End of training

README.md

@@ -2,10 +2,16 @@
 license: other
 base_model: HuggingFaceH4/zephyr-7b-gemma-sft-v0.1
 tags:
+- alignment-handbook
+- trl
+- dpo
+- generated_from_trainer
 - trl
 - dpo
 - alignment-handbook
 - generated_from_trainer
+datasets:
+- argilla/dpo-mix-7k
 model-index:
 - name: zephyr-7b-gemma-dpo
   results: []
@@ -16,17 +22,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-7b-gemma-dpo
 
-This model is a fine-tuned version of [HuggingFaceH4/zephyr-7b-gemma-sft-v0.1](https://huggingface.co/HuggingFaceH4/zephyr-7b-gemma-sft-v0.1) on an unknown dataset.
+This model is a fine-tuned version of [HuggingFaceH4/zephyr-7b-gemma-sft-v0.1](https://huggingface.co/HuggingFaceH4/zephyr-7b-gemma-sft-v0.1) on the argilla/dpo-mix-7k dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.4643
-- Rewards/chosen: -3.5909
-- Rewards/rejected: -5.3391
-- Rewards/accuracies: 0.75
-- Rewards/margins: 1.7481
-- Logps/rejected: -515.7638
-- Logps/chosen: -428.1683
-- Logits/rejected: 94.0722
-- Logits/chosen: 91.3541
+- Loss: 0.4567
+- Rewards/chosen: -3.5881
+- Rewards/rejected: -5.3433
+- Rewards/accuracies: 0.7660
+- Rewards/margins: 1.7552
+- Logps/rejected: -515.8483
+- Logps/chosen: -428.1110
+- Logits/rejected: 94.0535
+- Logits/chosen: 91.3372
 
 ## Model description
 

all_results.json

@@ -1,18 +1,18 @@
 {
     "epoch": 1.971563981042654,
-    "eval_logits/chosen": 96.71578216552734,
-    "eval_logits/rejected": 90.98221588134766,
-    "eval_logps/chosen": -423.6227722167969,
-    "eval_logps/rejected": -453.7782287597656,
-    "eval_loss": 0.468290776014328,
-    "eval_rewards/accuracies": 0.7708333134651184,
-    "eval_rewards/chosen": -3.0221338272094727,
-    "eval_rewards/margins": 1.6591955423355103,
-    "eval_rewards/rejected": -4.681329727172852,
-    "eval_runtime": 58.6185,
+    "eval_logits/chosen": 91.33722686767578,
+    "eval_logits/rejected": 94.05353546142578,
+    "eval_logps/chosen": -428.1109619140625,
+    "eval_logps/rejected": -515.8482666015625,
+    "eval_loss": 0.4566638171672821,
+    "eval_rewards/accuracies": 0.7659574747085571,
+    "eval_rewards/chosen": -3.5880744457244873,
+    "eval_rewards/margins": 1.7552103996276855,
+    "eval_rewards/rejected": -5.343284606933594,
+    "eval_runtime": 90.1839,
     "eval_samples": 750,
-    "eval_samples_per_second": 12.795,
-    "eval_steps_per_second": 0.409,
+    "eval_samples_per_second": 8.316,
+    "eval_steps_per_second": 0.521,
     "total_flos": 0.0,
     "train_loss": 0.3883641087091886,
     "train_runtime": 2802.2739,

config.json

@@ -24,6 +24,6 @@
   "rope_theta": 10000.0,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.40.1",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 256000
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 1.971563981042654,
-    "eval_logits/chosen": 96.71578216552734,
-    "eval_logits/rejected": 90.98221588134766,
-    "eval_logps/chosen": -423.6227722167969,
-    "eval_logps/rejected": -453.7782287597656,
-    "eval_loss": 0.468290776014328,
-    "eval_rewards/accuracies": 0.7708333134651184,
-    "eval_rewards/chosen": -3.0221338272094727,
-    "eval_rewards/margins": 1.6591955423355103,
-    "eval_rewards/rejected": -4.681329727172852,
-    "eval_runtime": 58.6185,
+    "eval_logits/chosen": 91.33722686767578,
+    "eval_logits/rejected": 94.05353546142578,
+    "eval_logps/chosen": -428.1109619140625,
+    "eval_logps/rejected": -515.8482666015625,
+    "eval_loss": 0.4566638171672821,
+    "eval_rewards/accuracies": 0.7659574747085571,
+    "eval_rewards/chosen": -3.5880744457244873,
+    "eval_rewards/margins": 1.7552103996276855,
+    "eval_rewards/rejected": -5.343284606933594,
+    "eval_runtime": 90.1839,
     "eval_samples": 750,
-    "eval_samples_per_second": 12.795,
-    "eval_steps_per_second": 0.409
+    "eval_samples_per_second": 8.316,
+    "eval_steps_per_second": 0.521
 }

runs/Apr29_18-08-25_gcp002/events.out.tfevents.1714417377.gcp002.13104.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca73e478eff34cd45b2044bec802df9dd3bcb6a171f83d71ae00847e0121e856
+size 815