itslukeypookie commited on
Commit
38f18d9
·
verified ·
1 Parent(s): a888d12

Delete main.py

Browse files
Files changed (1) hide show
  1. main.py +0 -560
main.py DELETED
@@ -1,560 +0,0 @@
1
- ###### Train CIFAR10 with PyTorch. ######
2
-
3
- ### IMPORT DEPENDENCIES
4
-
5
- from torch.utils.data import DataLoader
6
- import torch
7
- import torch.nn as nn
8
- import torch.optim as optim
9
- import torch.nn.functional as F
10
- import torch.backends.cudnn as cudnn
11
- import gradio as gr
12
- import wandb
13
- import math
14
- import numpy as np
15
- import matplotlib.pyplot as plt
16
-
17
-
18
- import torchvision
19
- import torchvision.transforms as transforms
20
- import torchvision.models as models
21
- import torch.optim.lr_scheduler as lr_scheduler
22
- import os
23
- import argparse
24
- import torchattacks
25
-
26
- from models import *
27
-
28
- from tqdm import tqdm
29
- from PIL import Image
30
- import gradio as gr
31
-
32
- # from utils import progress_bar
33
-
34
- # CSS theme styling
35
- theme = gr.themes.Base(
36
- font=[gr.themes.GoogleFont('Montserrat'), 'ui-sans-serif', 'system-ui', 'sans-serif'],
37
- primary_hue="emerald",
38
- secondary_hue="emerald",
39
- neutral_hue="zinc"
40
- ).set(
41
- body_text_color='*neutral_950',
42
- body_text_color_subdued='*neutral_950',
43
- block_shadow='*shadow_drop_lg',
44
- button_shadow='*shadow_drop_lg',
45
- block_title_text_color='*neutral_950',
46
- block_title_text_weight='500',
47
- slider_color='*secondary_600'
48
- )
49
-
50
- def normalize(img):
51
- min_im = np.min(img)
52
- np_img = img - min_im
53
- max_im = np.max(np_img)
54
- np_img /= max_im
55
- return np_img
56
-
57
- def imshow(img, fig_name = "test_input.png"):
58
- try:
59
- img = img.clone().detach().cpu().numpy()
60
- except:
61
- print('img already numpy')
62
-
63
- plt.imshow(normalize(np.transpose(img, (1, 2, 0))))
64
- plt.savefig(fig_name)
65
- print(f'Figure saved as {fig_name}')
66
- return fig_name
67
-
68
- def class_names(class_num, class_list): # converts the raw number label to text
69
- if (class_num < 0) and (class_num >= 10):
70
- gr.Warning("Class List Error")
71
- return
72
- return class_list[class_num]
73
-
74
-
75
- ### MAIN FUNCTION
76
- best_acc = 0
77
- def main(drop_type, epochs_sldr, train_sldr, test_sldr, learning_rate, optimizer, sigma_sldr, adv_attack, username, scheduler):
78
-
79
- ## Input protection
80
- if not drop_type:
81
- gr.Warning("Please select a model from the dropdown.")
82
- return
83
- if not username:
84
- gr.Warning("Please enter a WandB username.")
85
- return
86
- if(epochs_sldr % 1 != 0):
87
- gr.Warning("Number of epochs must be an integer.")
88
- return
89
- if(train_sldr % 1 != 0):
90
- gr.Warning("Training batch size must be an integer.")
91
- return
92
- if(test_sldr % 1 != 0):
93
- gr.Warning("Testing batch size must be an integer.")
