xai
check_for_nans(data, layer_index)
¶
Checks for NaN values in the tensor data.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
data |
Tensor
|
The tensor to check for NaN values. |
required |
layer_index |
int
|
The index of the layer for logging purposes. |
required |
Returns:
Name | Type | Description |
---|---|---|
bool |
bool
|
True if NaNs are found, False otherwise. |
Source code in spotpython/plot/xai.py
23 24 25 26 27 28 29 30 31 32 33 34 35 36 |
|
get_activations(net, fun_control, batch_size, device='cpu', normalize=False)
¶
Computes the activations for each layer of the network, the mean activations, and the sizes of the activations for each layer.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
Module
|
The neural network model. |
required |
fun_control |
dict
|
A dictionary containing the dataset. |
required |
batch_size |
int
|
The batch size for the data loader. |
required |
device |
str
|
The device to run the model on. Defaults to “cpu”. |
'cpu'
|
normalize |
bool
|
Whether to normalize the input data. Defaults to False. |
False
|
Returns:
Name | Type | Description |
---|---|---|
tuple |
tuple
|
A tuple containing the activations, mean activations, and layer sizes for each layer. |
Examples:
>>> from spotpython.plot.xai import get_activations
import torch
import numpy as np
import torch.nn as nn
from spotpython.utils.init import fun_control_init
from spotpython.data.diabetes import Diabetes
from spotpython.light.regression.nn_linear_regressor import NNLinearRegressor
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.hyperparameters.values import (
get_default_hyperparameters_as_array, get_one_config_from_X)
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.data.lightdatamodule import LightDataModule
from spotpython.plot.xai import get_gradients
fun_control = fun_control_init(
_L_in=10, # 10: diabetes
_L_out=1,
_torchmetric="mean_squared_error",
data_set=Diabetes(),
core_model=NNLinearRegressor,
hyperdict=LightHyperDict)
X = get_default_hyperparameters_as_array(fun_control)
config = get_one_config_from_X(X, fun_control)
_L_in = fun_control["_L_in"]
_L_out = fun_control["_L_out"]
_torchmetric = fun_control["_torchmetric"]
batch_size = 16
model = fun_control["core_model"](**config, _L_in=_L_in, _L_out=_L_out, _torchmetric=_torchmetric)
activations, mean_activations, layer_sizes = get_activations(net=model, fun_control=fun_control, batch_size=batch_size, device = "cpu")
plot_nn_values_scatter(nn_values=activations, layer_sizes=layer_sizes, nn_values_names="Activations")
Source code in spotpython/plot/xai.py
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 |
|
get_all_layers_conductance(spot_tuner, fun_control, device='cpu', remove_spot_attributes=False)
¶
Get the conductance of all layers.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
spot_tuner |
object
|
The spot tuner object. |
required |
fun_control |
dict
|
A dictionary with the function control. |
required |
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
remove_spot_attributes |
bool
|
Whether to remove the spot attributes. Defaults to False. |
False
|
Source code in spotpython/plot/xai.py
959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 |
|
get_attributions(spot_tuner, fun_control, attr_method='IntegratedGradients', baseline=None, abs_attr=True, n_rel=5, device='cpu', normalize=True, remove_spot_attributes=False)
¶
Get the attributions of a neural network.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
spot_tuner |
object
|
The spot tuner object. |
required |
fun_control |
dict
|
A dictionary with the function control. |
required |
attr_method |
str
|
The attribution method. Defaults to “IntegratedGradients”. |
'IntegratedGradients'
|
baseline |
Tensor
|
The baseline for the attribution methods. Defaults to None. |
None
|
abs_attr |
bool
|
Whether the method should sort by the absolute attribution values. Defaults to True. |
True
|
n_rel |
int
|
The number of relevant features. Defaults to 5. |
5
|
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
normalize |
bool
|
Whether to normalize the input data. Defaults to True. |
True
|
remove_spot_attributes |
bool
|
Whether to remove the spot attributes.
