mirror of https://github.com/tracel-ai/burn.git
![]() * Update cubecl * Update to scope merge * Fix bitwise shift * Initial JIT implementation * Move testgen to burn-jit * Improve HA4/8 algo * Terminate units past the predefined 32 plane size * move jit backend back into `burn-vision` and make tests work * Add initial CPU implementation without stats * Implement stats * Implement all backends except fusion * Fix autodiff to use GPU when available * Fixes and cleanup * Add docs * Update cubecl * Compact labels for JIT * Improve JIT backend implementation by adding label compaction * Use GPU reduction for max label * Manually fuse presence and prefix sum * Make prefix sum more generic over line size * Add vision tests to xtask * Fix CPU and other review stuff * Add publish job * Review fixes * Remove autodiff implementation --------- Co-authored-by: Guillaume Lagrange <lagrange.guillaume.1@gmail.com> |
||
---|---|---|
.. | ||
main.rs |