File size: 10,825 Bytes
712dbf0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 |
// Copyright © 2024 Apple Inc.
#include <nanobind/nanobind.h>
#include <nanobind/stl/optional.h>
#include <nanobind/stl/shared_ptr.h>
#include <nanobind/stl/string.h>
#include <nanobind/stl/variant.h>
#include <nanobind/stl/vector.h>
#include "mlx/distributed/distributed.h"
#include "mlx/distributed/ops.h"
#include "python/src/small_vector.h"
#include "python/src/utils.h"
namespace mx = mlx::core;
namespace nb = nanobind;
using namespace nb::literals;
void init_distributed(nb::module_& parent_module) {
auto m = parent_module.def_submodule(
"distributed", "mlx.core.distributed: Communication operations");
nb::class_<mx::distributed::Group>(
m,
"Group",
R"pbcopy(
An :class:`mlx.core.distributed.Group` represents a group of independent mlx
processes that can communicate.
)pbcopy")
.def(
"rank", &mx::distributed::Group::rank, "Get the rank of this process")
.def("size", &mx::distributed::Group::size, "Get the size of the group")
.def(
"split",
&mx::distributed::Group::split,
"color"_a,
"key"_a = -1,
nb::sig("def split(self, color: int, key: int = -1) -> Group"),
R"pbdoc(
Split the group to subgroups based on the provided color.
Processes that use the same color go to the same group. The ``key``
argument defines the rank in the new group. The smaller the key the
smaller the rank. If the key is negative then the rank in the
current group is used.
Args:
color (int): A value to group processes into subgroups.
key (int, optional): A key to optionally change the rank ordering
of the processes.
)pbdoc");
m.def(
"is_available",
&mx::distributed::is_available,
R"pbdoc(
Check if a communication backend is available.
)pbdoc");
m.def(
"init",
&mx::distributed::init,
"strict"_a = false,
"backend"_a = "any",
nb::sig("def init(strict: bool = False, backend: str = 'any') -> Group"),
R"pbdoc(
Initialize the communication backend and create the global communication group.
Example:
.. code:: python
import mlx.core as mx
group = mx.distributed.init(backend="ring")
Args:
strict (bool, optional): If set to False it returns a singleton group
in case ``mx.distributed.is_available()`` returns False otherwise
it throws a runtime error. Default: ``False``
backend (str, optional): Which distributed backend to initialize.
Possible values ``mpi``, ``ring``, ``nccl``, ``any``. If set to ``any`` all
available backends are tried and the first one that succeeds
becomes the global group which will be returned in subsequent
calls. Default: ``any``
Returns:
Group: The group representing all the launched processes.
)pbdoc");
m.def(
"all_sum",
[](const ScalarOrArray& x,
std::optional<mx::distributed::Group> group,
mx::StreamOrDevice s) {
return mx::distributed::all_sum(to_array(x), group, s);
},
"x"_a,
nb::kw_only(),
"group"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def all_sum(x: array, *, group: Optional[Group] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
All reduce sum.
Sum the ``x`` arrays from all processes in the group.
Args:
x (array): Input array.
group (Group): The group of processes that will participate in the
reduction. If set to ``None`` the global group is used. Default:
``None``.
stream (Stream, optional): Stream or device. Defaults to ``None``
in which case the default stream of the default device is used.
Returns:
array: The sum of all ``x`` arrays.
)pbdoc");
m.def(
"all_max",
[](const ScalarOrArray& x,
std::optional<mx::distributed::Group> group,
mx::StreamOrDevice s) {
return mx::distributed::all_max(to_array(x), group, s);
},
"x"_a,
nb::kw_only(),
"group"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def all_max(x: array, *, group: Optional[Group] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
All reduce max.
Find the maximum of the ``x`` arrays from all processes in the group.
Args:
x (array): Input array.
group (Group): The group of processes that will participate in the
reduction. If set to ``None`` the global group is used. Default:
``None``.
stream (Stream, optional): Stream or device. Defaults to ``None``
in which case the default stream of the default device is used.
Returns:
array: The maximum of all ``x`` arrays.