94
- return
95
-
96
- num_epochs = int(epochs_sldr)
97
- global learn_batch
98
- learn_batch = int(train_sldr)
99
- global test_batch
100
- test_batch = int(test_sldr)
101
- learning_rate = float(learning_rate)
102
- optimizer_choose = str(optimizer)
103
- sigma = float(sigma_sldr)
104
- attack = str(adv_attack)
105
- scheduler_choose = str(scheduler)
106
-
107
- # REPLACE ENTITY WITH USERNAME BELOW
108
- wandb.init(entity=username, project="model-training")
109
-
110
- parser = argparse.ArgumentParser(description='PyTorch CIFAR10 Training')
111
- parser.add_argument('--lr', default=0.1, type=float, help='learning rate')
112
- parser.add_argument('--resume', '-r', action='store_true',
113
- help='resume from checkpoint')
114
- args = parser.parse_args()
115
-
116
- if torch.cuda.is_available():
117
- device = 'cuda'
118
- gr.Info("Cuda detected - running on Cuda")
119
- elif torch.backends.mps.is_available():
120
- device = 'mps'
121
- gr.Info("MPS detected - running on Metal")
122
- else:
123
- device = 'cpu'
124
- gr.Info("No GPU Detected - running on CPU")
125
-
126
- start_epoch = 0 # start from epoch 0 or last checkpoint epoch
127
-
128
- ## Data
129
- try:
130
- print('==> Preparing data..')
131
- transform_train = transforms.Compose([
132
- transforms.RandomCrop(32, padding=4),
133
- transforms.RandomHorizontalFlip(),
134
- transforms.ToTensor(),
135
- transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
136
- ])
137
-
138
- transform_test = transforms.Compose([
139
- transforms.ToTensor(),
140
- transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
141
- ])
142
-
143
- trainset = torchvision.datasets.CIFAR10(
144
- root='./data', train=True, download=True, transform=transform_train)
145
- trainloader = DataLoader(
146
- trainset, batch_size=learn_batch, shuffle=True, num_workers=2)
147
-
148
- testset = torchvision.datasets.CIFAR10(
149
- root='./data', train=False, download=True, transform=transform_test)
150
- testloader = DataLoader(
151
- testset, batch_size=test_batch, shuffle=True, num_workers=2)
152
-
153
- classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
154
- except Exception as e:
155
- print(f"Error: {e}")
156
- gr.Warning(f"Data Loading Error: {e}")
157
-
158
- ## Model
159
- try:
160
- print('==> Building model..')
161
- net = models_dict.get(drop_type, None)
162
-
163
- # Make list of models containing either classifer or fc functions
164
- classifier_models = ['ConvNext_Small', 'ConvNext_Base', 'ConvNext_Large', 'DenseNet', 'EfficientNet_B0', 'MobileNetV2',
165
- 'MaxVit', 'MnasNet0_5', 'SqueezeNet', 'VGG19']
166
- fc_models = ['GoogLeNet', 'InceptionNetV3', 'RegNet_X_400MF', 'ResNet18', 'ShuffleNet_V2_X0_5']
167
-
168
- # Check dropdown choice for fc or classifier function implementation
169
- if net in classifier_models:
170
- num_ftrs = net.classifier[-1].in_features
171
- net.classifier[-1] = torch.nn.Linear(num_ftrs, len(classes))
172
- elif net in fc_models:
173
- num_ftrs = net.fc.in_features
174
- net.fc = torch.nn.Linear(num_ftrs, len(classes))
175
-
176
- net = net.to(device)
177
-
178
- except Exception as e:
179
- print(f"Error: {e}")
180
- gr.Warning(f"Model Building Error: {e}")
181
-
182
- # if args.resume:
183
- # # Load checkpoint.
184
- # print('==> Resuming from checkpoint..')
185
- # assert os.path.isdir('checkpoint'), 'Error: no checkpoint directory found!'
