MCPcopy Create free account

hub / github.com/ml-explore/mlx / functions

Functions4,946 in github.com/ml-explore/mlx

↓ 47 callersFunctionOp
mlx/backend/cpu/simd/neon_fp16_simd.h:99
↓ 47 callersFunctionbroadcast_shapes
mlx/utils.cpp:136
↓ 47 callersMethodget_library
mlx/backend/metal/device.cpp:527
↓ 47 callersMethodptr
mlx/backend/cuda/event.h:71
↓ 47 callersMethodrand
(shape)
python/tests/test_blas.py:1372
↓ 47 callersFunctionwhere
mlx/ops.cpp:1874
↓ 46 callersMethodnext
Iteration helper */
mlx/backend/metal/kernels/steel/attn/loader.h:258
↓ 45 callersMethoddequantize
(self)
python/mlx/nn/layers/quantized.py:379
↓ 45 callersFunctionmoveaxis
mlx/ops.cpp:1506
↓ 45 callersFunctionsplit
mlx/ops.cpp:1058
↓ 44 callersMethoddispatch_threadgroups
mlx/backend/metal/device.cpp:356
↓ 41 callersMethodlength
mlx/3rdparty/pocketfft.h:2566
↓ 41 callersFunctionmalloc_async
mlx/backend/cuda/allocator.cpp:397
↓ 40 callersMethodload_safe
mlx/backend/metal/kernels/steel/gemm/nax.h:743
↓ 40 callersMethodpop_back
mlx/small_vector.h:336
↓ 40 callersFunctionsize
The number of elements in the array. */
mlx/array.h:100
↓ 40 callersFunctionsquare
mlx/ops.cpp:2983
↓ 39 callersFunctionis_comparable_with_array
python/src/utils.h:44
↓ 39 callersFunctionpromote_types
mlx/dtype.cpp:86
↓ 39 callersMethodstep
mlx/backend/common/utils.h:104
↓ 37 callersFunctionbroadcast_arrays
Broadcast the input arrays against one another while ignoring the * axes specified in `ignore_axes`. Note, this API is internal only. * The `ignore_
mlx/ops.cpp:1630
↓ 37 callersMethodjvp
mlx/fast.cpp:29
↓ 37 callersFunctionkey
mlx/random.cpp:30
↓ 37 callersFunctionmin
mlx/ops.h:726
↓ 37 callersFunctionthrow_invalid_operation
python/src/utils.h:61
↓ 35 callersFunctionat_least_float
mlx/ops.cpp:57
↓ 35 callersMethodload
mlx/export.cpp:494
↓ 35 callersMethodsize
mlx/3rdparty/pocketfft.h:229
↓ 34 callersFunctionbench
(f, *args)
benchmarks/python/comparative/bench_mlx.py:37
↓ 34 callersFunctioncopy_cpu
mlx/backend/cpu/copy.cpp:312
↓ 34 callersFunctionequal
mlx/ops.cpp:1758
↓ 34 callersFunctiongather
mlx/ops.h:1060
↓ 34 callersMethodraw_ptr
mlx/backend/cuda/allocator.cpp:410
↓ 33 callersFunctionbench
(f, *args)
benchmarks/python/comparative/bench_torch.py:36
↓ 33 callersFunctionexp
mlx/backend/cpu/simd/math.h:28
↓ 33 callersFunctiongrad
* Returns a function which computes the gradient of the input function with * respect to a vector of input arrays. * * The function being differ
mlx/transforms.h:128
↓ 33 callersFunctionprod
mlx/ops.h:692
↓ 33 callersFunctiontake_along_axis
mlx/ops.cpp:3464
↓ 32 callersFunctionany
mlx/ops.h:568
↓ 32 callersFunctioncopy
mlx/backend/cpu/copy.cpp:179
↓ 32 callersFunctiondtype_to_cuda_type
mlx/backend/cuda/utils.cpp:28
↓ 32 callersMethodexport_function
mlx/export.cpp:758
↓ 32 callersFunctionnegative
mlx/ops.cpp:2779
↓ 31 callersFunctionbits
mlx/random.h:50
↓ 31 callersFunctioncontiguous_copy_gpu
mlx/backend/gpu/copy.cpp:37
↓ 31 callersMethoderase
mlx/compile.cpp:370
↓ 31 callersFunctionfn
(x)
examples/python/qqmm.py:96
↓ 31 callersFunctionnormal
mlx/random.h:104
↓ 31 callersMethodsort
mlx/backend/metal/kernels/sort.h:67
↓ 31 callersFunctionswapaxes
mlx/ops.cpp:1533
↓ 30 callersMethodassertCmpNumpy
( self, args: List[Union[Tuple[int], Any]], mx_fn: Callable[..., mx.array], np
python/tests/mlx_tests.py:79
↓ 30 callersMethodmaximum
python/src/array.cpp:59
↓ 30 callersFunctionselect
mlx/backend/cpu/simd/neon_fp16_simd.h:176
↓ 30 callersFunctionslice_update