If True, a torch model is created via |
False
|
Returns:
Type | Description |
---|---|
DataFrame
|
pd.DataFrame (object): A DataFrame with the attributions. |
Source code in spotpython/plot/xai.py
653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 |
|
get_gradients(net, fun_control, batch_size, device='cpu', normalize=False)
¶
Get the gradients of a neural network and the size of each layer.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
object
|
A neural network. |
required |
fun_control |
dict
|
A dictionary with the function control. |
required |
batch_size |
int
|
The batch size. |
required |
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
normalize |
bool
|
Whether to normalize the input data. Defaults to False. |
False
|
Returns:
Name | Type | Description |
---|---|---|
tuple |
tuple
|
A tuple containing: - grads: A dictionary with the gradients of the neural network. - layer_sizes: A dictionary with layer names as keys and their sizes as entries in NumPy array format. |
Examples:
>>> from spotpython.plot.xai import get_gradients
import torch
import numpy as np
import torch.nn as nn
from spotpython.utils.init import fun_control_init
from spotpython.data.diabetes import Diabetes
from spotpython.light.regression.nn_linear_regressor import NNLinearRegressor
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.hyperparameters.values import (
get_default_hyperparameters_as_array, get_one_config_from_X)
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.data.lightdatamodule import LightDataModule
# from spotpython.plot.xai import get_gradients
fun_control = fun_control_init(
_L_in=10, # 10: diabetes
_L_out=1,
_torchmetric="mean_squared_error",
data_set=Diabetes(),
core_model=NNLinearRegressor,
hyperdict=LightHyperDict)
X = get_default_hyperparameters_as_array(fun_control)
config = get_one_config_from_X(X, fun_control)
_L_in = fun_control["_L_in"]
_L_out = fun_control["_L_out"]
_torchmetric = fun_control["_torchmetric"]
batch_size = 16
model = fun_control["core_model"](**config, _L_in=_L_in, _L_out=_L_out, _torchmetric=_torchmetric)
gradients, layer_sizes = get_gradients(net=model)
gradients, layer_sizes
Source code in spotpython/plot/xai.py
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 |
|
get_layer_conductance(spot_tuner, fun_control, layer_idx, device='cpu', normalize=True, remove_spot_attributes=False)
¶
Compute the average layer conductance attributions for a specified layer in the model.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
spot_tuner |
Spot
|
The spot tuner object containing the trained model. |
required |
fun_control |
dict
|
The fun_control dictionary containing the hyperparameters used to train the model. |
required |
layer_idx |
int
|
Index of the layer for which to compute layer conductance attributions. |
required |
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
normalize |
bool
|
Whether to normalize the input data. Defaults to True. |
True
|
remove_spot_attributes |
bool
|
Whether to remove the spot attributes. Defaults to False. |
False
|
Returns:
Type | Description |
---|---|
ndarray
|
numpy.ndarray: An array containing the average layer conductance attributions for the specified layer. The shape of the array corresponds to the shape of the attributions. |
Source code in spotpython/plot/xai.py
820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 |
|
get_weights(net, return_index=False)
¶
Get the weights of a neural network and the size of each layer.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
object
|
A neural network. |
required |
return_index |
bool
|
Whether to return the index. Defaults to False. |
False
|
Returns:
Name | Type | Description |
---|---|---|
tuple |
tuple
|
A tuple containing: - weights: A dictionary with the weights of the neural network. - index: The layer index list (only if return_index is True). - layer_sizes: A dictionary with layer names as keys and their sizes as entries in NumPy array format. |
Examples:
>>> from spotpython.plot.xai import get_weights
import torch
import numpy as np
import torch.nn as nn
from spotpython.utils.init import fun_control_init
from spotpython.data.diabetes import Diabetes
from spotpython.light.regression.nn_linear_regressor import NNLinearRegressor
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.hyperparameters.values import (
get_default_hyperparameters_as_array, get_one_config_from_X)
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.data.lightdatamodule import LightDataModule
from spotpython.plot.xai import get_gradients
fun_control = fun_control_init(
_L_in=10, # 10: diabetes
_L_out=1,
_torchmetric="mean_squared_error",
data_set=Diabetes(),
core_model=NNLinearRegressor,
hyperdict=LightHyperDict)
X = get_default_hyperparameters_as_array(fun_control)
