PyPI page
Home page
Author:
None
License:
MIT
Summary:
Triton GPU kernels for transformer operations — softmax, FlashAttention-2, RMSNorm, SwiGLU, quantized matmul
Latest version:
0.2.2
Required dependencies:
torch
|
triton
|
triton-windows
Optional dependencies:
ninja
|
numpy
|
pytest
Downloads last day:
8
Downloads last week:
39
Downloads last month:
571