Update a slice from the source array */
mlx/ops.cpp:822
↓ 30 callersFunctionsync_if_needed
(x)
benchmarks/python/comparative/bench_torch.py:47
↓ 30 callersMethodtensor
mlx/backend/cuda/cudnn_utils.h:111
↓ 29 callersMethodadd_temporary
mlx/backend/metal/device.cpp:334
↓ 29 callersFunctiondefault_stream
mlx/stream.cpp:40
↓ 29 callersFunctionpower
mlx/ops.cpp:3863
↓ 29 callersFunctionserialize
mlx/export.cpp:114
↓ 28 callersFunctioncomplex_mul_conj
Complex mul followed by conjugate
mlx/backend/metal/kernels/fft/radix.h:24
↓ 28 callersFunctionlog
mlx/ops.cpp:3097
↓ 27 callersMethoderase
mlx/backend/cuda/lru_cache.h:108
↓ 27 callersFunctionnew_scoped_memory_pool
mlx/backend/metal/device.cpp:824
↓ 27 callersFunctionscatter
mlx/ops.h:1208
↓ 26 callersFunctioneval
mlx/backend/cpu/eval.cpp:10
↓ 26 callersFunctionisnan
mlx/ops.cpp:1839
↓ 26 callersFunctionless
mlx/ops.cpp:1799
↓ 26 callersFunctionmaximum
mlx/ops.cpp:2944
↓ 26 callersFunctionto_arrays
python/src/utils.cpp:50
↓ 25 callersFunctioncomplex_mul
mlx/backend/metal/kernels/fft/radix.h:19
↓ 25 callersMethodsize
mlx/distributed/jaccl/lib/jaccl/ring.h:33
↓ 25 callersFunctiontree_map
Applies ``fn`` to the leaves of the Python tree ``tree`` and returns a new collection with the results. If ``rest`` is provided, every item i
python/mlx/utils.py:8
↓ 24 callersFunctioncopy
mlx/ops.cpp:287
↓ 24 callersFunctiondispatch_bool
mlx/backend/metal/kernels/steel/utils/integral_constant.h:94
↓ 24 callersFunctionelem_to_loc
mlx/backend/metal/kernels/utils.h:98
↓ 24 callersMethodload_unsafe
Load from device memory into threadgroup memory - without bound checking */
mlx/backend/metal/kernels/steel/gemm/loader.h:74
↓ 24 callersFunctionunary_fp
mlx/backend/cpu/unary.h:141
↓ 23 callersFunctioncos
mlx/ops.cpp:3007
↓ 23 callersFunctionis_available
mlx/device.cpp:28
↓ 23 callersMethodsize
mlx/distributed/ring/ring.cpp:473
↓ 23 callersMethodstore_result
Store results from simdgroup_matrix results into device memory */
mlx/backend/metal/kernels/steel/gemm/mma.h:535
↓ 23 callersFunctionunflatten
mlx/ops.cpp:404
↓ 22 callersMethodapply_gradients
Apply the gradients to the parameters and return the updated parameters. Can be used to update a model via ``model.update(opt.apply_g
python/mlx/optimizers/optimizers.py:85
↓ 22 callersFunctioncumsum
mlx/ops.cpp:3873
↓ 22 callersMethodget_peak_memory
mlx/backend/cuda/allocator.cpp:351
↓ 22 callersFunctionget_reduce_axes
python/src/utils.h:31
↓ 21 callersMethodadd_temporaries
mlx/backend/metal/device.cpp:338
↓ 21 callersFunctionconcatenate
mlx/ops.cpp:1198
↓ 21 callersFunctionelem_to_loc_broadcast
mlx/backend/metal/kernels/steel/utils.h:7
↓ 21 callersMethodelems
mlx/backend/metal/kernels/steel/gemm/mma.h:263
↓ 21 callersMethodemplace
mlx/export.cpp:638
↓ 21 callersFunctionfft
Compute the one-dimensional Fourier Transform. */
mlx/fft.h:52
↓ 21 callersFunctionfftn
mlx/fft.cpp:154
↓ 21 callersFunctionlayer_norm
(x, weight, bias, eps)
python/tests/test_fast.py:57
↓ 21 callersMethodminimum
python/src/array.cpp:63
↓ 21 callersFunctionsin
mlx/ops.cpp:3001
↓ 21 callersFunctionsoftmax
Softmax of an array. */
mlx/ops.h:1304
↓ 21 callersMethodstore_result_safe
mlx/backend/metal/kernels/steel/gemm/mma.h:568
↓ 21 callersFunctiontree_flatten
Flattens a Python tree to a list of key, value tuples. The keys are using the dot notation to define trees of arbitrary depth and complexity.
python/mlx/utils.py:118
← previousnext →101–200 of 4,946, ranked by callers