chenyu 1692087db5 _one_hot_along_dim input needs to be int (#9179) ...

* _one_hot_along_dim input needs to be int

indexing and onehot compare with arange, and non-int dtype is likely a bug

2025-02-20 09:00:43 -05:00

..

codegen

rename can_pad arg to cache [pr] (#9170)

2025-02-19 12:24:59 +01:00

engine

hotfix: add output buffer back to kernel parents + comment [pr] (#9174)

2025-02-19 14:22:01 +01:00

nn

Revert "match torch rmsnorm implementation (#6799)" (#9052)

2025-02-13 14:42:45 +08:00

renderer

Only add enable f16 directive if ShaderF16 is supported (#9163)

2025-02-19 17:20:03 +08:00

runtime

cuda hooking (#9180)

2025-02-20 19:20:01 +08:00

shape

symbolic gets its own file [pr] (#9132)

2025-02-17 18:55:21 +08:00

viz

correctly step through bottom_up_rewrites in viz [pr] (#9176)

2025-02-19 19:20:57 +01:00

__init__.py

UOp.variable (#7010)

2024-10-12 18:20:44 +08:00

device.py

fix win32 CPUProgram missing cache flush (#9171)

2025-02-19 21:38:51 +08:00

dtype.py

add truncate_bf16 (#9078)

2025-02-15 07:59:09 +08:00

gradient.py

fix pow backward when base is 0 (#9075)

2025-02-13 21:06:01 -05:00

helpers.py

Use proper units when printing beam time (#9103)

2025-02-17 23:41:38 +08:00

ops.py

rename can_pad arg to cache [pr] (#9170)

2025-02-19 12:24:59 +01:00

py.typed

add a single py.typed (#6083)

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

spec.py

UNIQUE UOp for buffer instead of arg (#9156)

2025-02-18 16:59:59 +08:00

tensor.py

_one_hot_along_dim input needs to be int (#9179)

2025-02-20 09:00:43 -05:00