End of training

README.md

@@ -2,9 +2,15 @@
 license: other
 base_model: lewtun/gemma-7b-sft-full-deita-10k-v0
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- argilla/dpo-mix-7k
 model-index:
 - name: gemma-7b-dpo-full-mix1-beta-0.05
   results: []
@@ -15,7 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # gemma-7b-dpo-full-mix1-beta-0.05
 
-This model is a fine-tuned version of [lewtun/gemma-7b-sft-full-deita-10k-v0](https://huggingface.co/lewtun/gemma-7b-sft-full-deita-10k-v0) on the None dataset.
+This model is a fine-tuned version of [lewtun/gemma-7b-sft-full-deita-10k-v0](https://huggingface.co/lewtun/gemma-7b-sft-full-deita-10k-v0) on the argilla/dpo-mix-7k dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.5058
+- Rewards/chosen: -1.1605
+- Rewards/rejected: -2.1128
+- Rewards/accuracies: 0.7083
+- Rewards/margins: 0.9523
+- Logps/rejected: -493.8050
+- Logps/chosen: -476.6144
+- Logits/rejected: 100.2884
+- Logits/chosen: 106.6217
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.99,
+    "eval_logits/chosen": 106.62171173095703,
+    "eval_logits/rejected": 100.2883529663086,
+    "eval_logps/chosen": -476.6144104003906,
+    "eval_logps/rejected": -493.8050231933594,
+    "eval_loss": 0.5058174729347229,
+    "eval_rewards/accuracies": 0.7083333134651184,
+    "eval_rewards/chosen": -1.1604968309402466,
+    "eval_rewards/margins": 0.9522747993469238,
+    "eval_rewards/rejected": -2.11277174949646,
+    "eval_runtime": 51.3345,
+    "eval_samples": 750,
+    "eval_samples_per_second": 14.61,
+    "eval_steps_per_second": 0.468,
     "train_loss": 0.5840168778712933,
     "train_runtime": 572.7421,
     "train_samples": 6750,

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": 106.62171173095703,
+    "eval_logits/rejected": 100.2883529663086,
+    "eval_logps/chosen": -476.6144104003906,
+    "eval_logps/rejected": -493.8050231933594,
+    "eval_loss": 0.5058174729347229,
+    "eval_rewards/accuracies": 0.7083333134651184,
+    "eval_rewards/chosen": -1.1604968309402466,
+    "eval_rewards/margins": 0.9522747993469238,
+    "eval_rewards/rejected": -2.11277174949646,
+    "eval_runtime": 51.3345,
+    "eval_samples": 750,
+    "eval_samples_per_second": 14.61,
+    "eval_steps_per_second": 0.468
+}