End of training

README.md

@@ -2,9 +2,15 @@
 library_name: transformers
 base_model: allenai/tulu-2-7b
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: ultrafeedback-binarized-tulu-2-7b-dpo-full
   results: []
@@ -15,17 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # ultrafeedback-binarized-tulu-2-7b-dpo-full
 
-This model is a fine-tuned version of [allenai/tulu-2-7b](https://huggingface.co/allenai/tulu-2-7b) on an unknown dataset.
+This model is a fine-tuned version of [allenai/tulu-2-7b](https://huggingface.co/allenai/tulu-2-7b) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.6640
-- Rewards/chosen: 0.0423
-- Rewards/rejected: -0.0311
-- Rewards/accuracies: 0.6706
-- Rewards/margins: 0.0734
-- Logps/rejected: -317.2082
-- Logps/chosen: -335.1042
-- Logits/rejected: -1.2523
-- Logits/chosen: -1.1794
+- Loss: 0.6636
+- Rewards/chosen: 0.0407
+- Rewards/rejected: -0.0326
+- Rewards/accuracies: 0.6746
+- Rewards/margins: 0.0733
+- Logps/rejected: -317.3561
+- Logps/chosen: -335.2600
+- Logits/rejected: -1.2511
+- Logits/chosen: -1.1780
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -1.1779704093933105,
+    "eval_logits/rejected": -1.2511389255523682,
+    "eval_logps/chosen": -335.2599792480469,
+    "eval_logps/rejected": -317.35614013671875,
+    "eval_loss": 0.6636306643486023,
+    "eval_rewards/accuracies": 0.6746031641960144,
+    "eval_rewards/chosen": 0.04071044921875,
+    "eval_rewards/margins": 0.0732826218008995,
+    "eval_rewards/rejected": -0.032572176307439804,
+    "eval_runtime": 91.7551,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 21.797,
+    "eval_steps_per_second": 0.687,
     "total_flos": 0.0,
     "train_loss": 0.6732059223382543,
     "train_runtime": 6951.1203,

config.json

@@ -24,6 +24,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.44.1",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 1.0,
+    "eval_logits/chosen": -1.1779704093933105,
+    "eval_logits/rejected": -1.2511389255523682,
+    "eval_logps/chosen": -335.2599792480469,
+    "eval_logps/rejected": -317.35614013671875,
+    "eval_loss": 0.6636306643486023,
+    "eval_rewards/accuracies": 0.6746031641960144,
+    "eval_rewards/chosen": 0.04071044921875,
+    "eval_rewards/margins": 0.0732826218008995,
+    "eval_rewards/rejected": -0.032572176307439804,
+    "eval_runtime": 91.7551,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 21.797,
+    "eval_steps_per_second": 0.687
+}