Awni Hannun a54f06b16f Fast RMS Norm (#862) ...

* fast rmsnorm

* no rms gpu

* kernel

* fix shared mem

* looped rms and donation in softmax

* Make the squaring in float32 to avoid underflow

* Fix the default StreamOrDevice for rope and rms_norm in fast

* nits

---------

Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>

2024-03-21 07:20:54 -07:00

..

array.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

buffer.h

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

CMakeLists.txt

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

constants.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

convert.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

convert.h

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

device.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

fast.cpp

Fast RMS Norm (#862)

2024-03-21 07:20:54 -07:00

fft.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

indexing.cpp

Update set item (#861)

2024-03-21 02:48:13 -07:00

indexing.h

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

linalg.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

load.cpp

simple typo fille (#848)

2024-03-19 06:15:17 -07:00

load.h

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

metal.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

mlx.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

ops.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

random.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

stream.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

transforms.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

trees.cpp

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

trees.h

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00

utils.h

Switch to nanobind (#839)

2024-03-18 20:12:25 -07:00