config = get_one_config_from_X(X, fun_control)
_L_in = fun_control["_L_in"]
_L_out = fun_control["_L_out"]
_torchmetric = fun_control["_torchmetric"]
batch_size = 16
model = fun_control["core_model"](**config, _L_in=_L_in, _L_out=_L_out, _torchmetric=_torchmetric)
weights, layer_sizes = get_weights(net=model)
weights, layer_sizes
Source code in spotpython/plot/xai.py
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 |
|
get_weights_conductance_last_layer(spot_tuner, fun_control, device='cpu', remove_spot_attributes=False)
¶
Get the weights and the conductance of the last layer.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
spot_tuner |
object
|
The spot tuner object. |
required |
fun_control |
dict
|
A dictionary with the function control. |
required |
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
remove_spot_attributes |
bool
|
Whether to remove the spot attributes. Defaults to False. |
False
|
Source code in spotpython/plot/xai.py
888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 |
|
is_square(n)
¶
Check if a number is a square number.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
n |
int
|
The number to check. |
required |
Returns:
Name | Type | Description |
---|---|---|
bool |
bool
|
True if the number is a square number, False otherwise. |
Examples:
>>> is_square(4)
True
>>> is_square(5)
False
Source code in spotpython/plot/xai.py
802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 |
|
plot_attributions(df, attr_method='IntegratedGradients')
¶
Plot the attributions of a neural network.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
df |
DataFrame
|
A DataFrame with the attributions. |
required |
attr_method |
str
|
The attribution method. Defaults to “IntegratedGradients”. |
'IntegratedGradients'
|
Returns:
Type | Description |
---|---|
None
|
None |
Source code in spotpython/plot/xai.py
779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 |
|
plot_conductance_last_layer(weights_last, layer_conductance_last, figsize=(12, 6), show=True)
¶
Plot the conductance of the last layer.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
weights_last |
ndarray
|
The weights of the last layer. |
required |
layer_conductance_last |
ndarray
|
The conductance of the last layer. |
required |
figsize |
tuple
|
The figure size. Defaults to (12, 6). |
(12, 6)
|
show |
bool
|
Whether to show the plot. Defaults |
True
|
Examples:
>>> import numpy as np
from spotpython.plot.xai import plot_conductance_last_layer
weights_last = np.random.rand(10)
layer_conductance_last = np.random.rand(10)
plot_conductance_last_layer(weights_last, layer_conductance_last, show=True)
Source code in spotpython/plot/xai.py
921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 |
|
plot_nn_values_hist(nn_values, net, nn_values_names='', color='C0', columns=2)
¶
Plot the values of a neural network. Can be used to plot the weights, gradients, or activations of a neural network.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
nn_values |
dict
|
A dictionary with the values of the neural network. For example, the weights, gradients, or activations. |
required |
net |
object
|
A neural network. |
required |
color |
str
|
The color to use. Defaults to “C0”. |
'C0'
|
columns |
int
|
The number of columns. Defaults to 2. |
2
|
Source code in spotpython/plot/xai.py
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 |
|
plot_nn_values_scatter(nn_values, layer_sizes, nn_values_names='', absolute=True, cmap='gray', figsize=(6, 6), return_reshaped=False, show=True, colorbar_orientation='auto')
¶
Plot the values of a neural network including a marker for padding values.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
nn_values |
dict
|
A dictionary with the values of the neural network. For example, the weights, gradients, or activations. |
required |
layer_sizes |
dict
|
A dictionary with layer names as keys and their sizes as entries in NumPy array format. |
required |
nn_values_names |
str
|
The name of the values. Defaults to “”. |
''
|
absolute |
bool
|
Whether to use the absolute values. Defaults to True. |
True
|
cmap |
str
|
The colormap to use. Defaults to “gray”. |
'gray'
|
figsize |
tuple
|
The figure size. Defaults to (6, 6). |
(6, 6)
|
return_reshaped |
bool
|
Whether to return the reshaped values. Defaults to False. |
False
|
show |
bool
|
Whether to show the plot. Defaults to True. |
True
|
colorbar_orientation |
str
|
The orientation of the colorbar. Can be “auto”, “horizontal”, “vertical”, or “none”. “auto” will choose the orientation based on the geometry of the plot. “none” will not show the colorbar. Defaults to “auto”. |
'auto'
|
Returns:
Name | Type | Description |
---|---|---|
dict |
dict
|
A dictionary with the reshaped values. |
Source code in spotpython/plot/xai.py
