George Hotz 85dee83f5d amd flash attention cleanups + emulator fixes (#15431) ...

* amd flash attention cleanups

* simpler

* params

* fix emulator bugs

* fix idiv bug

* remove that test

* more emu fixes

2026-03-24 10:10:46 +08:00

..

apps

fix double after bug in rangeify (#15381)

2026-03-20 14:53:46 +08:00

codegen

removed unused named arg in rules [pr] (#15414)

2026-03-22 09:25:46 -04:00

engine

memplan on linears (#15422)

2026-03-23 19:50:16 +08:00

mixin

remove *0 hack in sign, gradient materializes zeros for unconnected nodes (#15416)

2026-03-22 12:49:26 -04:00

nn

dtypes.index -> dtypes.weakint (#15377)

2026-03-20 01:08:46 -04:00

renderer

sqtt: rdna4 decoder work (#15434)

2026-03-24 03:49:32 +09:00

runtime

add wmma support to amd_copy_matmul (#15384)

2026-03-20 19:02:19 +08:00

schedule

removed unused named arg in rules [pr] (#15414)

2026-03-22 09:25:46 -04:00

uop

amd flash attention cleanups + emulator fixes (#15431)

2026-03-24 10:10:46 +08:00

viz

viz: show dispatch to exec delay in sidebar (#15428)

2026-03-23 16:59:59 +09:00

__init__.py

start function and add walk rewrite (#14992)

2026-02-25 13:56:27 +08:00

device.py

FP8=1 llama works again, hipcc can run on macos (#15394)

2026-03-20 23:43:15 +09:00

dtype.py

remove match_dtype arg in Tensor._broadcasted (#15440)

2026-03-23 22:10:39 -04:00

function.py

more Tensor(UOp) cleanups (#15364)

2026-03-19 03:34:30 -04:00

gradient.py

add test for flat llama (#15327)

2026-03-18 15:16:33 +08:00

helpers.py

automatically emulate unsupported (tiny) floats [skip_process_replay] (#15366)

2026-03-20 02:31:44 -04:00

py.typed

add a single py.typed (#6083)

2024-08-14 17:31:46 -07:00

tensor.py

remove match_dtype arg in Tensor._broadcasted (#15440)

2026-03-23 22:10:39 -04:00