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
↓ 5 callersMethodseek
mlx/io/load.h:145
↓ 5 callersMethodsend
mlx/distributed/jaccl/lib/jaccl/tcp.cpp:137
↓ 5 callersMethodset_dtype
Set the dtype of the module's parameters. Args: dtype (Dtype): The new dtype. predicate (typing.Callable, optional):
python/mlx/nn/layers/base.py:597
↓ 5 callersFunctionshapes_without_reduction_axes
mlx/backend/common/reduce.cpp:7
↓ 5 callersMethodstore_safe
mlx/backend/metal/kernels/steel/gemm/nax.h:778
↓ 5 callersFunctiontan
mlx/ops.cpp:3013
↓ 5 callersFunctionto_fp8
mlx/ops.cpp:5252
↓ 5 callersFunctionto_scalar
python/src/convert.cpp:161
↓ 5 callersFunctiontopk
Returns topk elements of the flattened array. */
mlx/ops.cpp:2674
↓ 5 callersFunctiontree_flatten_with_structure
python/src/trees.cpp:277
↓ 5 callersFunctiontree_map
python/src/trees.cpp:17
↓ 5 callersFunctiontree_unflatten
Recreate a Python tree from its flat representation. .. code-block:: python from mlx.utils import tree_unflatten d = tree_unfla
python/mlx/utils.py:194
↓ 5 callersFunctionunary
mlx/backend/cpu/unary.h:56
↓ 5 callersFunctionzeros
mlx/ops.cpp:330
↓ 4 callersMethodDevice
mlx/backend/cuda/device.cpp:40
↓ 4 callersMethod__init__
(self, dims: int, eps: float = 1e-5)
python/mlx/nn/layers/normalization.py:134
↓ 4 callersMethod__init__
(self)
python/tests/test_export_import.py:354
↓ 4 callersFunction_calculate_fan_in_fan_out
(x)
python/mlx/nn/init.py:128
↓ 4 callersFunction_get_weight
(ind, grid, dist)
python/mlx/nn/layers/upsample.py:62
↓ 4 callersFunctionadd_gather_strides_and_shapes
mlx/backend/metal/quantized.cpp:158
↓ 4 callersFunctionadjust_matrix_offsets
mlx/backend/metal/kernels/fp_quantized_nax.h:454
↓ 4 callersFunctionall_gather
mlx/distributed/ops.cpp:79
↓ 4 callersFunctionall_max
mlx/distributed/ops.cpp:43
↓ 4 callersMethodall_max
mlx/distributed/mpi/mpi.cpp:404
↓ 4 callersFunctionall_min
mlx/distributed/ops.cpp:61
↓ 4 callersMethodall_min
mlx/distributed/mpi/mpi.cpp:419
↓ 4 callersFunctionbench
(f, m, v)
benchmarks/python/blas/bench_gemv.py:36
↓ 4 callersFunctionbinary_op_gpu_inplace
mlx/backend/metal/binary.cpp:65
↓ 4 callersFunctionbits
mlx/random.cpp:36
↓ 4 callersFunctionbitwise_invert
mlx/ops.cpp:6274
↓ 4 callersFunctionbitwise_or
mlx/ops.cpp:6232
↓ 4 callersFunctionbitwise_xor
mlx/ops.cpp:6239
↓ 4 callersFunctionceildiv
mlx/backend/metal/kernels/utils.h:307
↓ 4 callersMethodchildren
Return the direct descendants of this Module instance.
python/mlx/nn/layers/base.py:290
↓ 4 callersMethodclear
mlx/backend/metal/kernels/steel/attn/nax.h:566
↓ 4 callersMethodclear_cache
mlx/backend/cuda/allocator.cpp:380
↓ 4 callersFunctionclear_streams
mlx/scheduler.cpp:27
↓ 4 callersMethodcompute_capability_minor
mlx/backend/cuda/device.h:179
↓ 4 callersFunctioncompute_strided_indices
Strided IO for four step FFT
mlx/backend/metal/kernels/fft/readwrite.h:180
↓ 4 callersFunctioncontiguous_reduce
mlx/backend/cpu/reduce.cpp:99
↓ 4 callersFunctionconv2d
2D convolution with a filter */
mlx/ops.cpp:4112
↓ 4 callersFunctionconv_general
General convolution with a filter */
mlx/ops.h:1386
↓ 4 callersFunctionconv_general
General convolution with a filter */
mlx/ops.cpp:4256
↓ 4 callersMethodcreate_additive_causal_mask
(N: int, dtype: mx.Dtype = mx.float32)
python/mlx/nn/layers/transformer.py:96
↓ 4 callersFunctioncross
mlx/linalg.cpp:430
↓ 4 callersFunctioncummax
mlx/ops.cpp:3934
↓ 4 callersFunctioncummin
mlx/ops.cpp:3964
↓ 4 callersFunctioncurrent_binary_dir
mlx/backend/common/utils.cpp:9
↓ 4 callersFunctiondevice_count
mlx/device.cpp:39
↓ 4 callersMethoddivide
python/src/array.cpp:55
↓ 4 callersMethodenqueue
mlx/threadpool.h:60
↓ 4 callersFunctionensure_row_contiguous
mlx/backend/cuda/quantized/quantized_utils.h:7
↓ 4 callersMethodfloat16_t
mlx/distributed/jaccl/lib/jaccl/types.h:39
↓ 4 callersFunctiongather
mlx/ops.cpp:3269
↓ 4 callersFunctiongather_mv
mlx/backend/metal/matmul.cpp:2132
↓ 4 callersFunctiongemm_loop_aligned
mlx/backend/metal/kernels/quantized_utils.h:7
↓ 4 callersMethodget
mlx/distributed/jaccl/lib/jaccl/tcp.h:15
↓ 4 callersFunctionget_bytes_per_pack
mlx/backend/metal/kernels/fp_quantized_nax.h:26
↓ 4 callersFunctionget_cache_memory
mlx/backend/metal/allocator.cpp:261
↓ 4 callersMethodget_cache_memory
mlx/backend/cuda/allocator.cpp:370
↓ 4 callersFunctionget_env_var_or_throw
mlx/distributed/nccl/nccl.cpp:440
↓ 4 callersFunctionget_int_index
python/src/indexing.cpp:74
↓ 4 callersFunctionget_qmm_nax_kernel_wrapped
mlx/backend/metal/quantized.cpp:37
↓ 4 callersFunctionget_slice_int
python/src/indexing.cpp:45
↓ 4 callersFunctiongpu_merge_sort
mlx/backend/metal/sort.cpp:274
↓ 4 callersFunctiongroup_transpose
Transpose from (C_out, H, W, C_in / groups) to (C_in, H, W, C_out / groups).
