File size: 16,075 Bytes
46a43b0 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 | from __future__ import annotations
from typing import NamedTuple
import numpy
import cupy
import math
from cupy import _core
def delete(arr, indices, axis=None):
"""
Delete values from an array along the specified axis.
Args:
arr (cupy.ndarray):
Values are deleted from a copy of this array.
indices (slice, int or array of ints):
These indices correspond to values that will be deleted from the
copy of `arr`.
Boolean indices are treated as a mask of elements to remove.
axis (int or None):
The axis along which `indices` correspond to values that will be
deleted. If `axis` is not given, `arr` will be flattened.
Returns:
cupy.ndarray:
A copy of `arr` with values specified by `indices` deleted along
`axis`.
.. warning:: This function may synchronize the device.
.. seealso:: :func:`numpy.delete`.
"""
if axis is None:
arr = arr.ravel()
if isinstance(indices, cupy.ndarray) and indices.dtype == cupy.bool_:
return arr[~indices]
mask = cupy.ones(arr.size, dtype=bool)
mask[indices] = False
return arr[mask]
else:
if isinstance(indices, cupy.ndarray) and indices.dtype == cupy.bool_:
return cupy.compress(~indices, arr, axis=axis)
mask = cupy.ones(arr.shape[axis], dtype=bool)
mask[indices] = False
return cupy.compress(mask, arr, axis=axis)
# TODO(okuta): Implement insert
def append(arr, values, axis=None):
"""
Append values to the end of an array.
Args:
arr (array_like):
Values are appended to a copy of this array.
values (array_like):
These values are appended to a copy of ``arr``. It must be of the
correct shape (the same shape as ``arr``, excluding ``axis``). If
``axis`` is not specified, ``values`` can be any shape and will be
flattened before use.
axis (int or None):
The axis along which ``values`` are appended. If ``axis`` is not
given, both ``arr`` and ``values`` are flattened before use.
Returns:
cupy.ndarray:
A copy of ``arr`` with ``values`` appended to ``axis``. Note that
``append`` does not occur in-place: a new array is allocated and
filled. If ``axis`` is None, ``out`` is a flattened array.
.. seealso:: :func:`numpy.append`
"""
# TODO(asi1024): Implement fast path for scalar inputs.
arr = cupy.asarray(arr)
values = cupy.asarray(values)
if axis is None:
return _core.concatenate_method(
(arr.ravel(), values.ravel()), 0).ravel()
return _core.concatenate_method((arr, values), axis)
_resize_kernel = _core.ElementwiseKernel(
'raw T x, int64 size', 'T y',
'y = x[i % size]',
'cupy_resize',
)
def resize(a, new_shape):
"""Return a new array with the specified shape.
If the new array is larger than the original array, then the new
array is filled with repeated copies of ``a``. Note that this behavior
is different from a.resize(new_shape) which fills with zeros instead
of repeated copies of ``a``.
Args:
a (array_like): Array to be resized.
new_shape (int or tuple of int): Shape of resized array.
Returns:
cupy.ndarray:
The new array is formed from the data in the old array, repeated
if necessary to fill out the required number of elements. The
data are repeated in the order that they are stored in memory.
.. seealso:: :func:`numpy.resize`
"""
if numpy.isscalar(a):
return cupy.full(new_shape, a)
a = cupy.asarray(a)
if a.size == 0:
return cupy.zeros(new_shape, dtype=a.dtype)
out = cupy.empty(new_shape, a.dtype)
_resize_kernel(a, a.size, out)
return out
_first_nonzero_krnl = _core.ReductionKernel(
'T data, int64 len',
'int64 y',
'data == T(0) ? len : _j',
'min(a, b)',
'y = a',
'len',
'first_nonzero'
)
def trim_zeros(filt, trim='fb'):
"""Trim the leading and/or trailing zeros from a 1-D array or sequence.
Returns the trimmed array
Args:
filt(cupy.ndarray): Input array
trim(str, optional):
'fb' default option trims the array from both sides.
'f' option trim zeros from front.
'b' option trim zeros from back.
Returns:
cupy.ndarray: trimmed input
.. seealso:: :func:`numpy.trim_zeros`
"""
if filt.ndim == 0:
return filt
if filt.ndim > 1:
raise NotImplementedError('Multi-dimensional trim is not supported')
start = 0
end = filt.size
trim = trim.upper()
if 'F' in trim:
start = _first_nonzero_krnl(filt, filt.size).item()
if 'B' in trim:
end = filt.size - _first_nonzero_krnl(filt[::-1], filt.size).item()
return filt[start:end]
@_core.fusion.fuse()
def _unique_update_mask_equal_nan(mask, x0):
mask1 = cupy.logical_not(cupy.isnan(x0))
mask[:] = cupy.logical_and(mask, mask1)
def unique(ar, return_index=False, return_inverse=False,
return_counts=False, axis=None, *, equal_nan=True):
"""Find the unique elements of an array.
