MCPcopy Create free account

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

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

⨍ Functions 4,946◇ Types & classes 770↳ Endpoints 5
↓ 1 callersFunctionmlx_get_item_slice
python/src/indexing.cpp:87
↓ 1 callersFunctionmlx_load_gguf_helper
python/src/load.cpp:206
↓ 1 callersFunctionmlx_load_npy_helper
python/src/load.cpp:258
↓ 1 callersFunctionmlx_load_npz_helper
python/src/load.cpp:215
↓ 1 callersFunctionmlx_load_safetensor_helper
python/src/load.cpp:183
↓ 1 callersFunctionmlx_maximum_item
python/src/indexing.cpp:1051
↓ 1 callersFunctionmlx_minimum_item
python/src/indexing.cpp:1069
↓ 1 callersFunctionmlx_multiply_item
python/src/indexing.cpp:1015
↓ 1 callersFunctionmlx_scatter_args_int
python/src/indexing.cpp:476
↓ 1 callersFunctionmlx_scatter_args_nd
python/src/indexing.cpp:585
↓ 1 callersFunctionmlx_scatter_args_slice
python/src/indexing.cpp:532
↓ 1 callersFunctionmlx_subtract_item
python/src/indexing.cpp:997
↓ 1 callersFunctionmlx_to_dlpack
python/src/convert.cpp:157
↓ 1 callersFunctionmulti_block_sort
mlx/backend/metal/sort.cpp:114
↓ 1 callersFunctionmulti_upload_bluestein_fft
mlx/backend/metal/fft.cpp:349
↓ 1 callersMethodnamelist
python/src/load.cpp:72
↓ 1 callersFunctionnan_to_num
mlx/ops.cpp:1891
↓ 1 callersFunctionnative_bf16_max
mlx/distributed/jaccl/lib/jaccl/reduction_ops.h:58
↓ 1 callersFunctionnative_bf16_min
mlx/distributed/jaccl/lib/jaccl/reduction_ops.h:67
↓ 1 callersFunctionnative_bf16_sum
mlx/distributed/jaccl/lib/jaccl/reduction_ops.h:49
↓ 1 callersFunctionnccl_timeout
mlx/utils.h:176
↓ 1 callersFunctionnd_fft_op
mlx/backend/metal/fft.cpp:762
↓ 1 callersMethodneeds_commit
mlx/backend/metal/device.cpp:445
↓ 1 callersMethodneeds_commit
mlx/backend/cuda/device.cpp:458
↓ 1 callersFunctionnew_thread_local_stream
mlx/stream.cpp:77
↓ 1 callersMethodnext
mlx/random.cpp:20
↓ 1 callersFunctionnhwc_to_nchw
Return the shape and strides after transposing from NHWC to NCHW.
mlx/backend/cuda/cudnn_utils.cpp:40
↓ 1 callersFunctionnormal
r"""An initializer that returns samples from a normal distribution. Args: mean (float, optional): Mean of the normal distribution. Defaul
python/mlx/nn/init.py:37
↓ 1 callersFunctionnormal
mlx/random.cpp:174
↓ 1 callersFunctionnp_cumlogaddexp
(x1: np.ndarray, axis: int = -1)
python/tests/test_ops.py:47
↓ 1 callersFunctionnp_wrap_between
Wraps `x` between `[-a, a]`.
python/tests/test_ops.py:13
↓ 1 callersFunctionnum_threads
mlx/3rdparty/pocketfft.h:522
↓ 1 callersFunctionnumpy_segmented_mm_ref
Ground-truth reference in float64.
benchmarks/python/segmented_mm_bench.py:35
↓ 1 callersFunctionnvml_get_memory
mlx/backend/cuda/device_info.cpp:66
↓ 1 callersFunctionnvml_init
mlx/backend/cuda/device_info.cpp:38
↓ 1 callersMethodop_max
mlx/distributed/mpi/mpi.cpp:251
↓ 1 callersMethodop_min
mlx/distributed/mpi/mpi.cpp:264
↓ 1 callersMethodop_sum
mlx/distributed/mpi/mpi.cpp:240
↓ 1 callersMethodoperator[]
mlx/3rdparty/pocketfft.h:356
↓ 1 callersFunctionpack_and_quantize
mlx/ops.cpp:4688
↓ 1 callersFunctionpad_and_slice_conv_3D_gpu
mlx/backend/metal/conv.cpp:624
↓ 1 callersFunctionpage_aligned_alloc
mlx/distributed/jaccl/lib/jaccl/rdma.cpp:23
↓ 1 callersFunctionparse_cases
(cases)
benchmarks/python/segmented_mm_bench.py:16
↓ 1 callersFunctionparse_cases
(cases)
benchmarks/python/block_masked_mm_bench.py:16
↓ 1 callersFunctionparse_devices_json
mlx/distributed/jaccl/lib/jaccl/jaccl.cpp:17
↓ 1 callersFunctionparse_hardware_ports
(ports_string)
python/mlx/_distributed_utils/config.py:188
↓ 1 callersFunctionplan_fft
mlx/backend/metal/fft.cpp:113
↓ 1 callersFunctionplot_density
(ax_perf, ax_speedup, density, dtype)
benchmarks/python/masked_scatter.py:174
↓ 1 callersFunctionpread
There is no pread on Windows, emulate it with ReadFile.
mlx/io/load.cpp:120
↓ 1 callersFunctionprecompiled_cuda_kernel
mlx/backend/cuda/no_cuda.cpp:29
