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
|
import einx
from . import util
import numpy as np
from functools import partial
from typing import Callable, Union
import numpy.typing as npt
_any = any # Is overwritten below
@einx.jit(
trace=lambda t, c: lambda expr_in, tensor_in, expr_out, op, backend=None: c(
expr_in, t(tensor_in), expr_out, op=op
)
)
def reduce_stage3(expr_in, tensor_in, expr_out, op, backend=None):
for root in [expr_in, expr_out]:
for expr in root.all():
if isinstance(expr, einx.expr.stage3.Concatenation):
raise ValueError("Concatenation not allowed")
tensors_out, exprs_out = einx.vmap_with_axis_stage3(
[expr_in], [tensor_in], [expr_out], op, backend=backend
)
return tensors_out[0], exprs_out[0]
@einx.lru_cache
def parse(description, tensor_shape, keepdims=None, cse=True, **parameters):
description, parameters = einx.op.util._clean_description_and_parameters(
description, parameters
)
op = einx.expr.stage1.parse_op(description)
if len(op) == 1:
expr_in = einx.expr.solve(
[einx.expr.Equation(op[0][0], tensor_shape)]
+ [
einx.expr.Equation(k, np.asarray(v)[..., np.newaxis], depth1=None, depth2=None)
for k, v in parameters.items()
],
cse=cse,
cse_in_markers=True,
)[0]
if not _any(isinstance(expr, einx.expr.stage3.Marker) for expr in expr_in.all()):
raise ValueError("No axes are marked for reduction")
# Determine output expressions by removing markers from input expressions
def replace(expr):
if isinstance(expr, einx.expr.stage3.Marker):
if keepdims:
return [einx.expr.stage3.Axis(None, 1)]
else:
return []
expr_out = einx.expr.stage3.replace(expr_in, replace)
else:
if keepdims is not None:
raise ValueError("keepdims cannot be given when using '->'")
if len(op[0]) != 1:
raise ValueError(f"Expected 1 input expression, but got {len(op[0])}")
if len(op[1]) != 1:
raise ValueError(f"Expected 1 output expression, but got {len(op[1])}")
expr_in, expr_out = einx.expr.solve(
[einx.expr.Equation(op[0][0], tensor_shape)]
+ [einx.expr.Equation(op[1][0])]
+ [
einx.expr.Equation(k, np.asarray(v)[..., np.newaxis], depth1=None, depth2=None)
for k, v in parameters.items()
],
cse=cse,
cse_in_markers=True,
)[:2]
# If no axes are marked for reduction in expr_in, mark all axes that
# don't appear in expr_out
if not _any(einx.expr.stage3.is_marked(expr) for expr in expr_in.all()):
axes_names_out = {
axis.name for axis in expr_out.all() if isinstance(axis, einx.expr.stage3.Axis)
}
expr_in = einx.expr.stage3.mark(
expr_in,
lambda expr: isinstance(expr, einx.expr.stage3.Axis)
and expr.name not in axes_names_out,
)
return expr_in, expr_out
@einx.traceback_util.filter
@einx.jit(
trace=lambda t, c: lambda description, tensor, backend=None, **kwargs: c(
description, t(tensor), **kwargs
)
)
def reduce(
description: str,
tensor: einx.Tensor,
op: Union[Callable, str],
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Applies a reduction operation on the given tensors.
The operation reduces all marked axes in the input to a single scalar. It supports
the following shorthand notation:
* When no brackets are found, brackets are placed implicitly around all axes that do not
appear in the output.
Example: ``a b c -> a c`` resolves to ``a [b] c -> a c``.
* When no output is given, it is determined implicitly by removing marked subexpressions
from the input.
Example: ``a [b] c`` resolves to ``a [b] c -> a c``.
Args:
description: Description string for the operation in einx notation.
tensor: Input tensor or tensor factory matching the description string.
op: Backend reduction operation. Is called with ``op(tensor, axis=...)``. If ``op`` is
a string, retrieves the attribute of ``backend`` with the same name.
keepdims: Whether to replace marked expressions with 1s instead of dropping them. Must
be None when ``description`` already contains an output expression. Defaults to None.
backend: Backend to use for all operations. If None, determines the backend from the
input tensors. Defaults to None.
cse: Whether to apply common subexpression elimination to the expressions. Defaults
to True.
graph: Whether to return the graph representation of the operation instead of
computing the result. Defaults to False.
**parameters: Additional parameters that specify values for single axes, e.g. ``a=4``.
Returns:
The result of the reduction operation if ``graph=False``, otherwise the graph
representation of the operation.
