George HotzandGitHub f57380cbc2 simplify GatedDeltaNetBlock using two state tensors (#15704) ...

* test double after

* simpler ssm

* no double test

2026-04-16 21:14:00 +08:00

..

apps

simplify GatedDeltaNetBlock using two state tensors (#15704)

2026-04-16 21:14:00 +08:00

codegen

beam uop (#15660)

2026-04-09 19:13:03 +03:00

engine

jit: capturedjit is linear (#15743)

2026-04-16 14:54:39 +03:00

mixin

cat the stack to mixin (#15715)

2026-04-13 18:44:39 -04:00

nn

delete old UOp.size [pr] (#15756)

2026-04-15 23:21:00 -04:00

renderer

amd: EMU DPP support (#15719)

2026-04-14 14:58:41 +08:00

runtime

jit: capturedjit is linear (#15743)

2026-04-16 14:54:39 +03:00

schedule

put the ranges on store instead of after (#15759)

2026-04-16 19:06:40 +08:00

uop

jit: capturedjit is linear (#15743)

2026-04-16 14:54:39 +03:00

viz

CALL with return value is FUNCTION (#15758)

2026-04-16 13:25:07 +08:00

__init__.py

start function and add walk rewrite (#14992)

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

callify.py

CALL with return value is FUNCTION (#15758)

2026-04-16 13:25:07 +08:00

device.py

DEV: suggest alternatives (#15732)

2026-04-14 23:42:32 -04:00

dtype.py

print function timing with DEBUG=2 (#15468)

2026-03-25 19:07:32 +08:00

function.py

delete old UOp.size [pr] (#15756)

2026-04-15 23:21:00 -04:00

gradient.py

CALL with return value is FUNCTION (#15758)

2026-04-16 13:25:07 +08:00

helpers.py

DEV: secondary targets (#15748)

2026-04-15 17:26:20 -04:00

py.typed

add a single py.typed (#6083)

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

tensor.py

exclude __del__ from TRACEMETA wrapping (#15747)

2026-04-16 10:49:55 +08:00