Instructions to use kangdawei/DRA-GRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kangdawei/DRA-GRPO with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="kangdawei/DRA-GRPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("kangdawei/DRA-GRPO")
model = AutoModelForCausalLM.from_pretrained("kangdawei/DRA-GRPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use kangdawei/DRA-GRPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kangdawei/DRA-GRPO"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kangdawei/DRA-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/kangdawei/DRA-GRPO

SGLang

How to use kangdawei/DRA-GRPO with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "kangdawei/DRA-GRPO" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kangdawei/DRA-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "kangdawei/DRA-GRPO" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kangdawei/DRA-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use kangdawei/DRA-GRPO with Docker Model Runner:
```
docker model run hf.co/kangdawei/DRA-GRPO
```

kangdawei commited on Oct 1, 2025

Commit

40ba337

verified ·

1 Parent(s): 9c1d852

Model save

Browse files

Files changed (4) hide show

README.md +3 -5
all_results.json +4 -4
train_results.json +4 -4
trainer_state.json +0 -0

README.md CHANGED Viewed

@@ -1,11 +1,9 @@
 ---
 base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
-datasets: knoveleng/open-rs
 library_name: transformers
 model_name: DRA-GRPO
 tags:
 - generated_from_trainer
-- open-r1
 - trl
 - grpo
 licence: license
@@ -13,7 +11,7 @@ licence: license
 # Model Card for DRA-GRPO
-This model is a fine-tuned version of [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B) on the [knoveleng/open-rs](https://huggingface.co/datasets/knoveleng/open-rs) dataset.
 It has been trained using [TRL](https://github.com/huggingface/trl).
 ## Quick start
@@ -29,7 +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/kangdawei-texas-a-m-university/huggingface/runs/up34lf4c)
 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).
@@ -39,7 +37,7 @@ This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing
 - TRL: 0.16.0.dev0
 - Transformers: 4.49.0
 - Pytorch: 2.5.1
-- Datasets: 4.1.1
 - Tokenizers: 0.21.4
 ## Citations

 ---
 base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
 library_name: transformers
 model_name: DRA-GRPO
 tags:
 - generated_from_trainer
 - trl
 - grpo
 licence: license
 # Model Card for DRA-GRPO
+This model is a fine-tuned version of [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 ## Quick start
 ## 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/kangdawei-texas-a-m-university/huggingface/runs/e2dr6r95)
 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).
 - TRL: 0.16.0.dev0
 - Transformers: 4.49.0
 - Pytorch: 2.5.1
+- Datasets: 3.2.0
 - Tokenizers: 0.21.4
 ## Citations

all_results.json CHANGED Viewed

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 9.615809731966185e-07,
-    "train_runtime": 82.4093,
     "train_samples": 7000,
-    "train_samples_per_second": 145.615,
-    "train_steps_per_second": 6.067
 }

 {
     "total_flos": 0.0,
+    "train_loss": 0.012724974361105416,
+    "train_runtime": 59499.213,
     "train_samples": 7000,
+    "train_samples_per_second": 0.403,
+    "train_steps_per_second": 0.008
 }

train_results.json CHANGED Viewed

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 9.615809731966185e-07,
-    "train_runtime": 82.4093,
     "train_samples": 7000,
-    "train_samples_per_second": 145.615,
-    "train_steps_per_second": 6.067
 }

 {
     "total_flos": 0.0,
+    "train_loss": 0.012724974361105416,
+    "train_runtime": 59499.213,
     "train_samples": 7000,
+    "train_samples_per_second": 0.403,
+    "train_steps_per_second": 0.008
 }

trainer_state.json CHANGED Viewed

The diff for this file is too large to render. See raw diff