i

metal-flash-attention-rs

public

9 stars

0 forks

0 issues

Commits

List of commits on branch main.

Unverified

072de902aaf3840393b4f151182e34ec8c63c4b9

Update gitignore

iivarflakstad committed 10 months ago

Unverified

70368eec75d9c97330c98fb6726af0ba9ea09c76

Simplify datatype. Use primitive types instead of custom structs (#2)

iivarflakstad committed a year ago

Unverified

b00965ed9d20fd563d3a7c315e2cfc138aa2214f

Simpler encode_gemm signature. Better error handling

iivarflakstad committed a year ago

Unverified

f7b6d22b75354af8d731371b483a1f65aace7810

gemm config. improve benchmarks

iivarflakstad committed a year ago

Verified

c7aa6d3cc7cb0ec1d7f5522da616cf20fcadd801

Merge pull request #1 from ivarflakstad/gemm-fix

iivarflakstad committed a year ago

Unverified

76bb93c9c240d637fea1a126df7bc78dc18c13d4

GEMM works!

iivarflakstad committed a year ago

README

The README file for this repository.

Metal Flash Attention Rust API

Work in progress!

Use MFA from Rust. Metal Flash Attention is a faster alternative to Metal Performance Shaders, a reference implementation of modern GPU algorithms, and a step toward defragmenting the AI ecosystem. See metal flash attention