# cudaLLM

**Repository Path**: ByteDance-Seed/cudaLLM

## Basic Information

- **Project Name**: cudaLLM
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Apache-2.0
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-09-15
- **Last Updated**: 2026-09-17

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# CudaLLM: Training Language Models to Generate High-Performance CUDA Kernels

<a href="https://huggingface.co/ByteDance-Seed/cudaLLM-8B">
    <img 
        src="https://img.shields.io/badge/CudaLLM-Model-yellow?logo=huggingface&logoColor=yellow" 
        alt="CudaLLM Model"
    />
</a>

<a href="https://huggingface.co/datasets/ByteDance-Seed/cudaLLM-data">
    <img 
        src="https://img.shields.io/badge/CudaLLM-Data-blue?logo=huggingface&logoColor=yellow" 
        alt="CudaLLM Data"
    />
</a>

This project provides a complete pipeline for training LLMs to automatically generate efficient and correct CUDA kernels. By leveraging a two-stage process of SFT and RL, this framework fine-tunes a base model to write optimized CUDA code.

For demonstration purposes, this guide uses [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) as the base model.

## How It Works

The training methodology is composed of two main stages:

1.  **SFT:** The base LLM is first fine-tuned on a high-quality dataset of CUDA kernel examples. The data is generated by DeepSeek R1, DeepSeel Coder-7B, and Qwen2-32B.
2.  **RL:** After SFT, the model is further optimized through reinforcement learning. In this stage, the model generates CUDA kernels which are then compiled and tested. This feedback signal is used as a reward to train the model to produce valid kernels.

## Getting Started
To set up and run the training pipeline, follow these steps:
### Step 0: Prepare Datasets

First, you need to process the raw datasets for SFT and RL, and download the evaluation dataset. This script handles the necessary preprocessing.

  * **SFT Dataset:** `sft_cuda_llm_r1.parquet`
  * **RL Dataset:** `rl_cuda_llm_0424.parquet`
  * **Evaluation Dataset:** [KernelBench](https://huggingface.co/datasets/ScalingIntelligence/KernelBench)

Run the following command to begin:

```bash
# install verl, the git SHA is abb87bc147467589d1357dd80a1e3fefa188e11f
git clone https://github.com/volcengine/verl.git
cd verl
pip install --no-deps -e .
cd ..

python3 cuda_dataset.py
```

### Step 1: SFT

Next, fine-tune the base model on the prepared SFT dataset. This will adapt the model to the domain of CUDA code generation.

```bash
bash scripts/sft.sh
```

### Step 2: Evaluate the SFT Model

After the SFT stage is complete, evaluate the model's code generation accuracy on the **KernelBench** benchmark. This step provides a baseline measurement of the model's capabilities before reinforcement learning.

```bash
bash scripts/eval.sh
```

### Step 3: RL

Finally, use reinforcement learning to further enhance the SFT model's ability to generate performant code. For each node, this stage uses a hardware allocation as below:

  * **4x GPUs** are dedicated to the RL training loop.
  * **4x GPUs** are used to run the generated kernels, providing the reward needed for training.

```bash
bash scripts/rl.sh
```

Upon completion, you will have a model specifically trained to generate high-quality CUDA kernels. You can re-run the evaluation script (`eval.sh`) to measure the performance uplift from the RL stage.

### License
This project is licensed under the Apache License 2.0. See the [LICENSE](LICENSE) file for details.