API documentation
AutoeXplainer
AutoExplainer
The main class that evaluates a series of explanation methods and chooses the best one.
Attributes:
Name | Type | Description |
---|---|---|
raw_results |
Dict
|
Raw values of metrics computed for each observation for each explanation. |
first_aggregation_results |
Dict
|
Values of metrics aggregated across observations, i.e. each explanation function has value for each metric. |
second_aggregation_results |
Dict
|
Values of metrics aggregated for each explanation method. Each explanation method has single value, that represents overall quality. |
best_explanation_name |
str
|
Name of the selected best explanation found. |
Source code in autoexplainer/autoexplainer.py
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 |
|
__init__(model: torch.nn.Module, data: torch.Tensor, targets: torch.Tensor, device: str = 'cpu', seed: int = 42)
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
torch.nn.Module
|
Convolutional neural network to be explained. On this model some explanation and metric parameters will be inferred. |
required |
data |
torch.Tensor
|
Data that will be used for explanation method evaluation. shape: (N, C, H, W). |
required |
targets |
torch.Tensor
|
Labels for provided data. Encoded as integer vector with shape (N,). |
required |
Source code in autoexplainer/autoexplainer.py
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 |
|
aggregate(first_stage_aggregation_function_name: str = 'mean', second_stage_aggregation_function_name: str = 'rank_based', second_stage_aggregation_function_aggregation_parameters: Dict = None) -> None
Aggregates raw result computed in .evaluate() method in two steps. First, aggregates metric scores across provided observations, i.e. each explanation method has a value for each metric. Secondly, aggregates scores across available metrics, i.e. each explanation method has a single value that represents overall quality.
Stores both aggregation steps in the attributes first_aggregation_results
and second_aggregation_results
.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
first_stage_aggregation_function_name |
{"mean", "median", "min","max"}
|
Name of the function for the first stage aggregation. |
'mean'
|
second_stage_aggregation_function_name |
{"mean", "median", "min","max"}
|
Name of the function for second stage aggregaton. |
'rank_based'
|
second_stage_aggregation_function_aggregation_parameters |
Dict
|
Parameters for the second stage aggregation function. |
None
|
Source code in autoexplainer/autoexplainer.py
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 |
|
evaluate(explanations: List[str] = None, metrics: List[str] = None, explanation_params: Dict = None, metrics_params: Dict = None) -> None
Evaluates explanation methods. Stores results in .raw_results
attribute.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
explanations |
List[str]
|
List of names of explanation methods to be evaluated.
By default, uses all available explanation methods.
Accepts lists with subset of: |
None
|
metrics |
List[str]
|
List of names of evaluation metrics to be used. By default, uses all available metrics.
Accepts lists with subset of: |
None
|
explanation_params |
Dict[str, Dict]
|
Allows to override default parameters of selected explanation functions.
Accept Dictionary with form |
None
|
metrics_params |
Dict[str, Dict]
|
Allows to override default parameters of selected metrics.
Accept Dictionary with form |
None
|
Source code in autoexplainer/autoexplainer.py
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 |
|
get_best_explanation() -> BestExplanation
Returns an object with the selected best explanation method wrapped with a few additions, see BestExplanation for more details. Returns (BestExplanation): BestExplanation object
Source code in autoexplainer/autoexplainer.py
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 |
|
to_html(file_path: str, model_name: str = None, dataset_name: str = None, labels: Dict[int, str] = None) -> None
Generates evaluation report as HTML file.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
file_path |
str
|
Target file path. |
required |
model_name |
str
|
Name of model to show inside report. |
None
|
dataset_name |
str
|
Name of dataset to show inside report. |
None
|
labels |
Dict[int, str]
|
Mapping between class number and class names. e.g. |
None
|
Source code in autoexplainer/autoexplainer.py
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 |
|
to_pdf(folder_path: str = '', model_name: str = 'name of the model', dataset_name: str = 'name of the dataset', labels: Dict[int, str] = None) -> None
Creates PDF report from dict stored in the attribute first_aggregation_results
.
Needs Latex packages installed to run - see README.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
folder_path |
str
|
Path to directory, where the reports (PDF and tex) should be created. |
''
|
model_name |
str
|
Name of model to show inside report. |
'name of the model'
|
dataset_name |
str
|
Name of dataset to show inside report. |
'name of the dataset'
|
labels |
Dict[int, str]
|
Mapping between class number and class names. e.g. |
None
|
Source code in autoexplainer/autoexplainer.py
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 |
|
Metric handlers
AvgSensitivityHandler
Bases: MetricHandler
Metric Handler for Average Sensivity metric (Yeh et al., 2019).
Measures the average sensitivity of an explanation using a Monte Carlo sampling-based approximation.
