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:
+- yakazimir/llama3-ultrafeedback-armorm
 model-index:
 - name: llama3instruct_-orpo-10-0_5-1e-6-1_best
   results: []
@@ -16,17 +22,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # llama3instruct_-orpo-10-0_5-1e-6-1_best
 
-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 yakazimir/llama3-ultrafeedback-armorm dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.4129
-- Rewards/chosen: -11.3049
-- Rewards/rejected: -15.3895
-- Rewards/accuracies: 0.8133
-- Rewards/margins: 4.0846
-- Logps/rejected: -1.5389
-- Logps/chosen: -1.1305
-- Logits/rejected: -1.3507
-- Logits/chosen: -1.3829
+- Loss: 2.4050
+- Rewards/chosen: -11.3456
+- Rewards/rejected: -15.4559
+- Rewards/accuracies: 0.8102
+- Rewards/margins: 4.1103
+- Logps/rejected: -1.5456
+- Logps/chosen: -1.1346
+- Logits/rejected: -1.3837
+- Logits/chosen: -1.4182
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.9989071038251366,
+    "eval_logits/chosen": -1.4181982278823853,
+    "eval_logits/rejected": -1.3836873769760132,
+    "eval_logps/chosen": -1.1345627307891846,
+    "eval_logps/rejected": -1.5455946922302246,
+    "eval_loss": 2.404977321624756,
+    "eval_rewards/accuracies": 0.8102409839630127,
+    "eval_rewards/chosen": -11.345626831054688,
+    "eval_rewards/margins": 4.110320091247559,
+    "eval_rewards/rejected": -15.45594596862793,
+    "eval_runtime": 33.5188,
+    "eval_samples": 1318,
+    "eval_samples_per_second": 39.321,
+    "eval_steps_per_second": 2.476,
     "total_flos": 0.0,
     "train_loss": 3.252214796955267,
     "train_runtime": 5969.5229,

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

@@ -0,0 +1,16 @@
+{
+    "epoch": 0.9989071038251366,
+    "eval_logits/chosen": -1.4181982278823853,
+    "eval_logits/rejected": -1.3836873769760132,
+    "eval_logps/chosen": -1.1345627307891846,
+    "eval_logps/rejected": -1.5455946922302246,
+    "eval_loss": 2.404977321624756,
+    "eval_rewards/accuracies": 0.8102409839630127,
+    "eval_rewards/chosen": -11.345626831054688,
+    "eval_rewards/margins": 4.110320091247559,
+    "eval_rewards/rejected": -15.45594596862793,
+    "eval_runtime": 33.5188,
+    "eval_samples": 1318,
+    "eval_samples_per_second": 39.321,
+    "eval_steps_per_second": 2.476
+}