mmaemu 0.1.0

First release of mmaemu: bit-exact emulation of NVIDIA tensor-core FP8 GEMMs (torch._scaled_mm, Transformer Engine) as an H100 PCIe (the default) or RTX 4060 computes them with cuBLASLt 12.8.3.14, on the CPU or another CUDA GPU.

pip install mmaemu==0.1.0

Scope, validation and limitations: CHANGELOG.md and docs/validation.md. Documentation: https://astding.github.io/mmaemu/

SHA-256 (the files attached here are the ones on PyPI):

f9725a7334901574debde819b9f285ca00443e2b6c7f6ccc9a7993ec0cf7e44f  mmaemu-0.1.0-py3-none-any.whl
740b69ac7e10f0542399fbecbea216e638dfe7249ff70ec4c6474cf9a15a5a13  mmaemu-0.1.0.tar.gz