File size: 12,158 Bytes
f14e74e |
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 |
# Copyright © 2023 Apple Inc.
import math
import mlx.core as mx
from mlx.nn.layers.base import Module
def cross_entropy(
logits: mx.array,
targets: mx.array,
weights: mx.array = None,
axis: int = -1,
label_smoothing: float = 0.0,
reduction: str = "none",
) -> mx.array:
"""
Computes the cross entropy loss.
Args:
logits (array): The unnormalized predicted logits.
targets (array): The target values, as class indices.
weights (array, optional): Weights for each target. Default: ``None``.
axis (int, optional): The axis over which to compute softmax. Default: ``-1``.
label_smoothing (float, optional): Label smoothing factor. Default: ``0``.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: The computed cross entropy loss.
"""
if label_smoothing < 0 or label_smoothing >= 1:
raise ValueError(f"Label smoothing must in [0, 1), got {label_smoothing}.")
score = mx.take_along_axis(logits, targets[..., None], axis).squeeze(-1)
logsumexp_logits = mx.logsumexp(logits, axis=axis)
if label_smoothing > 0:
# Adjust the true class score with label smoothing
adjusted_score = (1 - label_smoothing) * score
# Calculate the mean logit across the classes for smoothed loss
mean_logits = logits.mean(axis=axis)
smoothed_loss = -mean_logits * label_smoothing
# Combine the adjusted score and smoothed loss with the logsumexp logits
loss = logsumexp_logits - adjusted_score + smoothed_loss
else:
loss = logsumexp_logits - score
# Apply weights if provided
if weights is not None:
if weights.shape != targets.shape:
raise ValueError(
f"Weights with shape {weights.shape} is not the same as "
f"targets with shape {targets.shape}."
)
loss *= weights
# Apply reduction
return _reduce(loss, reduction)
def binary_cross_entropy(
logits: mx.array, targets: mx.array, reduction: str = "none"
) -> mx.array:
"""
Computes the binary cross entropy loss.
Args:
logits (array): The unnormalized (pre-sigmoid) predicted logits.
targets (array): The binary target values in {0, 1}.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: The computed binary cross entropy loss.
Examples:
>>> import mlx.core as mx
>>> import mlx.nn as nn
>>> inputs = mx.array([0.105361, 0.223144, 1.20397, 0.916291])
>>> targets = mx.array([0, 0, 1, 1])
>>> loss = nn.losses.binary_cross_entropy(inputs, targets, "mean")
>>> loss
array([0.612192], dtype=float32)
"""
loss = mx.logaddexp(0.0, logits) - targets * logits
return _reduce(loss, reduction)
def l1_loss(
predictions: mx.array, targets: mx.array, reduction: str = "mean"
) -> mx.array:
"""
Computes the L1 loss.
Args:
predictions (array): The predicted values.
targets (array): The target values.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'mean'``.
Returns:
array: The computed L1 loss.
"""
if predictions.shape != targets.shape:
raise ValueError(
f"Predictions shape {predictions.shape} does not match "
f"targets shape {targets.shape}."
)
loss = mx.abs(predictions - targets)
return _reduce(loss, reduction)
def mse_loss(
predictions: mx.array, targets: mx.array, reduction: str = "mean"
) -> mx.array:
"""
Computes the mean squared error loss.
Args:
predictions (array): The predicted values.
targets (array): The target values.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'mean'``.
Returns:
array: The computed mean squared error loss.
"""
if predictions.shape != targets.shape:
raise ValueError(
f"Predictions shape {predictions.shape} does not match "
f"targets shape {targets.shape}."
)
loss = mx.square(predictions - targets)
return _reduce(loss, reduction)
def nll_loss(
inputs: mx.array, targets: mx.array, axis: int = -1, reduction: str = "none"
) -> mx.array:
"""
Computes the negative log likelihood loss.
Args:
inputs (array): The predicted distribution in log space.
targets (array): The target values.
axis (int, optional): The distribution axis. Default: ``-1``.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: The computed NLL loss.
"""
loss = -mx.take_along_axis(inputs, targets[..., None], axis).squeeze(-1)
return _reduce(loss, reduction)
def kl_div_loss(
inputs: mx.array, targets: mx.array, axis: int = -1, reduction: str = "none"
) -> mx.array:
"""
Computes the Kullback-Leibler divergence loss.
Computes the following when ``reduction == 'none'``:
.. code-block:: python
mx.exp(targets) * (targets - inputs).sum(axis)
Args:
inputs (array): Log probabilities for the predicted distribution.
targets (array): Log probabilities for the target distribution.
axis (int, optional): The distribution axis. Default: ``-1``.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: The computed Kullback-Leibler divergence loss.
"""
loss = mx.sum(mx.exp(targets) * (targets - inputs), axis)
return _reduce(loss, reduction)
def smooth_l1_loss(
predictions: mx.array, targets: mx.array, beta: float = 1.0, reduction: str = "mean"
) -> mx.array:
r"""
Computes the smooth L1 loss.
