Awni Hannun ec72b44417 Add quantize/dequantize for mxfp8 and nvfp4 (#2688) ...

* Add quantize/dequantize slow path for mxfp8 and nvfp4

* fast cuda kernel for mx/nv quantization

* fallback for cuda < 12.8 (#2697)

* format (#2700)

* fix (#2701)

* metal kernels

* docs

* fix jit

* add default bits and group sizes

* improve quant docs

* fix output type of mxfp4 matmuls

2025-10-28 16:23:12 -07:00

..

array.cpp

Fix cumulative operations when axis=None (#2653)

2025-10-08 15:25:38 -07:00

buffer.h

Double for lapack (#1904)

2025-02-25 11:39:36 -08:00

CMakeLists.txt

Fix a few ccache cache miss (#2573)

2025-09-09 07:41:05 +09:00

constants.cpp

Array api (#1289)

2024-07-26 10:40:49 -07:00

convert.cpp

Support pickling array for bfloat16 (#2586)

2025-09-22 20:12:15 -07:00

convert.h

Support pickling array for bfloat16 (#2586)

2025-09-22 20:12:15 -07:00

cuda.cpp

Custom cuda kernel (#2517)

2025-08-20 17:20:22 -07:00

device.cpp

start cuda circle config (#2256)

2025-06-10 21:19:47 -07:00

distributed.cpp

NCCL backend (#2476)

2025-08-21 11:56:15 -07:00

export.cpp

Export with callback (#2612)

2025-10-08 19:24:33 -07:00

fast.cpp

Add sdpa with sinks (#2558)

2025-09-10 14:53:00 -07:00

fft.cpp

Use SmallVector for shapes and strides (#2454)

2025-08-05 09:41:03 +09:00

indexing.cpp

Dynamic slicing (#1741)

2025-01-07 14:02:16 -08:00

indexing.h

Remove "using namespace mlx::core" in python/src (#1689)

2024-12-11 15:45:39 -08:00

linalg.cpp

typing: add type hints to mlx.core.array, linalg, distributed, and random (#2565)

2025-09-04 09:08:11 -07:00

load.cpp

allow pathlib.Path to save/load functions (#2541)

2025-08-25 14:58:49 -07:00

load.h

Remove "using namespace mlx::core" in python/src (#1689)

2024-12-11 15:45:39 -08:00

memory.cpp

move memory APIs into top level mlx.core (#1982)

2025-03-21 07:25:12 -07:00

metal.cpp

Use SmallVector for shapes and strides (#2454)

2025-08-05 09:41:03 +09:00

mlx_func.cpp

fix wraps compile (#2461)

2025-08-04 16:14:18 -07:00

mlx_func.h

fix wraps compile (#2461)

2025-08-04 16:14:18 -07:00

mlx.cpp

Fix a few ccache cache miss (#2573)

2025-09-09 07:41:05 +09:00

ops.cpp

Add quantize/dequantize for mxfp8 and nvfp4 (#2688)

2025-10-28 16:23:12 -07:00

random.cpp

typing: add type hints to mlx.core.array, linalg, distributed, and random (#2565)

2025-09-04 09:08:11 -07:00

small_vector.h

Use SmallVector for shapes and strides (#2454)

2025-08-05 09:41:03 +09:00

stream.cpp

Remove "using namespace mlx::core" in python/src (#1689)

2024-12-11 15:45:39 -08:00

transforms.cpp

Compile now can attach arbitrary data to an entry (#2634)

2025-09-30 13:33:27 -07:00

trees.cpp

Limit grad recursion depth by not recursing through non-grad inputs (#1764)

2025-01-14 14:33:18 -08:00

trees.h

Limit grad recursion depth by not recursing through non-grad inputs (#1764)

2025-01-14 14:33:18 -08:00

utils.cpp

Add batch offsets for mx.fast.rope (#2564)

2025-09-08 17:35:07 -07:00

utils.h

Bump nanobind to 2.4 + fix (#1710)

2024-12-17 10:57:54 -08:00