)pbdoc");
m.def(
"all_min",
[](const ScalarOrArray& x,
std::optional<mx::distributed::Group> group,
mx::StreamOrDevice s) {
return mx::distributed::all_min(to_array(x), group, s);
},
"x"_a,
nb::kw_only(),
"group"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def all_min(x: array, *, group: Optional[Group] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
All reduce min.
Find the minimum of the ``x`` arrays from all processes in the group.
Args:
x (array): Input array.
group (Group): The group of processes that will participate in the
reduction. If set to ``None`` the global group is used. Default:
``None``.
stream (Stream, optional): Stream or device. Defaults to ``None``
in which case the default stream of the default device is used.
Returns:
array: The minimum of all ``x`` arrays.
)pbdoc");
m.def(
"all_gather",
[](const ScalarOrArray& x,
std::optional<mx::distributed::Group> group,
mx::StreamOrDevice s) {
return mx::distributed::all_gather(to_array(x), group, s);
},
"x"_a,
nb::kw_only(),
"group"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def all_gather(x: array, *, group: Optional[Group] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Gather arrays from all processes.
Gather the ``x`` arrays from all processes in the group and concatenate
them along the first axis. The arrays should all have the same shape.
Args:
x (array): Input array.
group (Group): The group of processes that will participate in the
gather. If set to ``None`` the global group is used. Default:
``None``.
stream (Stream, optional): Stream or device. Defaults to ``None``
in which case the default stream of the default device is used.
Returns:
array: The concatenation of all ``x`` arrays.
)pbdoc");
m.def(
"send",
[](const ScalarOrArray& x,
int dst,
std::optional<mx::distributed::Group> group,
mx::StreamOrDevice s) {
return mx::distributed::send(to_array(x), dst, group, s);
},
"x"_a,
"dst"_a,
nb::kw_only(),
"group"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def send(x: array, dst: int, *, group: Optional[Group] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Send an array from the current process to the process that has rank
``dst`` in the group.
Args:
x (array): Input array.
dst (int): Rank of the destination process in the group.
group (Group): The group of processes that will participate in the
sned. If set to ``None`` the global group is used. Default:
``None``.
stream (Stream, optional): Stream or device. Defaults to ``None``
in which case the default stream of the default device is used.
Returns:
array: An array identical to ``x`` which when evaluated the send is performed.
)pbdoc");
m.def(
"recv",
&mx::distributed::recv,
"shape"_a,
"dtype"_a,
"src"_a,
nb::kw_only(),
"group"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def recv(shape: Sequence[int], dtype: Dtype, src: int, *, group: Optional[Group] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Recv an array with shape ``shape`` and dtype ``dtype`` from process
with rank ``src``.
Args:
shape (Tuple[int]): The shape of the array we are receiving.
dtype (Dtype): The data type of the array we are receiving.
src (int): Rank of the source process in the group.
group (Group): The group of processes that will participate in the
recv. If set to ``None`` the global group is used. Default:
``None``.
stream (Stream, optional): Stream or device. Defaults to ``None``
in which case the default stream of the default device is used.
Returns:
array: The array that was received from ``src``.
)pbdoc");
m.def(
"recv_like",
[](const ScalarOrArray& x,
int src,
std::optional<mx::distributed::Group> group,
mx::StreamOrDevice s) {
return mx::distributed::recv_like(to_array(x), src, group, s);
},
"x"_a,
"src"_a,
nb::kw_only(),
"group"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def recv_like(x: array, src: int, *, group: Optional[Group] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Recv an array with shape and type like ``x`` from process with rank
``src``.
It is equivalent to calling ``mx.distributed.recv(x.shape, x.dtype, src)``.
Args:
x (array): An array defining the shape and dtype of the array we are
receiving.
src (int): Rank of the source process in the group.
group (Group): The group of processes that will participate in the
recv. If set to ``None`` the global group is used. Default:
``None``.
stream (Stream, optional): Stream or device. Defaults to ``None``
in which case the default stream of the default device is used.
Returns:
array: The array that was received from ``src``.
)pbdoc");
}
|