FastVideo Kernel

CUDA kernels for FastVideo video generation.

Installation

Standard Installation (Local Development)

This will automatically detect your GPU architecture. If an NVIDIA Hopper (H100/sm_90a) GPU is detected, ThunderKittens kernels will be enabled. Otherwise, they will be skipped, and the package will use Triton fallbacks at runtime.

Before installation, set CUDA toolchain paths:

export CUDA_HOME=/usr/local/cuda
export CUDACXX=$CUDA_HOME/bin/nvcc

git submodule update --init --recursive
cd fastvideo-kernel
./build.sh

Rocm Build

If you are in a rocm environment without the compilation toolchaine of CUDA.

cd fastvideo-kernel
./build.sh --rocm

Optional: FA4 CuTe block-sparse backend (VSA-256 fastpath)

The VSA-256 fastpath (tile volume 256, on NVIDIA Blackwell / sm_100) routes to the FlashAttention-4 CuTe-DSL block-sparse kernel exposed as flash_attn.cute. This is an optional dependency: it is imported lazily, and video_sparse_attn transparently falls back to the Triton backend when it is absent (so the package is fully usable without it).

The symbols the fastpath needs (flash_attn.cute.block_sparsity.BlockSparseTensorsTorch, flash_attn.cute.interface._flash_attn_fwd) are provided upstream by Dao-AILab/flash-attention. Pin to commit c19cd20e: the wrapper targets that revision's _flash_attn_fwd signature (m_block_size / n_block_size); later upstream revisions reshaped it into a tile_mn tuple and are not drop-in compatible.

pip install "nvidia-cutlass-dsl>=4.5.0" torchvision
pip install "git+https://github.com/Dao-AILab/flash-attention.git@c19cd20e#subdirectory=flash_attn/cute"

The CuTe kernel JIT-compiles on first use. Verified on Blackwell (sm_100) against tests/test_vsa256_forward*.py.

Usage

Sliding Tile Attention (STA) & Video Sparse Attention (VSA)

For detailed usage, please check the Attention Documentation.

from fastvideo_kernel import sliding_tile_attention, video_sparse_attn, moba_attn_varlen

# Example: Sliding Tile Attention
out = sliding_tile_attention(q, k, v, window_sizes, text_len)

# Example: Video Sparse Attention (with Triton fallback)
out = video_sparse_attn(q, k, v, block_sizes, block_sizes, topk=5)

# Example: VMoBA
out = moba_attn_varlen(q, k, v, cu_seqlens_q, cu_seqlens_k, ...)

Benchmark

VSA (block-sparse) TFLOPs

After building/installing fastvideo-kernel, run:

cd fastvideo-kernel
python benchmarks/bench_vsa.py --batch_size 1 --num_heads 16 --head_dim 128 --q_seq_lens 49152 --topk 64

TurboDiffusion Kernels

This package also includes kernels from TurboDiffusion, including INT8 GEMM, Quantization, RMSNorm and LayerNorm.

Requirements

  • Runtime:
    • NVIDIA H100 (sm_90a) for C++ optimized kernels.
    • Any CUDA GPU for Triton-based fallbacks.
  • Build:
    • CUDA Toolkit 12.3+
    • CUDA_HOME must be set (for example, /usr/local/cuda)
    • CUDACXX must be set (for example, $CUDA_HOME/bin/nvcc)
    • C++20 compatible compiler (GCC 10+, Clang 11+)

Acknowledgement

This package structure and build system are based on sgl-kernel from the SGLang project.

The implementation of turbodiffusion kernels is adapted from TurboDiffusion. If you use these kernels, please cite:

@article{zhang2025turbodiffusion,
  title={TurboDiffusion: Accelerating Video Diffusion Models by 100-200 Times},
  author={Zhang, Jintao and Zheng, Kaiwen and Jiang, Kai and Wang, Haoxu and Stoica, Ion and Gonzalez, Joseph E and Chen, Jianfei and Zhu, Jun},
  journal={arXiv preprint arXiv:2512.16093},
  year={2025}
}

Project links

Data verified by PyPI on Jun 29, 2026
Data provided by the project maintainers, verified at the time the release was uploaded to PyPI.

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Newer release available (0.3.2)

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for SolitaryThinker from gravatar.com SolitaryThinker

Credits

Author: Hao AI Lab

GitHub Statistics

Data verified by PyPI on Jun 29, 2026
The GitHub source repository was provided by the project maintainers and verified by PyPI at the time of upload. Stars, forks, and open issues/PRs are derived from that repository and have not been independently verified.

License

Apache License

Requires

Python >=3.10

Classifiers

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

fastvideo_kernel-0.3.0.tar.gz (20.6 MB view details)

Uploaded Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl (21.3 MB view details)

Uploaded CPython 3.12manylinux: glibc 2.34+ x86-64manylinux: glibc 2.35+ x86-64

fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl (5.0 MB view details)

Uploaded CPython 3.12manylinux: glibc 2.34+ ARM64manylinux: glibc 2.35+ ARM64

File details

Details for the file fastvideo_kernel-0.3.0.tar.gz.

File metadata

  • Download URL: fastvideo_kernel-0.3.0.tar.gz
  • Upload date:
  • Size: 20.6 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for fastvideo_kernel-0.3.0.tar.gz
Algorithm Hash digest
SHA256 12f3bbeed0e74907d6724e1ce19d48cbec0d93868791a934d71afc6a51812f5d
MD5 f7d92b7c652005de37ab7252f4e08dfa
BLAKE2b-256 39c2db9df659683cdde41d8b3fd18ac4949955b7b5bcbf5b782bf01e9d291dd1

See more details on using hashes here.

Provenance

The following attestation bundles were made for fastvideo_kernel-0.3.0.tar.gz:

Publisher: publish-kernel.yml on hao-ai-lab/FastVideo

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl.

File metadata

File hashes

Hashes for fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl
Algorithm Hash digest
SHA256 eeed60424c2fa5da5e1c73be7f2f153969c14cf7250877e4024496fd838591e3
MD5 0288945947db733334271053d2e5864e
BLAKE2b-256 ec36ff8527fff26afdc2d74ae78e809a8e3b48ddc8b2cbcd09d0be16909ec141

See more details on using hashes here.

Provenance

The following attestation bundles were made for fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl:

Publisher: publish-kernel.yml on hao-ai-lab/FastVideo

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl.

File metadata

File hashes

Hashes for fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl
Algorithm Hash digest
SHA256 41979805f9a5e0c57be7f0d26a62717f5c6baeaf7c4b14cced28d883d3136b64
MD5 678d05d341f7b9ef2397c9165ba80dbb
BLAKE2b-256 c3e9c3944cad09a40c06f12dff103858863030b0c80529323b4a0d8b87557033

See more details on using hashes here.

Provenance

The following attestation bundles were made for fastvideo_kernel-0.3.0-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl:

Publisher: publish-kernel.yml on hao-ai-lab/FastVideo

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page