End of training

README.md

@@ -3,10 +3,17 @@ library_name: transformers
 license: llama3.2
 base_model: tanliboy/llama-3.2-3b-sft
 tags:
+- alignment-handbook
+- trl
+- dpo
+- generated_from_trainer
 - trl
 - dpo
 - alignment-handbook
 - generated_from_trainer
+datasets:
+- HuggingFaceH4/orca_dpo_pairs
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: llama-3.2-3b-dpo
   results: []
@@ -17,17 +24,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # llama-3.2-3b-dpo
 
-This model is a fine-tuned version of [tanliboy/llama-3.2-3b-sft](https://huggingface.co/tanliboy/llama-3.2-3b-sft) on the None dataset.
+This model is a fine-tuned version of [tanliboy/llama-3.2-3b-sft](https://huggingface.co/tanliboy/llama-3.2-3b-sft) on the HuggingFaceH4/orca_dpo_pairs and the HuggingFaceH4/ultrafeedback_binarized datasets.
 It achieves the following results on the evaluation set:
-- Loss: 0.6284
-- Rewards/chosen: 0.8504
-- Rewards/rejected: -3.7058
-- Rewards/accuracies: 0.7437
-- Rewards/margins: 4.5562
-- Logps/rejected: -368.9125
-- Logps/chosen: -337.3143
-- Logits/rejected: 0.4571
-- Logits/chosen: 0.3820
+- Loss: 0.6289
+- Rewards/chosen: 0.7479
+- Rewards/rejected: -3.8379
+- Rewards/accuracies: 0.7405
+- Rewards/margins: 4.5857
+- Logps/rejected: -370.2327
+- Logps/chosen: -338.3392
+- Logits/rejected: 0.4475
+- Logits/chosen: 0.3731
 
 ## Model description
 

all_results.json

@@ -1,18 +1,18 @@
 {
     "epoch": 3.0,
-    "eval_logits/chosen": 0.3599458634853363,
-    "eval_logits/rejected": 0.514390230178833,
-    "eval_logps/chosen": -483.24920654296875,
-    "eval_logps/rejected": -561.0516357421875,
-    "eval_loss": 0.4862797260284424,
-    "eval_rewards/accuracies": 0.7215189933776855,
-    "eval_rewards/chosen": -1.453184962272644,
-    "eval_rewards/margins": 1.1619733572006226,
-    "eval_rewards/rejected": -2.6151583194732666,
-    "eval_runtime": 56.3308,
+    "eval_logits/chosen": 0.3730663061141968,
+    "eval_logits/rejected": 0.4475269019603729,
+    "eval_logps/chosen": -338.3392028808594,
+    "eval_logps/rejected": -370.232666015625,
+    "eval_loss": 0.6289177536964417,
+    "eval_rewards/accuracies": 0.7405063509941101,
+    "eval_rewards/chosen": 0.7478683590888977,
+    "eval_rewards/margins": 4.585729122161865,
+    "eval_rewards/rejected": -3.8378612995147705,
+    "eval_runtime": 70.1775,
     "eval_samples": 2500,
-    "eval_samples_per_second": 44.381,
-    "eval_steps_per_second": 1.402,
+    "eval_samples_per_second": 35.624,
+    "eval_steps_per_second": 1.126,
     "total_flos": 0.0,
     "train_loss": 0.5009220597491634,
     "train_runtime": 6227.6413,

config.json

@@ -35,6 +35,6 @@
   "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.44.2",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 128256
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 0.999564649542882,
-    "eval_logits/chosen": 0.3599458634853363,
-    "eval_logits/rejected": 0.514390230178833,
-    "eval_logps/chosen": -483.24920654296875,
-    "eval_logps/rejected": -561.0516357421875,
-    "eval_loss": 0.4862797260284424,
-    "eval_rewards/accuracies": 0.7215189933776855,
-    "eval_rewards/chosen": -1.453184962272644,
-    "eval_rewards/margins": 1.1619733572006226,
-    "eval_rewards/rejected": -2.6151583194732666,
-    "eval_runtime": 56.3308,
+    "epoch": 3.0,
+    "eval_logits/chosen": 0.3730663061141968,
+    "eval_logits/rejected": 0.4475269019603729,
+    "eval_logps/chosen": -338.3392028808594,
+    "eval_logps/rejected": -370.232666015625,
+    "eval_loss": 0.6289177536964417,
+    "eval_rewards/accuracies": 0.7405063509941101,
+    "eval_rewards/chosen": 0.7478683590888977,
+    "eval_rewards/margins": 4.585729122161865,
+    "eval_rewards/rejected": -3.8378612995147705,
+    "eval_runtime": 70.1775,
     "eval_samples": 2500,
-    "eval_samples_per_second": 44.381,
-    "eval_steps_per_second": 1.402
+    "eval_samples_per_second": 35.624,
+    "eval_steps_per_second": 1.126
 }

runs/Sep29_22-34-07_action-graph-trainer/events.out.tfevents.1727656053.action-graph-trainer.718565.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71c573641baaf0e3920cfaf4291e3c5827addd7c33a74c2c3a9e9e8b40a32325
+size 828