Model save

README.md

@@ -27,7 +27,7 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/bae_bbae/huggingface/runs/t7zaxofy) 
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/bae_bbae/huggingface/runs/ex3bpmc2) 
 
 
 This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 6.126944848091303e-07,
-    "train_runtime": 21.3548,
+    "train_loss": 0.08691239206201022,
+    "train_runtime": 9457.957,
     "train_samples": 45000,
-    "train_samples_per_second": 1011.481,
-    "train_steps_per_second": 42.145
+    "train_samples_per_second": 2.284,
+    "train_steps_per_second": 0.095
 }

config.json

@@ -23,7 +23,7 @@
   "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.48.1",
-  "use_cache": false,
+  "use_cache": true,
   "use_sliding_window": false,
   "vocab_size": 151936
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 6.126944848091303e-07,
-    "train_runtime": 21.3548,
+    "train_loss": 0.08691239206201022,
+    "train_runtime": 9457.957,
     "train_samples": 45000,
-    "train_samples_per_second": 1011.481,
-    "train_steps_per_second": 42.145
+    "train_samples_per_second": 2.284,
+    "train_steps_per_second": 0.095
 }