↓ 1 callersFunctionprelu
r"""Applies the element-wise parametric ReLU. .. math:: \text{PReLU}(x) = \max(0,x) + a * \min(0,x) where :math:`a` is an array.
python/mlx/nn/layers/activations.py:261
↓ 1 callersFunctionprepare_ethernet_hostfile
(args, hosts)
python/mlx/_distributed_utils/config.py:434
↓ 1 callersFunctionprepare_inputs
(B, qL, kL, D, qH, kH, mask, transpose, dtype)
python/tests/test_fast_sdpa.py:78
↓ 1 callersFunctionprepare_inputs
(B, qL, kL, D, qH, kH, mask, transpose, dtype)
benchmarks/python/sdpa_bench.py:31
↓ 1 callersFunctionprepare_tb_hostfile
(args, hosts, sshinfo)
python/mlx/_distributed_utils/config.py:515
↓ 1 callersFunctionpreprocess_einsum_inputs
Collapse repeat indices and sum single dimensions. For example: - "aa" becomes "a" - "ij,jk->k" becoms "j,jk->k"
mlx/einsum.cpp:490
↓ 1 callersMethodpreprocess_output
(self, data, is_stdout=False)
python/mlx/_distributed_utils/launch.py:76
↓ 1 callersFunctionprimitive_root
mlx/backend/metal/fft.cpp:243
↓ 1 callersMethodprint_functions
mlx/export.cpp:571
↓ 1 callersFunctionprint_table
(headers, rows)
benchmarks/python/segmented_mm_bench.py:82
↓ 1 callersFunctionprint_table
(headers, rows)
benchmarks/python/block_masked_mm_bench.py:68
↓ 1 callersMethodprod
mlx/3rdparty/pocketfft.h:456
↓ 1 callersMethodpush
mlx/3rdparty/pocketfft.h:577
↓ 1 callersFunctionpy_vmap
python/src/transforms.cpp:309
↓ 1 callersFunctionqmm_nax
mlx/backend/metal/quantized.cpp:473
↓ 1 callersFunctionqmm_splitk
mlx/backend/metal/quantized.cpp:774
↓ 1 callersFunctionqmv_quad
mlx/backend/metal/quantized.cpp:177
↓ 1 callersFunctionqqmm_impl
mlx/backend/cuda/quantized/qqmm_impl.cpp:8
↓ 1 callersFunctionquantize
mlx/ops.cpp:4908
↓ 1 callersFunctionquantize_dequantize
mlx/backend/metal/quantized.cpp:1564
↓ 1 callersFunctionqvm
mlx/backend/metal/quantized.cpp:419
↓ 1 callersFunctionqvm_split_k
mlx/backend/metal/quantized.cpp:298
↓ 1 callersFunctionradix_func
mlx/backend/metal/kernels/hadamard.h:11
↓ 1 callersFunctionread_cached_ptx
Try to read the cached |ptx| and |ptx_kernels| from |cache_dir|.
mlx/backend/cuda/jit_module.cpp:144
↓ 1 callersFunctionrecip
mlx/backend/cpu/simd/base_simd.h:80
↓ 1 callersMethodrecv
mlx/distributed/distributed.cpp:83
↓ 1 callersFunctionrecvAll
mlx/distributed/nccl/nccl.cpp:116
↓ 1 callersFunctionrecv_like
mlx/distributed/ops.cpp:152
↓ 1 callersMethodregister_output_array
mlx/backend/metal/device.cpp:323
↓ 1 callersFunctionrelease
mlx/allocator.h:42
↓ 1 callersFunctionrepeat_input_to_compiled
tests/compile_tests.cpp:591
↓ 1 callersFunctionreset_peak_memory
mlx/backend/cuda/allocator.cpp:427
↓ 1 callersMethodreset_peak_memory
mlx/backend/no_gpu/allocator.cpp:36
↓ 1 callersMethodreset_peak_memory
mlx/backend/metal/allocator.h:32
↓ 1 callersMethodreset_to_inline_storage
Clear and go back to inline storage. Dynamic storage is *not* freed. For internal use only.
mlx/small_vector.h:482
↓ 1 callersMethodresize
mlx/backend/metal/resident.cpp:52
↓ 1 callersFunctionretain_graph
mlx/transforms_impl.h:84
↓ 1 callersFunctionrfft2
Compute the two-dimensional Fourier Transform on a real input. */
mlx/fft.h:183
↓ 1 callersFunctionrfftfreq
mlx/fft.cpp:328
↓ 1 callersFunctionrow_reduce_general_dispatch
mlx/backend/metal/reduce.cpp:539
↓ 1 callersFunctionrow_reduce_looped
mlx/backend/metal/reduce.cpp:489
↓ 1 callersFunctionrow_reduce_simple
mlx/backend/metal/reduce.cpp:449
↓ 1 callersFunctionrow_reduce_small
mlx/backend/metal/reduce.cpp:393
↓ 1 callersFunctionrsqrt
mlx/backend/cpu/simd/base_simd.h:75
↓ 1 callersMethodrun
mlx/backend/cpu/eig.cpp:155
↓ 1 callersMethodrun
mlx/backend/cpu/svd.cpp:158
↓ 1 callersMethodrun
mlx/backend/cpu/eigh.cpp:111
↓ 1 callersFunctionrun_conv_checks
mlx/ops.cpp:4029
↓ 1 callersFunctionsafe_div
mlx/backend/metal/reduce.cpp:188
↓ 1 callersFunctionsave_cuda_graphs_dot_file
mlx/backend/cuda/device.cpp:22
← previousnext →1,601–1,700 of 4,946, ranked by callers