forked from ml-explore/mlx
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathrandom.cpp
More file actions
505 lines (459 loc) · 18.6 KB
/
random.cpp
File metadata and controls
505 lines (459 loc) · 18.6 KB
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
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
// Copyright © 2023-2024 Apple Inc.
#include <nanobind/nanobind.h>
#include <nanobind/stl/optional.h>
#include <nanobind/stl/variant.h>
#include <nanobind/stl/vector.h>
#include <chrono>
#include "python/src/utils.h"
#include "mlx/ops.h"
#include "mlx/random.h"
namespace mx = mlx::core;
namespace nb = nanobind;
using namespace nb::literals;
class PyKeySequence {
public:
explicit PyKeySequence(uint64_t seed) {
state_.append(mx::random::key(seed));
}
void seed(uint64_t seed) {
state_[0] = mx::random::key(seed);
}
mx::array next() {
auto out = mx::random::split(nb::cast<mx::array>(state_[0]));
state_[0] = out.first;
return out.second;
}
nb::list state() {
return state_;
}
void release() {
nb::gil_scoped_acquire gil;
state_.release().dec_ref();
}
private:
nb::list state_;
};
PyKeySequence& default_key() {
auto get_current_time_seed = []() {
auto now = std::chrono::system_clock::now();
return std::chrono::duration_cast<std::chrono::milliseconds>(
now.time_since_epoch())
.count();
};
static PyKeySequence ks(get_current_time_seed());
return ks;
}
void init_random(nb::module_& parent_module) {
auto m = parent_module.def_submodule(
"random",
"mlx.core.random: functionality related to random number generation");
m.attr("state") = default_key().state();
m.def(
"seed",
[](uint64_t seed) { default_key().seed(seed); },
"seed"_a,
R"pbdoc(
Seed the global PRNG.
Args:
seed (int): Seed for the global PRNG.
)pbdoc");
m.def(
"key",
&mx::random::key,
"seed"_a,
R"pbdoc(
Get a PRNG key from a seed.
Args:
seed (int): Seed for the PRNG.
Returns:
array: The PRNG key array.
)pbdoc");
m.def(
"split",
nb::overload_cast<const mx::array&, int, mx::StreamOrDevice>(
&mx::random::split),
"key"_a,
"num"_a = 2,
"stream"_a = nb::none(),
nb::sig(
"def split(key: array, num: int = 2, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Split a PRNG key into sub keys.
Args:
key (array): Input key to split.
num (int, optional): Number of sub keys. Default: ``2``.
Returns:
array: The array of sub keys with ``num`` as its first dimension.
)pbdoc");
m.def(
"uniform",
[](const ScalarOrArray& low,
const ScalarOrArray& high,
const mx::Shape& shape,
std::optional<mx::Dtype> type,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
return mx::random::uniform(
to_array(low),
to_array(high),
shape,
type.value_or(mx::float32),
key,
s);
},
"low"_a = 0,
"high"_a = 1,
"shape"_a = mx::Shape{},
"dtype"_a.none() = mx::float32,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def uniform(low: Union[scalar, array] = 0, high: Union[scalar, array] = 1, shape: Sequence[int] = [], dtype: Optional[Dtype] = float32, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Generate uniformly distributed random numbers.
The values are sampled uniformly in the half-open interval ``[low, high)``.
The lower and upper bound can be scalars or arrays and must be
broadcastable to ``shape``.
Args:
low (scalar or array, optional): Lower bound of the distribution.
Default: ``0``.
high (scalar or array, optional): Upper bound of the distribution.
Default: ``1``.
shape (list(int), optional): Shape of the output. Default:``()``.
dtype (Dtype, optional): Type of the output. Default: ``float32``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array: The output array random values.
)pbdoc");
m.def(
"normal",
[](const mx::Shape& shape,
std::optional<mx::Dtype> type,
float loc,
float scale,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
return mx::random::normal(
shape, type.value_or(mx::float32), loc, scale, key, s);
},
"shape"_a = mx::Shape{},
"dtype"_a.none() = mx::float32,
"loc"_a = 0.0,
"scale"_a = 1.0,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def normal(shape: Sequence[int] = [], dtype: Optional[Dtype] = float32, loc: float = 0.0, scale: float = 1.0, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Generate normally distributed random numbers.
Args:
shape (list(int), optional): Shape of the output. Default is ``()``.
dtype (Dtype, optional): Type of the output. Default is ``float32``.
loc (float, optional): Mean of the distribution. Default is ``0.0``.
scale (float, optional): Standard deviation of the distribution. Default is ``1.0``.
key (array, optional): A PRNG key. Default: None.
Returns:
array: The output array of random values.