Dictionary with parameters to override must be in the form:
metric_parameters = {"init": <dictionary with parameters used in metric's __init__>,
"call": <dictionary with parameters used in metric's __call__>}
Parameters accepted in metric_parameters
:
"init":
abs
: a bool stating if absolute operation should be taken on the attributionsnormalise
: a bool stating if the attributions should be normalisednormalise_func
: a Callable that make a normalising transformation of the attributionslower_bound
(float): lower Bound of Perturbation, default=0.2upper_bound
(None, float): upper Bound of Perturbation, default=Nonenr_samples
(integer): the number of samples iterated, default=200.norm_numerator
(callable): function for norm calculations on the numerator, default=fro_norm.norm_denominator
(callable): function for norm calculations on the denominator, default=fro_norm.perturb_func
(callable): input perturbation function, default=uniform_noise.similarity_func
(callable): similarity function applied to compare input and perturbed input.
"call": No parameters are used.
Source code in autoexplainer/metrics.py
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 |
|
FaithfulnessEstimateHandler
Bases: MetricHandler
Metric handler for Faithfulness Estimate metric (Alvarez-Melis et al., 2018).
Computes the correlation between probability drops and attribution scores on various points.
Dictionary with parameters to override must be in the form:
metric_parameters = {"init": <dictionary with parameters used in metric's __init__>,
"call": <dictionary with parameters used in metric's __call__>}
Parameters accepted in metric_parameters
:
"init":
abs
: a bool stating if absolute operation should be taken on the attributionsnormalise
: a bool stating if the attributions should be normalisednormalise_func
: a Callable that make a normalising transformation of the attributionsnr_runs
(integer): the number of runs (for each input and explanation pair), default=100.subset_size
(integer): the size of subset, default=224.perturb_baseline
(string): indicates the type of baseline: "mean", "random", "uniform", "black" or "white", default="mean".similarity_func
(callable): Similarity function applied to compare input and perturbed input, default=correlation_spearman.perturb_func
(callable): input perturbation function, default=baseline_replacement_by_indices.features_in_step
(integer): the size of the step, default=256.softmax
(boolean): indicates wheter to use softmax probabilities or logits in model prediction, default=True.
"call": No parameters are used.
Source code in autoexplainer/metrics.py
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 |
|
IROFHandler
Bases: MetricHandler
Metric handler for Iterative Removal Of Features metric (Rieger at el., 2020).
Computes the area over the curve per class for sorted mean importances of feature segments (superpixels) as they are iteratively removed (and prediction scores are collected), averaged over several test samples.
Dictionary with parameters to override must be in the form:
metric_parameters = {"init": <dictionary with parameters used in metric's __init__>,
"call": <dictionary with parameters used in metric's __call__>}
Parameters accepted in metric_parameters
:
"init":
- abs: a bool stating if absolute operation should be taken on the attributions
- normalise: a bool stating if the attributions should be normalised
- normalise_func: a Callable that make a normalising transformation of the attributions
- segmentation_method (string): Image segmentation method: 'slic' or 'felzenszwalb', default="slic"
- perturb_baseline (string): indicates the type of baseline: "mean", "random", "uniform", "black" or "white", default="mean"
- perturb_func (callable): input perturbation function, default=baseline_replacement_by_indices
- softmax (boolean): indicates wheter to use softmax probabilities or logits in model prediction
"call": No parameters are used.
Source code in autoexplainer/metrics.py
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 |
|
MetricHandler
Bases: ABC
Abstract class for metrics handlers.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
torch.nn.Module
|
Model used for metrics' parameters inference. |
required |
data |
torch.Tensor
|
Data used for metrics' parameters inference. |
required |
targets |
torch.Tensor
|
Target used for metrics' parameters inference. |
required |
metric_parameters |
Dict
|
Metric parameters to overwrite inferred parameters. Dictionary must be in the form:
|
None
|
Attributes:
Name | Type | Description |
---|---|---|
metric |
quantus.Metric
|
Attribute that stores created metric object after determinig its parameters. |
metric_parameters |
Dict
|
Dictionary with parameters used for this metric. |
Source code in autoexplainer/metrics.py
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 |
|
compute_metric_values(model: torch.nn.Module, data: torch.Tensor, targets: torch.Tensor, attributions: torch.Tensor = None, explanation_func: Callable = None) -> np.ndarray
Computes metric values for given model, dataset, and explanation function.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
torch.nn.Module
|
required | |
data |
torch.Tensor
|
required | |
targets |
torch.Tensor
|
required | |
attributions |
torch.Tensor
|
None
|
|
explanation_func |
Callable
|
None
|
Returns:
Type | Description |
---|---|
np.ndarray
|
NumPy array with metric values for each given image. |
Source code in autoexplainer/metrics.py
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 |
|
SparsenessHandler
Bases: MetricHandler
Metric Handler for Sparseness metric (Chalasani et al., 2020).
Uses the Gini Index for measuring, if only highly attributed features are truly predictive of the model output.
Dictionary with parameters to override must be in the form:
metric_parameters = {"init": <dictionary with parameters used in metric's __init__>,
"call": <dictionary with parameters used in metric's __call__>}
Parameters accepted in metric_parameters
:
"init":
- abs: a bool stating if absolute operation should be taken on the attributions
- normalise: a bool stating if the attributions should be normalised
- normalise_func: a Callable that make a normalising transformation of the attributions
"call": No parameters are used.
Source code in autoexplainer/metrics.py
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 |
|
Explanation methods handlers
BestExplanation
Class for an object that wraps the best explanation method selected during the evaluation process.
