End of training

README.md

@@ -3,9 +3,15 @@ library_name: transformers
 license: llama3
 base_model: meta-llama/Meta-Llama-3-8B-Instruct
 tags:
+- alignment-handbook
 - trl
 - simpo
 - generated_from_trainer
+- trl
+- simpo
+- generated_from_trainer
+datasets:
+- princeton-nlp/llama3-ultrafeedback-armorm
 model-index:
 - name: simpo-exps
   results: []
@@ -16,17 +22,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # simpo-exps
 
-This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) on an unknown dataset.
+This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) on the princeton-nlp/llama3-ultrafeedback-armorm dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.1795
-- Rewards/chosen: -17.3452
-- Rewards/rejected: -23.3517
+- Loss: 1.1745
+- Rewards/chosen: -17.4859
+- Rewards/rejected: -23.5403
 - Rewards/accuracies: 0.8435
-- Rewards/margins: 6.0065
-- Logps/rejected: -2.3352
-- Logps/chosen: -1.7345
-- Logits/rejected: -1.4942
-- Logits/chosen: -1.4832
+- Rewards/margins: 6.0544
+- Logps/rejected: -2.3540
+- Logps/chosen: -1.7486
+- Logits/rejected: -1.5014
+- Logits/chosen: -1.4906
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.9982631930527722,
+    "eval_logits/chosen": -1.4905521869659424,
+    "eval_logits/rejected": -1.501386284828186,
+    "eval_logps/chosen": -1.7485934495925903,
+    "eval_logps/rejected": -2.3540291786193848,
+    "eval_loss": 1.1744760274887085,
+    "eval_rewards/accuracies": 0.8434959053993225,
+    "eval_rewards/chosen": -17.485933303833008,
+    "eval_rewards/margins": 6.054358005523682,
+    "eval_rewards/rejected": -23.54029083251953,
+    "eval_runtime": 63.2847,
+    "eval_samples": 1961,
+    "eval_samples_per_second": 30.987,
+    "eval_steps_per_second": 1.944,
     "total_flos": 0.0,
     "train_loss": 1.7882541670789045,
     "train_runtime": 8154.3088,

config.json

@@ -24,6 +24,6 @@
   "tie_word_embeddings": false,
   "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": 0.9987943737441393,
-    "eval_logits/chosen": -2.756258010864258,
-    "eval_logits/rejected": -2.754289388656616,
-    "eval_logps/chosen": -0.6834676861763,
-    "eval_logps/rejected": -0.7465164661407471,
-    "eval_loss": 0.7904670238494873,
-    "eval_rewards/accuracies": 0.605053186416626,
-    "eval_rewards/chosen": -1.7086691856384277,
-    "eval_rewards/margins": 0.15762200951576233,
-    "eval_rewards/rejected": -1.8662911653518677,
-    "eval_runtime": 102.313,
-    "eval_samples": 2994,
-    "eval_samples_per_second": 29.263,
-    "eval_steps_per_second": 1.837
+    "epoch": 0.9982631930527722,
+    "eval_logits/chosen": -1.4905521869659424,
+    "eval_logits/rejected": -1.501386284828186,
+    "eval_logps/chosen": -1.7485934495925903,
+    "eval_logps/rejected": -2.3540291786193848,
+    "eval_loss": 1.1744760274887085,
+    "eval_rewards/accuracies": 0.8434959053993225,
+    "eval_rewards/chosen": -17.485933303833008,
+    "eval_rewards/margins": 6.054358005523682,
+    "eval_rewards/rejected": -23.54029083251953,
+    "eval_runtime": 63.2847,
+    "eval_samples": 1961,
+    "eval_samples_per_second": 30.987,
+    "eval_steps_per_second": 1.944
 }