Model save

README.md

@@ -1,19 +1,17 @@
 ---
 base_model: Qwen/Qwen2.5-Math-7B
-datasets: DigitalLearningGmbH/MATH-lighteval
 library_name: transformers
-model_name: Qwen-2.5-7B-GRPO-Base-32Action_647
+model_name: Qwen-2.5-7B-GRPO-Base-32Action_173
 tags:
 - generated_from_trainer
-- open-r1
 - trl
 - grpo
 licence: license
 ---
 
-# Model Card for Qwen-2.5-7B-GRPO-Base-32Action_647
+# Model Card for Qwen-2.5-7B-GRPO-Base-32Action_173
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-Math-7B](https://huggingface.co/Qwen/Qwen2.5-Math-7B) on the [DigitalLearningGmbH/MATH-lighteval](https://huggingface.co/datasets/DigitalLearningGmbH/MATH-lighteval) dataset.
+This model is a fine-tuned version of [Qwen/Qwen2.5-Math-7B](https://huggingface.co/Qwen/Qwen2.5-Math-7B).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start
@@ -22,14 +20,14 @@ It has been trained using [TRL](https://github.com/huggingface/trl).
 from transformers import pipeline
 
 question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
-generator = pipeline("text-generation", model="luckeciano/Qwen-2.5-7B-GRPO-Base-32Action_647", device="cuda")
+generator = pipeline("text-generation", model="luckeciano/Qwen-2.5-7B-GRPO-Base-32Action_173", device="cuda")
 output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
 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/max-ent-llms/PolicyGradientStability/runs/qj1xvx50) 
+[<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/max-ent-llms/PolicyGradientStability/runs/34t8qyp3) 
 
 
 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": 2.261561673488899e-09,
-    "train_runtime": 16912.9255,
+    "train_loss": -5.137796188492417e-10,
+    "train_runtime": 18587.2058,
     "train_samples": 7500,
-    "train_samples_per_second": 0.568,
-    "train_steps_per_second": 0.006
+    "train_samples_per_second": 0.516,
+    "train_steps_per_second": 0.005
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 2.261561673488899e-09,
-    "train_runtime": 16912.9255,
+    "train_loss": -5.137796188492417e-10,
+    "train_runtime": 18587.2058,
     "train_samples": 7500,
-    "train_samples_per_second": 0.568,
-    "train_steps_per_second": 0.006
+    "train_samples_per_second": 0.516,
+    "train_steps_per_second": 0.005
 }