updated dist files

Teach HIP grouped_gemm about autograd

Teach HIP grouped_gemm about autograd

Gate ROCm grouped_gemm hipBLASLt behind env flag

Clean ROCm grouped_gemm fallback and add tests

Fix ROCm grouped_gemm accumulation corruption

Add ROCm build debugging utilities

Add ROCm build artifacts and HIP backend

fix: readme tweak for new commit

feat: bump build for torch compile

fix: include torch compile flag

fix: support torch compile via fake tensors

fix: re upload build

fix: bump build

fix: add quickstart and avoid autotune when no cuda

feat: bump build for shared experts

feat: support shared experts layer and tests

fix: extract expert device mesh for group from unused prehook

Build from kernel-builder job

fix: prefer using passed parallel group

feat: full rebuild

fix: adjust layer params in source

feat: adjust layer params

feat: bump build

fix: improve expert parallel implementation and refactors

feat: bump build

fix: add parallel forward functional logic

feat: bump build for fully vendored version

fix: fully vendor stk and fix imports

fix: bump build

fix: vendor stk decorators

feat: bump builds

feat: add functional version of moe class and layer

fix: bump build and imports

fix: prefer relative imports

fix: bump builds

feat: vendor grouped gemm

feat: add build output

fix: prefer nix based library inclusion

feat: validate build with original test suite

fix: bump flake for github builder

feat: initial port of megablocks to builder format