The smooth L1 loss is a variant of the L1 loss which replaces the absolute
difference with a squared difference when the absolute difference is less
than ``beta``.
The formula for the smooth L1 Loss is:
.. math::
l =
\begin{cases}
0.5 (x - y)^2, & \text{ if } & (x - y) < \beta \\
|x - y| - 0.5 \beta, & & \text{otherwise}
\end{cases}
Args:
predictions (array): Predicted values.
targets (array): Ground truth values.
beta (float, optional): The threshold after which the loss changes
from the squared to the absolute difference. Default: ``1.0``.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'mean'``.
Returns:
array: The computed smooth L1 loss.
"""
if predictions.shape != targets.shape:
raise ValueError(
f"Predictions shape {predictions.shape} does not match "
f"targets shape {targets.shape}."
)
diff = predictions - targets
loss = mx.where(
diff < beta, 0.5 * mx.square(diff) / beta, mx.abs(diff) - 0.5 * beta
)
return _reduce(loss, reduction)
def triplet_loss(
anchors: mx.array,
positives: mx.array,
negatives: mx.array,
axis: int = -1,
p: int = 2,
margin: float = 1.0,
eps: float = 1e-6,
reduction: str = "none",
) -> mx.array:
r"""
Computes the triplet loss for a set of anchor, positive, and negative samples.
Margin is represented with alpha in the math section.
.. math::
L_{\text{triplet}} = \max\left(\|A - P\|_p - \|A - N\|_p + \alpha, 0\right)
Args:
anchors (array): The anchor samples.
positives (array): The positive samples.
negatives (array): The negative samples.
axis (int, optional): The distribution axis. Default: ``-1``.
p (int, optional): The norm degree for pairwise distance. Default: ``2``.
margin (float, optional): Margin for the triplet loss. Defaults to ``1.0``.
eps (float, optional): Small positive constant to prevent numerical instability. Defaults to ``1e-6``.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: Computed triplet loss. If reduction is "none", returns a tensor of the same shape as input;
if reduction is "mean" or "sum", returns a scalar tensor.
"""
loss = mx.maximum(
mx.sqrt(mx.power(anchors - positives, p).sum(axis) + eps)
- mx.sqrt(mx.power(anchors - negatives, p).sum(axis) + eps)
+ margin,
0,
)
return _reduce(loss, reduction)
def _reduce(loss: mx.array, reduction: str = "none"):
if reduction == "mean":
return mx.mean(loss)
elif reduction == "sum":
return mx.sum(loss)
elif reduction == "none":
return loss
else:
raise ValueError("Invalid reduction. Must be 'none', 'mean', or 'sum'.")
def hinge_loss(
inputs: mx.array, targets: mx.array, reduction: str = "none"
) -> mx.array:
r"""
Computes the hinge loss between inputs and targets.
.. math::
\text{hinge}(y, y_{\text{pred}}) = \max(0, 1 - y \cdot y_{\text{pred}})
Args:
inputs (array): The predicted values.
targets (array): The target values. They should be -1 or 1.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: The computed hinge loss.
"""
loss = mx.maximum(1 - inputs * targets, 0)
return _reduce(loss, reduction)
def huber_loss(
inputs: mx.array, targets: mx.array, delta: float = 1.0, reduction: str = "none"
) -> mx.array:
r"""
Computes the Huber loss between inputs and targets.
.. math::
L_{\delta}(a) =
\left\{ \begin{array}{ll}
\frac{1}{2} a^2 & \text{for } |a| \leq \delta, \\
\delta \left( |a| - \frac{1}{2} \delta \right) & \text{otherwise.}
\end{array} \right.
Args:
inputs (array): The predicted values.
targets (array): The target values.
delta (float, optional): The threshold at which to change between L1 and L2 loss.
Default: ``1.0``.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: The computed Huber loss.
"""
errors = inputs - targets
abs_errors = mx.abs(errors)
quadratic = mx.minimum(abs_errors, delta)
linear = abs_errors - quadratic
loss = 0.5 * quadratic**2 + delta * linear
return _reduce(loss, reduction)
def log_cosh_loss(
inputs: mx.array, targets: mx.array, reduction: str = "none"
) -> mx.array:
r"""
Computes the log cosh loss between inputs and targets.
Logcosh acts like L2 loss for small errors, ensuring stable gradients,
and like the L1 loss for large errors, reducing sensitivity to outliers. This
dual behavior offers a balanced, robust approach for regression tasks.
.. math::
\text{logcosh}(y_{\text{true}}, y_{\text{pred}}) =
\frac{1}{n} \sum_{i=1}^{n}
\log(\cosh(y_{\text{pred}}^{(i)} - y_{\text{true}}^{(i)}))
Args:
inputs (array): The predicted values.
targets (array): The target values.
reduction (str, optional): Specifies the reduction to apply to the output:
``'none'`` | ``'mean'`` | ``'sum'``. Default: ``'none'``.
Returns:
array: The computed log cosh loss.
"""
errors = inputs - targets
loss = mx.logaddexp(errors, -errors) - math.log(2)
return _reduce(loss, reduction)
|