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 |
|
sort_layers(data_dict)
¶
Sorts a dictionary with keys in the format “Layer X” based on the numerical value X.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
data_dict |
dict
|
A dictionary with keys in the format “Layer X”. |
required |
Returns:
Name | Type | Description |
---|---|---|
dict |
dict
|
A dictionary with the keys sorted based on the numerical value X. |
Examples:
>>> data_dict = {
... "Layer 1": [1, 2, 3],
... "Layer 3": [4, 5, 6],
... "Layer 2": [7, 8, 9]
... }
>>> sort_layers(data_dict)
{'Layer 1': [1, 2, 3], 'Layer 2': [7, 8, 9], 'Layer 3': [4,
Source code in spotpython/plot/xai.py
989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 |
|
visualize_activations_distributions(activations, net, color='C0', columns=4, bins=50, show=True)
¶
Plots the distribution of activations for each layer that were determined via the get_activations function.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
activations |
dict
|
A dictionary containing activations for each layer. |
required |
net |
Module
|
The neural network model. |
required |
color |
str
|
The color for the plot histogram. Defaults to “C0”. |
'C0'
|
columns |
int
|
The number of columns for the subplots. Defaults to 4. |
4
|
bins |
int
|
The number of bins for the histogram. Defaults to 50. |
50
|
show |
bool
|
Whether to show the plot. Defaults to True. |
True
|
Returns:
Type | Description |
---|---|
None
|
None |
Source code in spotpython/plot/xai.py
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 |
|
visualize_gradient_distributions(net, fun_control, batch_size, device='cpu', color='C0', xlabel=None, stat='count', use_kde=True, columns=2, normalize=True)
¶
Plot the gradients distributions of a neural network.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
object
|
A neural network. |
required |
fun_control |
dict
|
A dictionary with the function control. |
required |
batch_size |
int
|
The batch size. |
required |
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
color |
str
|
The color to use. Defaults to “C0”. |
'C0'
|
xlabel |
str
|
The x label. Defaults to None. |
None
|
stat |
str
|
The stat. Defaults to “count”. |
'count'
|
use_kde |
bool
|
Whether to use kde. Defaults to True. |
True
|
columns |
int
|
The number of columns. Defaults to 2. |
2
|
normalize |
bool
|
Whether to normalize the input data. Defaults to True. |
True
|
Returns:
Type | Description |
---|---|
None
|
None |
Source code in spotpython/plot/xai.py
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 |
|
visualize_gradients(net, fun_control, batch_size, absolute=True, cmap='gray', figsize=(6, 6), device='cpu', normalize=True)
¶
Scatter plots the gradients of a neural network.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
object
|
A neural network. |
required |
fun_control |
dict
|
A dictionary with the function control. |
required |
batch_size |
int
|
The batch size. |
required |
absolute |
bool
|
Whether to use the absolute values. Defaults to True. |
True
|
cmap |
str
|
The colormap to use. Defaults to “gray”. |
'gray'
|
figsize |
tuple
|
The figure size. Defaults to (6, 6). |
(6, 6)
|
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
normalize |
bool
|
Whether to normalize the input data. Defaults to True. |
True
|
Returns:
Type | Description |
---|---|
None
|
None |
Source code in spotpython/plot/xai.py
611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 |
|
visualize_mean_activations(mean_activations, layer_sizes, absolute=True, cmap='gray', figsize=(6, 6))
¶
Scatter plots the mean activations of a neural network for each layer. means_activations is a dictionary with the mean activations of the neural network computed via the get_activations function.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
mean_activations |
dict
|
A dictionary with the mean activations of the neural network. |
required |
layer_sizes |
dict
|
A dictionary with layer names as keys and their sizes as entries in NumPy array format. |
required |
absolute |
bool
|
Whether to use the absolute values. Defaults to True. |
True
|
cmap |
str
|
The colormap to use. Defaults to “gray”. |
'gray'
|
figsize |
tuple
|
The figure size. Defaults to (6, 6). |
(6, 6)
|
Returns:
Type | Description |
---|---|
None
|
None |
Examples:
>>> from spotpython.plot.xai import get_activations
activations, mean_activations, layer_sizes = get_activations(net, fun_control)
visualize_mean_activations(mean_activations, layer_sizes)
Source code in spotpython/plot/xai.py
522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 |
|
visualize_weights(net, absolute=True, cmap='gray', figsize=(6, 6))
¶
Scatter plots the weights of a neural network.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
object
|