Examples:
Compute mean along rows of a matrix:
>>> x = np.random.uniform(size=(16, 20))
>>> einx.mean("a b -> b", x).shape
(20,)
>>> einx.mean("[a] b -> b", x).shape
(20,)
>>> einx.mean("[a] b", x).shape
(20,)
Compute sum along rows of a matrix and broadcast to the original shape:
>>> x = np.random.uniform(size=(16, 20))
>>> einx.sum("[a] b -> a b", x).shape
(16, 20,)
Sum pooling with kernel size 2:
>>> x = np.random.uniform(size=(4, 16, 16, 3))
>>> einx.sum("b (s [s2])... c", x, s2=2).shape
(4, 8, 8, 3)
Compute variance per channel over an image:
>>> x = np.random.uniform(size=(256, 256, 3))
>>> einx.var("[...] c", x).shape
(3,)
"""
expr_in, expr_out = parse(
description, einx.tracer.get_shape(tensor), keepdims=keepdims, cse=cse, **parameters
)
tensor, expr_out = reduce_stage3(expr_in, tensor, expr_out, op=op, backend=backend)
return tensor
reduce.parse = parse
@einx.traceback_util.filter
def sum(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="sum"``"""
return reduce(
description, tensor, op="sum", keepdims=keepdims, backend=backend, cse=cse, **parameters
)
def sum_stage3(*args, **kwargs):
return reduce_stage3(*args, op="sum", **kwargs)
@einx.traceback_util.filter
def mean(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="mean"``"""
return reduce(
description, tensor, op="mean", keepdims=keepdims, backend=backend, cse=cse, **parameters
)
def mean_stage3(*args, **kwargs):
return reduce_stage3(*args, op="mean", **kwargs)
@einx.traceback_util.filter
def var(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="var"``"""
return reduce(
description, tensor, op="var", keepdims=keepdims, backend=backend, cse=cse, **parameters
)
def var_stage3(*args, **kwargs):
return reduce_stage3(*args, op="var", **kwargs)
@einx.traceback_util.filter
def std(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="std"``"""
return reduce(
description, tensor, op="std", keepdims=keepdims, backend=backend, cse=cse, **parameters
)
def std_stage3(*args, **kwargs):
return reduce_stage3(*args, op="std", **kwargs)
@einx.traceback_util.filter
def prod(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="prod"``"""
return reduce(
description, tensor, op="prod", keepdims=keepdims, backend=backend, cse=cse, **parameters
)
def prod_stage3(*args, **kwargs):
return reduce_stage3(*args, op="prod", **kwargs)
@einx.traceback_util.filter
def count_nonzero(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="count_nonzero"``"""
return reduce(
description,
tensor,
op="count_nonzero",
keepdims=keepdims,
backend=backend,
cse=cse,
**parameters,
)
def count_nonzero_stage3(*args, **kwargs):
return reduce_stage3(*args, op="count_nonzero", **kwargs)
@einx.traceback_util.filter
def any(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="any"``"""
return reduce(
description, tensor, op="any", keepdims=keepdims, backend=backend, cse=cse, **parameters
)
def any_stage3(*args, **kwargs):
return reduce_stage3(*args, op="any", **kwargs)
@einx.traceback_util.filter
def all(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
cse: bool = True,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="all"``"""
return reduce(
description, tensor, op="all", keepdims=keepdims, backend=backend, cse=cse, **parameters
)
def all_stage3(*args, **kwargs):
return reduce_stage3(*args, op="all", **kwargs)
@einx.traceback_util.filter
def max(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="max"``"""
return reduce(description, tensor, op="max", keepdims=keepdims, backend=backend, **parameters)
def max_stage3(*args, **kwargs):
return reduce_stage3(*args, op="max", **kwargs)
@einx.traceback_util.filter
def min(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="min"``"""
return reduce(description, tensor, op="min", keepdims=keepdims, backend=backend, **parameters)
def min_stage3(*args, **kwargs):
return reduce_stage3(*args, op="min", **kwargs)
@einx.traceback_util.filter
def logsumexp(
description: str,
tensor: einx.Tensor,
keepdims: Union[bool, None] = None,
backend: Union[einx.Backend, str, None] = None,
**parameters: npt.ArrayLike,
) -> einx.Tensor:
"""Specialization of :func:`einx.reduce` with ``op="logsumexp"``"""
return reduce(
description, tensor, op="logsumexp", keepdims=keepdims, backend=backend, **parameters
)
def logsumexp_stage3(*args, **kwargs):
return reduce_stage3(*args, op="logsumexp", **kwargs)
|