mlx/backend/cuda/conv.cpp:147
↓ 4 callersFunctionimag
mlx/ops.cpp:6409
↓ 4 callersFunctionindices_or_default
mlx/ops.cpp:61
↓ 4 callersMethodinstantiate
mlx/backend/cuda/utils.cpp:72
↓ 4 callersFunctionis_big_endian
mlx/io/load.cpp:36
↓ 4 callersFunctionis_istream_object
python/src/load.cpp:34
↓ 4 callersMethodis_module
(value)
python/mlx/nn/layers/base.py:227
↓ 4 callersFunctionisnan
mlx/backend/cpu/simd/accelerate_simd.h:147
↓ 4 callersFunctionkindof
mlx/dtype.cpp:91
↓ 4 callersMethodlib_name
mlx/primitives.h:657
↓ 4 callersFunctionload_colocated_library
Firstly, search for the metallib in the same path as this binary
mlx/backend/metal/device.cpp:118
↓ 4 callersFunctionload_safe
mlx/backend/metal/kernels/steel/attn/loader.h:25
↓ 4 callersMethodload_safe
mlx/backend/metal/kernels/gemv_masked.h:105
↓ 4 callersMethodload_unsafe
mlx/backend/metal/kernels/gemv_masked.h:96
↓ 4 callersFunctionlog2
mlx/backend/cpu/simd/base_simd.h:134
↓ 4 callersFunctionmake_contiguous_strides
mlx/backend/common/utils.h:34
↓ 4 callersFunctionmalloc_with_same_layout
mlx/backend/cuda/scaled_dot_product_attention.cpp:40
↓ 4 callersFunctionmeshgrid
mlx/ops.cpp:1145
↓ 4 callersFunctionmlx_atomic_compare_exchange_weak_explicit
mlx/backend/metal/kernels/atomic.h:102
↓ 4 callersMethodmma
(BM, BK) X (BK, BN) multiply accumulate function */
mlx/backend/metal/kernels/steel/attn/mma.h:533
↓ 4 callersFunctionmod_exp
Rader
mlx/backend/metal/fft.cpp:231
↓ 4 callersFunctionn_active_tasks
mlx/scheduler.h:96
↓ 4 callersMethodnext
mlx/backend/metal/kernels/utils.h:210
↓ 4 callersFunctionnext_power_of_2
mlx/utils.h:137
↓ 4 callersFunctionnumber_of_elements
mlx/ops.cpp:6158
↓ 4 callersFunctionoffset_neg_idx
mlx/backend/cpu/indexing.cpp:19
↓ 4 callersMethodout_of_bounds
mlx/backend/metal/kernels/fft/readwrite.h:256
↓ 4 callersFunctionoutput_grid_for_col_reduce
mlx/backend/metal/reduce.cpp:233
↓ 4 callersFunctionparse
Parse the comma separated subscripts into a vector of strings. If the output subscripts are missing they are inferred. For example: "ij,jk -> ik" bec
mlx/einsum.cpp:60
↓ 4 callersFunctionpartition
* Returns a partitioned copy of the flattened array * such that the smaller kth elements are first. **/
mlx/ops.cpp:2595
↓ 4 callersFunctionplan_stockham_fft
mlx/backend/metal/fft.cpp:89
↓ 4 callersMethodpoll
mlx/distributed/jaccl/lib/jaccl/rdma.h:215
↓ 4 callersMethodpost_recv
mlx/distributed/jaccl/lib/jaccl/rdma.h:197
↓ 4 callersMethodpost_send
mlx/distributed/jaccl/lib/jaccl/rdma.h:177
↓ 4 callersFunctionprepare_slice
mlx/backend/common/slicing.cpp:7
← previousnext →601–700 of 4,946, ranked by callers