A neural network. |
required |
absolute |
bool
|
Whether to use the absolute values. Defaults to True. |
True
|
cmap |
str
|
The colormap to use. Defaults to “gray”. |
'gray'
|
figsize |
tuple
|
The figure size. Defaults to (6, 6). |
(6, 6)
|
Returns:
Type | Description |
---|---|
None
|
None |
Examples:
>>> from spotpython.utils.init import fun_control_init
from spotpython.data.diabetes import Diabetes
from spotpython.light.regression.nn_linear_regressor import NNLinearRegressor
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.hyperparameters.values import (
get_default_hyperparameters_as_array, get_one_config_from_X)
from spotpython.plot.xai import visualize_weights
fun_control = fun_control_init(
_L_in=10, # 10: diabetes
_L_out=1,
_torchmetric="mean_squared_error",
data_set=Diabetes(),
core_model=NNLinearRegressor,
hyperdict=LightHyperDict)
X = get_default_hyperparameters_as_array(fun_control)
config = get_one_config_from_X(X, fun_control)
_L_in = fun_control["_L_in"]
_L_out = fun_control["_L_out"]
_torchmetric = fun_control["_torchmetric"]
batch_size = 16
model = fun_control["core_model"](**config, _L_in=_L_in, _L_out=_L_out, _torchmetric=_torchmetric)
visualize_weights(net=model, absolute=True, cmap="gray", figsize=(6, 6))
Source code in spotpython/plot/xai.py
559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 |
|
visualize_weights_distributions(net, color='C0', columns=2)
¶
Plot the weights distributions of a neural network.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
object
|
A neural network. |
required |
color |
str
|
The color to use. Defaults to “C0”. |
'C0'
|
columns |
int
|
The number of columns. Defaults to 2. |
2
|
Returns:
Type | Description |
---|---|
None
|
None |
Source code in spotpython/plot/xai.py
457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 |
|
viz_net(net, device='cpu', show_attrs=False, show_saved=False, max_attr_chars=50, filename='model_architecture', format='png')
¶
Visualize the architecture of a linear neural network.
Produces Graphviz representation of PyTorch autograd graph.
If a node represents a backward function, it is gray. Otherwise, the node represents a tensor and is either blue, orange, or green:
- Blue: reachable leaf tensors that requires grad (tensors whose .grad fields will be populated during .backward())
- Orange: saved tensors of custom autograd functions as well as those saved by built-in backward nodes
- Green: tensor passed in as outputs
- Dark green: if any output is a view, we represent its base tensor with a dark green node.
If show_attrs
=True and show_saved
=True it is shown what autograd saves for the backward pass.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
net |
Module
|
The neural network model. |
required |
device |
str
|
The device to use. Defaults to “cpu”. |
'cpu'
|
show_attrs |
bool
|
whether to display non-tensor attributes of backward nodes (Requires PyTorch version >= 1.9) |
False
|
show_saved |
bool
|
whether to display saved tensor nodes that are not by custom autograd functions. Saved tensor nodes for custom functions, if present, are always displayed. (Requires PyTorch version >= 1.9) |
False
|
max_attr_chars |
int
|
if show_attrs is True, sets max number of characters to display for any given attribute. Defaults to 50. |
50
|
filename |
str
|
The filename. Defaults to “model_architecture”. |
'model_architecture'
|
format |
str
|
The output format. Defaults to “png”. |
'png'
|
Returns:
Type | Description |
---|---|
None
|
None |
Raises:
Type | Description |
---|---|
ValueError
|
If the model does not have a linear layer. |
TypeError
|
If the network structure or parameters are invalid. |
RuntimeError
|
If an unexpected error occurs. |
Examples:
>>> from spotpython.plot.xai import viz_net
from spotpython.utils.init import fun_control_init
from spotpython.data.diabetes import Diabetes
from spotpython.light.regression.nn_linear_regressor import NNLinearRegressor
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
from spotpython.hyperparameters.values import (
get_default_hyperparameters_as_array, get_one_config_from_X)
from spotpython.hyperdict.light_hyper_dict import LightHyperDict
_L_in=10
_L_out=1
_torchmetric="mean_squared_error"
fun_control = fun_control_init(
_L_in=_L_in,
_L_out=_L_out,
_torchmetric=_torchmetric,
data_set=Diabetes(),
core_model=NNLinearRegressor,
hyperdict=LightHyperDict)
X = get_default_hyperparameters_as_array(fun_control)
config = get_one_config_from_X(X, fun_control)
model = fun_control["core_model"](**config, _L_in=_L_in, _L_out=_L_out, _torchmetric=_torchmetric)
viz_net(net=model, device="cpu", show_attrs=True, show_saved=True, filename="model_architecture3", format="png")
Source code in spotpython/plot/xai.py
1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 |
|