Instructions to use AscendKernelGen/KernelGen-LM-32B-RL with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AscendKernelGen/KernelGen-LM-32B-RL with Transformers:

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

pipe = pipeline("text-generation", model="AscendKernelGen/KernelGen-LM-32B-RL")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AscendKernelGen/KernelGen-LM-32B-RL")
model = AutoModelForCausalLM.from_pretrained("AscendKernelGen/KernelGen-LM-32B-RL")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use AscendKernelGen/KernelGen-LM-32B-RL with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AscendKernelGen/KernelGen-LM-32B-RL"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AscendKernelGen/KernelGen-LM-32B-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AscendKernelGen/KernelGen-LM-32B-RL

SGLang

How to use AscendKernelGen/KernelGen-LM-32B-RL 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 "AscendKernelGen/KernelGen-LM-32B-RL" \
    --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": "AscendKernelGen/KernelGen-LM-32B-RL",
		"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 "AscendKernelGen/KernelGen-LM-32B-RL" \
        --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": "AscendKernelGen/KernelGen-LM-32B-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AscendKernelGen/KernelGen-LM-32B-RL with Docker Model Runner:
```
docker model run hf.co/AscendKernelGen/KernelGen-LM-32B-RL
```

AscendKernelGen commited on Jan 26

Commit

796bca4

verified ·

1 Parent(s): 754670c

Update README.md

Browse files

Files changed (1) hide show

README.md +4 -4

README.md CHANGED Viewed

@@ -19,10 +19,10 @@ The Ascend KernelGen Technical Report is published at https://arxiv.org/abs/2601
 Our framework, **Ascend KernelGen (AKGen)**, bridges the gap between general-purpose code generation and hardware-specific programming through a closed-loop system of data construction, training, and evaluation. Key innovations include:
-* [cite_start]**Ascend-CoT Dataset:** A high-quality, domain-specific dataset incorporating **Chain-of-Thought (CoT)** reasoning. It combines documentation-based reasoning, code-centric reasoning derived from real-world kernel implementations, and general reasoning chains to capture the structured logic required for low-level NPU programming.
-* [cite_start]**Domain-Adaptive Post-Training:** A two-stage optimization process. We first employ **Supervised Fine-Tuning (SFT)** with error-derived supervision (correcting API misuse and numerical errors). This is followed by **Reinforcement Learning (RL)** using Direct Preference Optimization (DPO), driven by execution-based correctness and performance signals.
-* [cite_start]**Hardware-Grounded Evaluation:** Validated using **NPUKernelBench**, a comprehensive benchmark that assesses compilation success, functional correctness, and performance (latency) on real Ascend hardware across varying complexity levels.
-* [cite_start]**Performance:** The model demonstrates siginificant improvement on complex Level-2 kernels compared to baselines, and effectively solving tasks where general-purpose models (like Qwen3, Llama3.1) fail completely.
 ## Citation
 @article{cao2026ascendkernelgen,

 Our framework, **Ascend KernelGen (AKGen)**, bridges the gap between general-purpose code generation and hardware-specific programming through a closed-loop system of data construction, training, and evaluation. Key innovations include:
+* **Ascend-CoT Dataset:** A high-quality, domain-specific dataset incorporating **Chain-of-Thought (CoT)** reasoning. It combines documentation-based reasoning, code-centric reasoning derived from real-world kernel implementations, and general reasoning chains to capture the structured logic required for low-level NPU programming.
+* **Domain-Adaptive Post-Training:** A two-stage optimization process. We first employ **Supervised Fine-Tuning (SFT)** with error-derived supervision (correcting API misuse and numerical errors). This is followed by **Reinforcement Learning (RL)** using Direct Preference Optimization (DPO), driven by execution-based correctness and performance signals.
+* **Hardware-Grounded Evaluation:** Validated using **NPUKernelBench**, a comprehensive benchmark that assesses compilation success, functional correctness, and performance (latency) on real Ascend hardware across varying complexity levels.
+* **Performance:** The model demonstrates siginificant improvement on complex Level-2 kernels compared to baselines, and effectively solving tasks where general-purpose models (like Qwen3, Llama3.1) fail completely.
 ## Citation
 @article{cao2026ascendkernelgen,