": -1.7157304286956787,
-    "eval_logits/rejected": -1.2709392309188843,
-    "eval_logps/chosen": -759.6055297851562,
-    "eval_logps/rejected": -1354.8134765625,
-    "eval_loss": 0.24732786417007446,
-    "eval_rewards/accuracies": 0.8524844646453857,
-    "eval_rewards/chosen": -4.681485176086426,
-    "eval_rewards/margins": 5.831614017486572,
-    "eval_rewards/rejected": -10.513099670410156,
-    "eval_runtime": 461.8444,
+    "eval_logits/chosen": -1.575805425643921,
+    "eval_logits/rejected": -1.1250745058059692,
+    "eval_logps/chosen": -786.4600830078125,
+    "eval_logps/rejected": -1372.0833740234375,
+    "eval_loss": 0.2487475574016571,
+    "eval_rewards/accuracies": 0.85326087474823,
+    "eval_rewards/chosen": -4.950030326843262,
+    "eval_rewards/margins": 5.735767841339111,
+    "eval_rewards/rejected": -10.685796737670898,
+    "eval_runtime": 461.9104,
     "eval_samples": 5126,
-    "eval_samples_per_second": 11.099,
+    "eval_samples_per_second": 11.097,
     "eval_steps_per_second": 0.349
 }

runs/Aug25_02-05-26_ip-10-0-9-154.ec2.internal/events.out.tfevents.1724566667.ip-10-0-9-154.ec2.internal.56487.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15ea715dfb6844de75cf024d35d29050886eef742177bb8b2f8fbd47b1d2785e
+size 828