)pbdoc");
m.def(
"multivariate_normal",
[](const mx::array& mean,
const mx::array& cov,
const mx::Shape& shape,
std::optional<mx::Dtype> type,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
return mx::random::multivariate_normal(
mean, cov, shape, type.value_or(mx::float32), key, s);
},
"mean"_a,
"cov"_a,
"shape"_a = mx::Shape{},
"dtype"_a.none() = mx::float32,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def multivariate_normal(mean: array, cov: array, shape: Sequence[int] = [], dtype: Optional[Dtype] = float32, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Generate jointly-normal random samples given a mean and covariance.
The matrix ``cov`` must be positive semi-definite. The behavior is
undefined if it is not. The only supported ``dtype`` is ``float32``.
Args:
mean (array): array of shape ``(..., n)``, the mean of the
distribution.
cov (array): array of shape ``(..., n, n)``, the covariance
matrix of the distribution. The batch shape ``...`` must be
broadcast-compatible with that of ``mean``.
shape (list(int), optional): The output shape must be
broadcast-compatible with ``mean.shape[:-1]`` and ``cov.shape[:-2]``.
If empty, the result shape is determined by broadcasting the batch
shapes of ``mean`` and ``cov``. Default: ``[]``.
dtype (Dtype, optional): The output type. Default: ``float32``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array: The output array of random values.
)pbdoc");
m.def(
"randint",
[](const ScalarOrArray& low,
const ScalarOrArray& high,
const mx::Shape& shape,
std::optional<mx::Dtype> type,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
return mx::random::randint(
to_array(low),
to_array(high),
shape,
type.value_or(mx::int32),
key,
s);
},
"low"_a,
"high"_a,
"shape"_a = mx::Shape{},
"dtype"_a.none() = mx::int32,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def randint(low: Union[scalar, array], high: Union[scalar, array], shape: Sequence[int] = [], dtype: Optional[Dtype] = int32, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Generate random integers from the given interval.
The values are sampled with equal probability from the integers in
half-open interval ``[low, high)``. The lower and upper bound can be
scalars or arrays and must be broadcastable to ``shape``.
Args:
low (scalar or array): Lower bound of the interval.
high (scalar or array): Upper bound of the interval.
shape (list(int), optional): Shape of the output. Default: ``()``.
dtype (Dtype, optional): Type of the output. Default: ``int32``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array: The array of random integers.
)pbdoc");
m.def(
"bernoulli",
[](const ScalarOrArray& p_,
const std::optional<mx::Shape> shape,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
auto p = to_array(p_);
if (shape.has_value()) {
return mx::random::bernoulli(p, shape.value(), key, s);
} else {
return mx::random::bernoulli(p, key, s);
}
},
"p"_a = 0.5,
"shape"_a = nb::none(),
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def bernoulli(p: Union[scalar, array] = 0.5, shape: Optional[Sequence[int]] = None, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Generate Bernoulli random values.
The values are sampled from the bernoulli distribution with parameter
``p``. The parameter ``p`` can be a :obj:`float` or :obj:`array` and
must be broadcastable to ``shape``.
Args:
p (float or array, optional): Parameter of the Bernoulli
distribution. Default: ``0.5``.
shape (list(int), optional): Shape of the output.
Default: ``p.shape``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array: The array of random integers.
)pbdoc");
m.def(
"truncated_normal",
[](const ScalarOrArray& lower_,
const ScalarOrArray& upper_,
const std::optional<mx::Shape> shape_,
std::optional<mx::Dtype> type,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
auto lower = to_array(lower_);
auto upper = to_array(upper_);
auto t = type.value_or(mx::float32);
if (shape_.has_value()) {
return mx::random::truncated_normal(
lower, upper, shape_.value(), t, key, s);
} else {
return mx::random::truncated_normal(lower, upper, t, key, s);
}
},
"lower"_a,
"upper"_a,
"shape"_a = nb::none(),
"dtype"_a.none() = mx::float32,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def truncated_normal(lower: Union[scalar, array], upper: Union[scalar, array], shape: Optional[Sequence[int]] = None, dtype: Optional[Dtype] = float32, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Generate values from a truncated normal distribution.
The values are sampled from the truncated normal distribution
on the domain ``(lower, upper)``. The bounds ``lower`` and ``upper``
can be scalars or arrays and must be broadcastable to ``shape``.
Args:
lower (scalar or array): Lower bound of the domain.
upper (scalar or array): Upper bound of the domain.
shape (list(int), optional): The shape of the output.
Default:``()``.
dtype (Dtype, optional): The data type of the output.
Default: ``float32``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array: The output array of random values.
