Skip to main content

High-performance sigmoid attention kernels using Triton

Project description

Triton Sigmoid Attention

Python 3.11+ PyTorch 2.6+ License: MIT Code style: black

Fused sigmoid attention kernels for NVIDIA GPUs built with Triton.

Overview

Two implementations:

  • Dense - Same-length sequences (fastest)
  • Padded - Variable-length with padding masks

Features torch.compile support and causal masking. Requires Ampere architecture or newer.

Installation

pip install triton-sigmoid
Development install from source
git clone https://github.com/MSDLLCpapers/triton-sigmoid.git
cd triton-sigmoid
uv sync --extra dev
source .venv/bin/activate

Quick Start

Dense Attention (Same-Length Sequences)

import torch
from triton_sigmoid import sigmoid_attention

batch, seq_len, n_heads, head_dim = 2, 1024, 8, 64
q = torch.randn(batch, seq_len, n_heads, head_dim, device='cuda', dtype=torch.float16)
k = torch.randn(batch, seq_len, n_heads, head_dim, device='cuda', dtype=torch.float16)
v = torch.randn(batch, seq_len, n_heads, head_dim, device='cuda', dtype=torch.float16)

output = sigmoid_attention(q, k, v, is_causal=False)
output_causal = sigmoid_attention(q, k, v, is_causal=True)

Padded Attention (Variable-Length Sequences)

import torch
from triton_sigmoid import sigmoid_attention_padded

batch, seq_len, n_heads, head_dim = 2, 1024, 8, 64
q = torch.randn(batch, seq_len, n_heads, head_dim, device='cuda', dtype=torch.float16)
k = torch.randn(batch, seq_len, n_heads, head_dim, device='cuda', dtype=torch.float16)
v = torch.randn(batch, seq_len, n_heads, head_dim, device='cuda', dtype=torch.float16)

seq_lens_k = torch.tensor([800, 950], device='cuda', dtype=torch.int32)
seq_lens_q = torch.tensor([800, 950], device='cuda', dtype=torch.int32)

output = sigmoid_attention_padded(q, k, v, seq_lens_k=seq_lens_k, seq_lens_q=seq_lens_q)

compiled_fn = torch.compile(sigmoid_attention_padded)
output = compiled_fn(q, k, v, seq_lens_k=seq_lens_k, seq_lens_q=seq_lens_q, is_causal=True)

Performance

TFLOPS Comparison

See benchmarks/README.md for details.

Documentation

License

MIT License - see LICENSE file for details.

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

Citation

If you use this work in your research, please cite our paper:

@misc{sadashivaiah2026bettermodelsfastertraining,
      title={Better Models, Faster Training: Sigmoid Attention for single-cell Foundation Models}, 
      author={Vijay Sadashivaiah and Georgios Dasoulas and Judith Mueller and Soumya Ghosh},
      year={2026},
      eprint={2604.27124},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2604.27124}, 
}

Acknowledgments

Project details


Download files

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

Source Distribution

triton_sigmoid-0.1.2.tar.gz (30.6 kB view details)

Uploaded Source

Built Distribution

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

triton_sigmoid-0.1.2-py3-none-any.whl (21.8 kB view details)

Uploaded Python 3

File details

Details for the file triton_sigmoid-0.1.2.tar.gz.

File metadata

  • Download URL: triton_sigmoid-0.1.2.tar.gz
  • Upload date:
  • Size: 30.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for triton_sigmoid-0.1.2.tar.gz
Algorithm Hash digest
SHA256 67ad4b8b42b5970a20a056ac99666c28b8207928e8afaec404921248987769d9
MD5 bb30997fc4fa18a86b2132f02b29e22d
BLAKE2b-256 5d7788b7b2077039e347c81b4b32e436c5ad752eaaddecf3b7e9d2a44772ca93

See more details on using hashes here.

Provenance

The following attestation bundles were made for triton_sigmoid-0.1.2.tar.gz:

Publisher: publish.yml on MSDLLCpapers/triton-sigmoid

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

File details

Details for the file triton_sigmoid-0.1.2-py3-none-any.whl.

File metadata

  • Download URL: triton_sigmoid-0.1.2-py3-none-any.whl
  • Upload date:
  • Size: 21.8 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for triton_sigmoid-0.1.2-py3-none-any.whl
Algorithm Hash digest
SHA256 137dd0b32e55d9062c7650645ce28161e3a371c8fe390fd1577fb854663630dc
MD5 32424c2e51ce9e65814c7a0f2dc8504e
BLAKE2b-256 d168ba726cceb9964d03822fbac0049e09be10155d7c75db3f05b6e90e829708

See more details on using hashes here.

Provenance

The following attestation bundles were made for triton_sigmoid-0.1.2-py3-none-any.whl:

Publisher: publish.yml on MSDLLCpapers/triton-sigmoid

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