Attributes:
Name | Type | Description |
---|---|---|
attributions |
torch.Tensor
|
Attributions computed during evaluation using this explanation method only. |
explanation_function |
torch.Tensor
|
Function that computes attributions for the provided model and data. |
name |
str
|
Name of this explanation method. |
parameters |
Dict
|
Parameters used in this explanation method. |
metric_handlers |
Dict
|
Dictionary with metric handlers that this explanation method was evaluated with. |
aggregation_parameters |
Dict
|
Parameter that were used during aggregation of metric values. |
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|
evaluate(model: torch.nn.Module, data: torch.Tensor, targets: torch.Tensor, attributions: torch.Tensor = None, aggregate: bool = False) -> Dict
Evaluate the selected best explanation method again on new data.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
torch.nn.Module
|
Convolutional neural network to be explained. |
required |
data |
torch.Tensor
|
Data that will be used for the evaluation of the explanation method. shape: (N, C, H, W). |
required |
targets |
torch.Tensor
|
Labels for provided data. Encoded as integer vector with shape (N,). |
required |
attributions |
torch.Tensor
|
Attributions for this data that were previously computed, to skip computing them once more. |
None
|
aggregate |
bool
|
Indicates whether results should be aggregated (in the same manner as in AutoExplainer). |
False
|
Returns:
Name | Type | Description |
---|---|---|
results |
Dict
|
Results of evaluation. |
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|
explain(model: torch.nn.Module, data: torch.Tensor, targets: torch.Tensor) -> torch.Tensor
Compute new attributions.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
torch.nn.Module
|
CNN neural network to be explained. |
required |
data |
torch.Tensor
|
Data for which attributions will be computed. shape: (N, C, H, W). |
required |
targets |
torch.Tensor
|
Labels for provided data. Encoded as integer vector with shape (N,). |
required |
Returns:
Type | Description |
---|---|
torch.Tensor
|
attributions (torch.Tensor) |
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|
ExplanationHandler
Bases: ABC
Abstract class for explanation methods handlers. Handlers manage explanation methods: they read and adapt parameters for given model and data. They also create explanation functions that may be used by the user or can be passed to metric handlers.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
model |
torch.nn.Module
|
Model used for methods' parameter adaptation. |
required |
data |
torch.Tensor
|
Data used for method's parameters adaptation. Tensor with shape (N, C, H, W) |
required |
targets |
torch.Tensor
|
Target used for method's parameters inference - integer vector with shape (N,) |
required |
explanation_parameters |
Dict
|
Explanation method parameters to be overwritten. |
None
|
Attributes:
Name | Type | Description |
---|---|---|
explanation_function |
Callable
|
Explanation method as function ready to be used with already set parameters. |
explanation_parameters |
Dict
|
Parameters chosen for given explanation method. |
attributions |
torch.Tensor
|
Computed attributions, only most recent. |
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|
get_explanation_function() -> Callable
Return function that can be run by Quantus metrics.
Source code in autoexplainer/explanations/explanation_handlers.py
66 67 68 |
|
GradCamHandler
Bases: ExplanationHandler
Handler for GradCam explanation method. Uses captum implementation of GradCam.
By default, the last convolutional layer is chosen as a parameter for GradCam.
To overwrite default parameters, passed dictionary must be in the form:
explanation_parameters = {"explanation_parameters":{ <parameters accepted by GradCam in Captum> }}
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|
IntegratedGradients
Bases: ExplanationHandler
Handler for Integrated Gradients explanation method. Uses implementation in Quantus library.
To overwrite default parameters, passed dictionary must be in the form:
explanation_parameters = {"explanation_parameters":{ <parameters> }}
Integrated Gradients method accepts paramteters:
normalise
(Bool) - Normalize attribution values. default=Falseabs
(bool) - Return absolute values of attribtuion. default=False- 'pos_only` (bool) - Clip negative values of attribution to 0.0. default=False
neg_only
(bool) - Clip positive values of attribution to 0.0. default=False
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|
KernelShapHandler
Bases: ExplanationHandler
Handler for Kernel Shap explanation. Uses captum implementation of Kernel Shap. Accepts parameters with form:
To overwrite default parameters, passed dictionary must be in the form:
explanation_parameters = {"mask_parameters": { "mask_function_name":<str>, <other parameters for chosen mask function> },
"explanation_parameters":{ "baseline_function_name":<str>, <parameters accepted by KernelShap in Captum> }}
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|
SaliencyHandler
Bases: ExplanationHandler
Handler for Saliency explanation method. Uses implementation in Quantus library.
To overwrite default parameters, passed dictionary must be in the form:
explanation_parameters = {"explanation_parameters":{ <parameters> }}
Saliency method accepts paramteters:
normalise
(Bool) - Normalize attribution values. default=Falseabs
(bool) - Return absolute values of attribtuion. default=False- 'pos_only` (bool) - Clip negative values of attribution to 0.0. default=False
neg_only
(bool) - Clip positive values of attribution to 0.0. default=False
Source code in autoexplainer/explanations/explanation_handlers.py
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 |
|