)pbdoc");
m.def(
"gumbel",
[](const mx::Shape& shape,
std::optional<mx::Dtype> type,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
return mx::random::gumbel(shape, type.value_or(mx::float32), key, s);
},
"shape"_a = mx::Shape{},
"dtype"_a.none() = mx::float32,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def gumbel(shape: Sequence[int] = [], dtype: Optional[Dtype] = float32, key: Union[None, Stream, Device] = None, stream: Optional[array] = None) -> array"),
R"pbdoc(
Sample from the standard Gumbel distribution.
The values are sampled from a standard Gumbel distribution
which CDF ``exp(-exp(-x))``.
Args:
shape (list(int)): The shape of the output.
dtype (Dtype, optional): The data type of the output.
Default: ``float32``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array:
The :class:`array` with shape ``shape`` and distributed according
to the Gumbel distribution.
)pbdoc");
m.def(
"categorical",
[](const mx::array& logits,
int axis,
const std::optional<mx::Shape> shape,
const std::optional<int> num_samples,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
if (shape.has_value() && num_samples.has_value()) {
throw std::invalid_argument(
"[categorical] At most one of shape or num_samples can be specified.");
} else if (shape.has_value()) {
return mx::random::categorical(logits, axis, shape.value(), key, s);
} else if (num_samples.has_value()) {
return mx::random::categorical(
logits, axis, num_samples.value(), key, s);
} else {
return mx::random::categorical(logits, axis, key, s);
}
},
"logits"_a,
"axis"_a = -1,
"shape"_a = nb::none(),
"num_samples"_a = nb::none(),
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def categorical(logits: array, axis: int = -1, shape: Optional[Sequence[int]] = None, num_samples: Optional[int] = None, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Sample from a categorical distribution.
The values are sampled from the categorical distribution specified by
the unnormalized values in ``logits``. Note, at most one of ``shape``
or ``num_samples`` can be specified. If both are ``None``, the output
has the same shape as ``logits`` with the ``axis`` dimension removed.
Args:
logits (array): The *unnormalized* categorical distribution(s).
axis (int, optional): The axis which specifies the distribution.
Default: ``-1``.
shape (list(int), optional): The shape of the output. This must
be broadcast compatable with ``logits.shape`` with the ``axis``
dimension removed. Default: ``None``
num_samples (int, optional): The number of samples to draw from each
of the categorical distributions in ``logits``. The output will have
``num_samples`` in the last dimension. Default: ``None``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array: The ``shape``-sized output array with type ``uint32``.
)pbdoc");
m.def(
"laplace",
[](const mx::Shape& shape,
std::optional<mx::Dtype> type,
float loc,
float scale,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
return mx::random::laplace(
shape, type.value_or(mx::float32), loc, scale, key, s);
},
"shape"_a = mx::Shape{},
"dtype"_a.none() = mx::float32,
"loc"_a = 0.0,
"scale"_a = 1.0,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def laplace(shape: Sequence[int] = [], dtype: Optional[Dtype] = float32, loc: float = 0.0, scale: float = 1.0, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Sample numbers from a Laplace distribution.
Args:
shape (list(int), optional): Shape of the output. Default: ``()``.
dtype (Dtype, optional): Type of the output. Default: ``float32``.
loc (float, optional): Mean of the distribution. Default: ``0.0``.
scale (float, optional): The scale "b" of the Laplace distribution.
Default:``1.0``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array: The output array of random values.
)pbdoc");
m.def(
"permutation",
[](const std::variant<nb::int_, mx::array>& x,
int axis,
const std::optional<mx::array>& key_,
mx::StreamOrDevice s) {
auto key = key_ ? key_.value() : default_key().next();
if (auto pv = std::get_if<nb::int_>(&x); pv) {
return mx::random::permutation(nb::cast<int>(*pv), key, s);
} else {
return mx::random::permutation(std::get<mx::array>(x), axis, key, s);
}
},
"x"_a,
"axis"_a = 0,
"key"_a = nb::none(),
"stream"_a = nb::none(),
nb::sig(
"def permutation(x: Union[int, array], axis: int = 0, key: Optional[array] = None, stream: Union[None, Stream, Device] = None) -> array"),
R"pbdoc(
Generate a random permutation or permute the entries of an array.
Args:
x (int or array, optional): If an integer is provided a random
permtuation of ``mx.arange(x)`` is returned. Otherwise the entries
of ``x`` along the given axis are randomly permuted.
axis (int, optional): The axis to permute along. Default: ``0``.
key (array, optional): A PRNG key. Default: ``None``.
Returns:
array:
The generated random permutation or randomly permuted input array.
)pbdoc");
// Register static Python object cleanup before the interpreter exits
auto atexit = nb::module_::import_("atexit");
atexit.attr("register")(nb::cpp_function([]() { default_key().release(); }));
}