End of training

README.md

@@ -2,9 +2,15 @@
 license: other
 base_model: lewtun/gemma-7b-sft-full-ultrachat-v0
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/orca_dpo_pairs
 model-index:
 - name: gemma-7b-dpo-full-orca-v0
   results: []
@@ -15,7 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # gemma-7b-dpo-full-orca-v0
 
-This model is a fine-tuned version of [lewtun/gemma-7b-sft-full-ultrachat-v0](https://huggingface.co/lewtun/gemma-7b-sft-full-ultrachat-v0) on the None dataset.
+This model is a fine-tuned version of [lewtun/gemma-7b-sft-full-ultrachat-v0](https://huggingface.co/lewtun/gemma-7b-sft-full-ultrachat-v0) on the HuggingFaceH4/orca_dpo_pairs dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0131
+- Rewards/chosen: 4.5525
+- Rewards/rejected: -7.7149
+- Rewards/accuracies: 0.9922
+- Rewards/margins: 12.2674
+- Logps/rejected: -860.7157
+- Logps/chosen: -725.1588
+- Logits/rejected: 141.1811
+- Logits/chosen: 94.1054
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.99,
+    "eval_logits/chosen": 94.10542297363281,
+    "eval_logits/rejected": 141.18112182617188,
+    "eval_logps/chosen": -725.1587524414062,
+    "eval_logps/rejected": -860.7156982421875,
+    "eval_loss": 0.013092889450490475,
+    "eval_rewards/accuracies": 0.9921875,
+    "eval_rewards/chosen": 4.552516937255859,
+    "eval_rewards/margins": 12.267425537109375,
+    "eval_rewards/rejected": -7.714908599853516,
+    "eval_runtime": 44.617,
+    "eval_samples": 1000,
+    "eval_samples_per_second": 22.413,
+    "eval_steps_per_second": 0.717,
     "train_loss": 0.0926785598124337,
     "train_runtime": 1161.69,
     "train_samples": 11859,

config.json

@@ -23,6 +23,6 @@
   "rope_theta": 10000.0,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.39.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 256000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 0.99,
+    "eval_logits/chosen": 94.10542297363281,
+    "eval_logits/rejected": 141.18112182617188,
+    "eval_logps/chosen": -725.1587524414062,
+    "eval_logps/rejected": -860.7156982421875,
+    "eval_loss": 0.013092889450490475,
+    "eval_rewards/accuracies": 0.9921875,
+    "eval_rewards/chosen": 4.552516937255859,
+    "eval_rewards/margins": 12.267425537109375,
+    "eval_rewards/rejected": -7.714908599853516,
+    "eval_runtime": 44.617,
+    "eval_samples": 1000,
+    "eval_samples_per_second": 22.413,
+    "eval_steps_per_second": 0.717
+}

runs/Feb29_16-29-02_ip-26-0-167-9/events.out.tfevents.1709225527.ip-26-0-167-9.3787591.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0539dd01e543956e118b75520b9e8f0e190e2f3d45857723b88e3d867200496b
+size 815