186
- # checkpoint = torch.load('./checkpoint/ckpt.pth')
187
- # net.load_state_dict(checkpoint['net'])
188
- # best_acc = checkpoint['acc']
189
- # start_epoch = checkpoint['epoch']
190
-
191
- SGDopt = optim.SGD(net.parameters(), lr=learning_rate,momentum=0.9, weight_decay=5e-4)
192
- Adamopt = optim.Adam(net.parameters(), lr=learning_rate, weight_decay=5e-4)
193
-
194
- criterion = nn.CrossEntropyLoss()
195
-
196
- if optimizer_choose == "SGD":
197
- optimizer = SGDopt
198
- elif optimizer_choose == "Adam":
199
- optimizer = Adamopt
200
- print (f'optimizer: {optimizer}')
201
-
202
- #scheduler = lr_scheduler.LinearLR(optimizer, start_factor=learning_rate, end_factor=0.0001, total_iters=10)
203
- if scheduler_choose == "CosineAnnealingLR":
204
- scheduler = lr_scheduler.CosineAnnealingLR(optimizer, T_max=200)
205
- elif scheduler_choose == "ReduceLROnPlateau":
206
- scheduler = lr_scheduler.ReduceLROnPlateau(optimizer, mode='max', patience=5)
207
- elif scheduler_choose == "StepLR":
208
- scheduler = lr_scheduler.StepLR(optimizer, step_size=30)
209
- print (f'scheduler: {scheduler_choose}')
210
-
211
- img_labels = [] # initialize list for label generation
212
- raw_image_list = [] # initialize list for image generation
213
- img_list1 = [] # initialize list for combined image/labels
214
- img_list2 = [] # initialize list for gaussian image generation
215
- img_list3 = [] # initialize list for adversarial attack image generation
216
-
217
- # The following lists are used when generating all images in an epoch instead of 10:
218
- full_img_labels = []
219
- full_raw_image_list = []
220
- full_img_list1 = []
221
-
222
- adv_num = 1 # initialize adversarial image number for naming purposes
223
- global gaussian_num
224
- gaussian_num = 1 # initialize gaussian noise image number for naming purposes
225
-
226
- for epoch in range(start_epoch, start_epoch+epochs_sldr):
227
- if sigma == 0:
228
- train(epoch, net, trainloader, device, optimizer, criterion, sigma)
229
- else:
230
- gaussian_fig = train(epoch, net, trainloader, device, optimizer, criterion, sigma)
231
- acc, predicted = test(epoch, net, testloader, device, criterion)
232
-
233
- if scheduler_choose == "ReduceLROnPlateau":
234
- scheduler.step(metrics=acc)
235
- elif not scheduler_choose == "None":
236
- scheduler.step()
237
-
238
- if (((epoch-1) % 10 == 0) or (epoch == 0)) and (epoch != 1): # generate images every 10 epochs (and the 0th epoch)
239
- dataiter = iter(testloader)
240
- imgs, labels = next(dataiter)
241
- normalized_imgs = (imgs-imgs.min())/(imgs.max()-imgs.min())
242
- atk = torchattacks.PGD(net, eps=0.00015, alpha=0.0000000000000001, steps=7)
243
- if attack == "Yes":
244
- if normalized_imgs is None:
245
- print("error occured")
246
- else:
247
- print(torch.std(normalized_imgs))
248
- atk.set_normalization_used(mean = torch.mean(normalized_imgs,axis=[0,2,3]), std=torch.std(normalized_imgs,axis=[0,2,3])/1.125)
249
- adv_images = atk(imgs, labels)
250
- fig_name = imshow(adv_images[0], fig_name = f'figures/adversarial_attack{adv_num}.png')
251
- attack_fig = Image.open(fig_name)
252
- for i in range(1): # generate 1 image per epoch
253
- img_list3.append(attack_fig)
254
- adv_num = adv_num + 1
255
- for i in range(10): # generate 10 images per epoch
256
- gradio_imgs = transforms.functional.to_pil_image(normalized_imgs[i])
257
- raw_image_list.append(gradio_imgs)
258
- predicted_text = class_names(predicted[i].item(), classes)
259
- actual_text = class_names(labels[i].item(), classes)
260