Returns the sorted unique elements of an array. There are three optional
outputs in addition to the unique elements:
* the indices of the input array that give the unique values
* the indices of the unique array that reconstruct the input array
* the number of times each unique value comes up in the input array
Args:
ar(array_like): Input array. This will be flattened if it is not
already 1-D.
return_index(bool, optional): If True, also return the indices of `ar`
(along the specified axis, if provided, or in the flattened array)
that result in the unique array.
return_inverse(bool, optional): If True, also return the indices of the
unique array (for the specified axis, if provided) that can be used
to reconstruct `ar`.
return_counts(bool, optional): If True, also return the number of times
each unique item appears in `ar`.
axis(int or None, optional): The axis to operate on. If None, ar will
be flattened. If an integer, the subarrays indexed by the given
axis will be flattened and treated as the elements of a 1-D array
with the dimension of the given axis, see the notes for more
details. The default is None.
equal_nan(bool, optional): If True, collapse multiple NaN values in the
return array into one.
Returns:
cupy.ndarray or tuple:
If there are no optional outputs, it returns the
:class:`cupy.ndarray` of the sorted unique values. Otherwise, it
returns the tuple which contains the sorted unique values and
following.
* The indices of the first occurrences of the unique values in the
original array. Only provided if `return_index` is True.
* The indices to reconstruct the original array from the
unique array. Only provided if `return_inverse` is True.
* The number of times each of the unique values comes up in the
original array. Only provided if `return_counts` is True.
Notes:
When an axis is specified the subarrays indexed by the axis are sorted.
This is done by making the specified axis the first dimension of the
array (move the axis to the first dimension to keep the order of the
other axes) and then flattening the subarrays in C order.
.. warning::
This function may synchronize the device.
.. seealso:: :func:`numpy.unique`
"""
if axis is None:
ret = _unique_1d(ar, return_index=return_index,
return_inverse=return_inverse,
return_counts=return_counts,
equal_nan=equal_nan, inverse_shape=ar.shape)
return ret
ar = cupy.moveaxis(ar, axis, 0)
# The array is reshaped into a contiguous 2D array
orig_shape = ar.shape
idx = cupy.arange(0, orig_shape[0], dtype=cupy.intp)
ar = ar.reshape(orig_shape[0], math.prod(orig_shape[1:]))
ar = cupy.ascontiguousarray(ar)
is_unsigned = cupy.issubdtype(ar.dtype, cupy.unsignedinteger)
is_complex = cupy.iscomplexobj(ar)
ar_cmp = ar
if is_unsigned:
ar_cmp = ar.astype(cupy.intp)
def compare_axis_elems(idx1, idx2):
left, right = ar_cmp[idx1], ar_cmp[idx2]
comp = cupy.trim_zeros(left - right, 'f')
if comp.shape[0] > 0:
diff = comp[0]
if is_complex and cupy.isnan(diff):
return True
return diff < 0
return False
# The array is sorted lexicographically using the first item of each
# element on the axis
sorted_indices = cupy.empty(orig_shape[0], dtype=cupy.intp)
queue = [(idx.tolist(), 0)]
while queue != []:
current, off = queue.pop(0)
if current == []:
continue
mid_elem = current[0]
left = []
right = []
for i in range(1, len(current)):
if compare_axis_elems(current[i], mid_elem):
left.append(current[i])
else:
right.append(current[i])
elem_pos = off + len(left)
queue.append((left, off))
queue.append((right, elem_pos + 1))
sorted_indices[elem_pos] = mid_elem
ar = ar[sorted_indices]
if ar.size > 0:
mask = cupy.empty(ar.shape, dtype=cupy.bool_)
mask[:1] = True
mask[1:] = ar[1:] != ar[:-1]
mask = cupy.any(mask, axis=1)
else:
# If empty, then the mask should grab the first empty array as the
# unique one
mask = cupy.ones((ar.shape[0]), dtype=cupy.bool_)
mask[1:] = False
# Index the input array with the unique elements and reshape it into the
# original size and dimension order
ar = ar[mask]
ar = ar.reshape(mask.sum().item(), *orig_shape[1:])
ar = cupy.moveaxis(ar, 0, axis)
ret = ar,
if return_index:
ret += sorted_indices[mask],
if return_inverse:
imask = cupy.cumsum(mask) - 1
inv_idx = cupy.empty(mask.shape, dtype=cupy.intp)
inv_idx[sorted_indices] = imask
ret += inv_idx,
if return_counts:
nonzero = cupy.nonzero(mask)[0] # may synchronize
idx = cupy.empty((nonzero.size + 1,), nonzero.dtype)
idx[:-1] = nonzero
