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}
}

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-0.3.1.tar.gz (25.6 MB view details)

Uploaded Jun 30, 2026 Source

Built Distributions

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

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl (21.3 MB view details)

Uploaded Jun 30, 2026 CPython 3.12manylinux: glibc 2.34+ x86-64manylinux: glibc 2.35+ x86-64

fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl (5.0 MB view details)

Uploaded Jun 30, 2026 CPython 3.12manylinux: glibc 2.34+ ARM64manylinux: glibc 2.35+ ARM64

File details

Details for the file fastvideo_kernel-0.3.1.tar.gz.

File metadata

Download URL: fastvideo_kernel-0.3.1.tar.gz
Upload date: Jun 30, 2026
Size: 25.6 MB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for fastvideo_kernel-0.3.1.tar.gz
Algorithm	Hash digest
SHA256	`0cf25c6c7975c724e97ca5a43054758b41d1a64be4076f82ed11efca3dfaaf21`
MD5	`1936962815e667a9a9396ba43adab3cf`
BLAKE2b-256	`fec5c59e43c1d5d97c179cb168405b7a19e39dcfa8a876590fbf19b3fd2f413a`

See more details on using hashes here.

Provenance

The following attestation bundles were made for fastvideo_kernel-0.3.1.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.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: fastvideo_kernel-0.3.1.tar.gz
- Subject digest: 0cf25c6c7975c724e97ca5a43054758b41d1a64be4076f82ed11efca3dfaaf21
- Sigstore transparency entry: 2025926883
- Sigstore integration time: Jun 30, 2026
Source repository:
- Permalink: hao-ai-lab/FastVideo@31e26abec43bec9b25a68dd1f2618db4972fe293
- Branch / Tag: refs/heads/main
- Owner: https://github.com/hao-ai-lab
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: publish-kernel.yml@31e26abec43bec9b25a68dd1f2618db4972fe293
- Trigger Event: push

File details

Details for the file fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl.

File metadata

Download URL: fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl
Upload date: Jun 30, 2026
Size: 21.3 MB
Tags: CPython 3.12, manylinux: glibc 2.34+ x86-64, manylinux: glibc 2.35+ x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl
Algorithm	Hash digest
SHA256	`351d8c419a398bc63c7a9d3777b57ce806dca8a356292d563ce4f9d304060f83`
MD5	`0530871b532ffe2e9435360eac8ffff5`
BLAKE2b-256	`6dfbaee3afd273aa8b5e712451b495d7e64712ea8084fa9fbf3372ea96f20a25`

See more details on using hashes here.

Provenance

The following attestation bundles were made for fastvideo_kernel-0.3.1-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.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_x86_64.manylinux_2_35_x86_64.whl
- Subject digest: 351d8c419a398bc63c7a9d3777b57ce806dca8a356292d563ce4f9d304060f83
- Sigstore transparency entry: 2025926976
- Sigstore integration time: Jun 30, 2026
Source repository:
- Permalink: hao-ai-lab/FastVideo@31e26abec43bec9b25a68dd1f2618db4972fe293
- Branch / Tag: refs/heads/main
- Owner: https://github.com/hao-ai-lab
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: publish-kernel.yml@31e26abec43bec9b25a68dd1f2618db4972fe293
- Trigger Event: push

File details

Details for the file fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl.

File metadata

Download URL: fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl
Upload date: Jun 30, 2026
Size: 5.0 MB
Tags: CPython 3.12, manylinux: glibc 2.34+ ARM64, manylinux: glibc 2.35+ ARM64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl
Algorithm	Hash digest
SHA256	`e6890ccb986d4a3480426768ef1c2d16e6312fce15ad9c4c32a3659eff7f0ba8`
MD5	`1931855c0c6b6b536ce9fb62bd410a25`
BLAKE2b-256	`83c7f8710eba11d46758c026f8e64a2339a85e209638b571940bdee374872a6c`

See more details on using hashes here.

Provenance

The following attestation bundles were made for fastvideo_kernel-0.3.1-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.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: fastvideo_kernel-0.3.1-cp312-cp312-manylinux_2_34_aarch64.manylinux_2_35_aarch64.whl
- Subject digest: e6890ccb986d4a3480426768ef1c2d16e6312fce15ad9c4c32a3659eff7f0ba8
- Sigstore transparency entry: 2025927092
- Sigstore integration time: Jun 30, 2026
Source repository:
- Permalink: hao-ai-lab/FastVideo@31e26abec43bec9b25a68dd1f2618db4972fe293
- Branch / Tag: refs/heads/main
- Owner: https://github.com/hao-ai-lab
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: publish-kernel.yml@31e26abec43bec9b25a68dd1f2618db4972fe293
- Trigger Event: push

Release history Release notifications | RSS feed

0.3.2

Jul 3, 2026

This release

0.3.1

Jun 30, 2026

0.3.0

Jun 29, 2026

0.2.6

Feb 14, 2026

0.2.5

Jan 22, 2026

0.2.4

Jan 20, 2026

0.2.3

Jan 17, 2026

0.2.2

Jan 5, 2026

0.2.1

Dec 26, 2025

0.2.0

Dec 26, 2025

0.1.0

Dec 24, 2025

fastvideo-kernel 0.3.1

FastVideo Kernel

Installation

Standard Installation (Local Development)

Rocm Build

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

Usage

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

Benchmark

VSA (block-sparse) TFLOPs

TurboDiffusion Kernels

Requirements

Acknowledgement

Project links

Key dates

1 maintainer

Credits

GitHub Statistics

License

Requires

Classifiers

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

Release history Release notifications | RSS feed