idx[-1] = mask.size
ret += idx[1:] - idx[:-1],
if len(ret) == 1:
ret = ret[0]
return ret
def _unique_1d(ar, return_index=False, return_inverse=False,
return_counts=False, equal_nan=True, inverse_shape=None):
ar = cupy.asarray(ar).flatten()
if return_index or return_inverse:
perm = ar.argsort()
aux = ar[perm]
else:
ar.sort()
aux = ar
mask = cupy.empty(aux.shape, dtype=cupy.bool_)
mask[:1] = True
mask[1:] = aux[1:] != aux[:-1]
if equal_nan:
_unique_update_mask_equal_nan(mask[1:], aux[:-1])
ret = aux[mask]
if not return_index and not return_inverse and not return_counts:
return ret
ret = ret,
if return_index:
ret += perm[mask],
if return_inverse:
imask = cupy.cumsum(mask) - 1
inv_idx = cupy.empty(mask.shape, dtype=cupy.intp)
inv_idx[perm] = imask
ret += inv_idx.reshape(inverse_shape),
if return_counts:
nonzero = cupy.nonzero(mask)[0] # may synchronize
idx = cupy.empty((nonzero.size + 1,), nonzero.dtype)
idx[:-1] = nonzero
idx[-1] = mask.size
ret += idx[1:] - idx[:-1],
return ret
# Array API compatible unique_XXX wrappers
class UniqueAllResult(NamedTuple):
values: cupy.ndarray
indices: cupy.ndarray
inverse_indices: cupy.ndarray
counts: cupy.ndarray
class UniqueCountsResult(NamedTuple):
values: cupy.ndarray
counts: cupy.ndarray
class UniqueInverseResult(NamedTuple):
values: cupy.ndarray
inverse_indices: cupy.ndarray
def unique_all(x):
"""
Find the unique elements of an array, and counts, inverse and indices.
This function is an Array API compatible alternative to:
>>> x = cupy.array([1, 1, 2])
>>> np.unique(x, return_index=True, return_inverse=True,
... return_counts=True, equal_nan=False)
(array([1, 2]), array([0, 2]), array([0, 0, 1]), array([2, 1]))
Parameters
----------
x : ndarray
Input array. It will be flattened if it is not already 1-D.
Returns
-------
out : namedtuple
The result containing:
* values - The unique elements of an input array.
* indices - The first occurring indices for each unique element.
* inverse_indices - The indices from the set of unique elements
that reconstruct `x`.
* counts - The corresponding counts for each unique element.
See Also
--------
unique : Find the unique elements of an array.
numpy.unique_all
"""
result = unique(
x,
return_index=True,
return_inverse=True,
return_counts=True,
equal_nan=False
)
return UniqueAllResult(*result)
def unique_counts(x):
"""
Find the unique elements and counts of an input array `x`.
This function is an Array API compatible alternative to:
>>> x = cupy.array([1, 1, 2])
>>> cupy.unique(x, return_counts=True, equal_nan=False)
(array([1, 2]), array([2, 1]))
Parameters
----------
x : ndarray
Input array. It will be flattened if it is not already 1-D.
Returns
-------
out : namedtuple
The result containing:
* values - The unique elements of an input array.
* counts - The corresponding counts for each unique element.
See Also
--------
unique : Find the unique elements of an array.
np.unique_counts
"""
result = unique(
x,
return_index=False,
return_inverse=False,
return_counts=True,
equal_nan=False
)
return UniqueCountsResult(*result)
def unique_inverse(x):
"""
Find the unique elements of `x` and indices to reconstruct `x`.
This function is Array API compatible alternative to:
>>> x = cupy.array([1, 1, 2])
>>> cupy.unique(x, return_inverse=True, equal_nan=False)
(array([1, 2]), array([0, 0, 1]))
Parameters
----------
x : ndarray
Input array. It will be flattened if it is not already 1-D.
Returns
-------
out : namedtuple
The result containing:
* values - The unique elements of an input array.
* inverse_indices - The indices from the set of unique elements
that reconstruct `x`.
See Also
--------
unique : Find the unique elements of an array.
numpy.unique_inverse
"""
result = unique(
x,
return_index=False,
return_inverse=True,
return_counts=False,
equal_nan=False
)
return UniqueInverseResult(*result)
def unique_values(x):
"""
Returns the unique elements of an input array `x`.
This function is Array API compatible alternative to:
>>> x = cupy.array([1, 1, 2])
>>> cupy.unique(x, equal_nan=False)
array([1, 2])
Parameters
----------
x : ndarray
Input array. It will be flattened if it is not already 1-D.
Returns
-------
out : ndarray
The unique elements of an input array.
See Also
--------
unique : Find the unique elements of an array.
numpy.unique_values
"""
return unique(
x,
return_index=False,
return_inverse=False,
return_counts=False,
equal_nan=False
)
|