- label_text = f'Epoch: {epoch} | Predicted: {predicted_text} | Actual: {actual_text}'
261
- img_labels.append(label_text)
262
- for i in range(test_batch): # generate all images per epoch
263
- full_gradio_imgs = transforms.functional.to_pil_image(normalized_imgs[i])
264
- full_raw_image_list.append(full_gradio_imgs)
265
- full_predicted_text = class_names(predicted[i].item(), classes)
266
- full_actual_text = class_names(labels[i].item(), classes)
267
- full_label_text = f'Epoch: {epoch} | Predicted: {full_predicted_text} | Actual: {full_actual_text}'
268
- full_img_labels.append(full_label_text)
269
- for i in range(len(raw_image_list)):
270
- img_tuple = (raw_image_list[i], img_labels[i])
271
- img_list1.append(img_tuple)
272
- for i in range(len(full_raw_image_list)):
273
- full_img_tuple = (full_raw_image_list[i], full_img_labels[i])
274
- full_img_list1.append(full_img_tuple)
275
- if sigma != 0:
276
- for i in range(1): # generate 1 image per epoch
277
- img_list2.append(gaussian_fig)
278
- gaussian_num = gaussian_num + 1
279
- if (sigma == 0) and (attack == "No"):
280
- return str(acc)+"%", img_list1, full_img_list1, None, None
281
- elif (sigma != 0) and (attack == "No"):
282
- return str(acc)+"%", img_list1, full_img_list1, img_list2, None
283
- elif (sigma == 0) and (attack == "Yes"):
284
- return str(acc)+"%", img_list1, full_img_list1, None, img_list3
285
- else:
286
- return str(acc)+"%", img_list1, full_img_list1, img_list2, img_list3
287
-
288
-
289
-
290
- ### TRAINING
291
- def train(epoch, net, trainloader, device, optimizer, criterion, sigma, progress=gr.Progress()):
292
- try:
293
- print('\nEpoch: %d' % epoch)
294
- net.train()
295
- train_loss = 0
296
- correct = 0
297
- total = 0
298
-
299
- iter_float = 50000/learn_batch
300
- iterations = math.ceil(iter_float)
301
- iter_prog = 0
302
-
303
- for batch_idx, (inputs, targets) in tqdm(enumerate(trainloader)):
304
- if sigma == 0:
305
- inputs, targets = inputs.to(device), targets.to(device)
306
- optimizer.zero_grad()
307
- outputs = net(inputs)
308
- else:
309
- noise = np.random.normal(0, sigma, inputs.shape)
310
- inputs += torch.tensor(noise)
311
- inputs, targets = inputs.to(device), targets.to(device)
312
- optimizer.zero_grad()
313
- outputs = net(inputs)
314
- n_inputs = inputs.clone().detach().cpu().numpy()
315
- if(batch_idx%99 == 0):
316
- fig_name = imshow(n_inputs[0], fig_name= f'figures/gaussian_noise{gaussian_num}.png')
317
- gaussian_fig = Image.open(fig_name)
318
-
319
- loss = criterion(outputs, targets)
320
- loss.backward()
321
- optimizer.step()
322
-
323
- train_loss += loss.item()
324
- _, predicted = outputs.max(1)
325
- total += targets.size(0)
326
- correct += predicted.eq(targets).sum().item()
327
-
328
- iter_prog = iter_prog + 1 # Iterating iteration amount
329
- progress(iter_prog/iterations, desc=f"Training Epoch {epoch}", total=iterations)
330
-
331
-
332
- # progress_bar(batch_idx, len(trainloader), 'Loss: %.3f | Acc: %.3f%% (%d/%d)'
333
- # % (train_loss/(batch_idx+1), 100.*correct/total, correct, total))
334
-
335
- except Exception as e:
336
- print(f"Error: {e}")
337
- gr.Warning(f"Training Error: {e}")
338
- if sigma != 0:
339
- return gaussian_fig
340
-
341
-
342
- ### TESTING
343
-
344
- def test(epoch, net, testloader, device, criterion, progress = gr.Progress()):
345
- try:
346
- net.eval()
347
- test_loss = 0
348
- correct = 0
349
- total = 0
350
-
351
- iter_float = 10000/test_batch
352
- iterations = math.ceil(iter_float)
353
- iter_prog = 0
354
-
355
- with torch.no_grad():
356
- for batch_idx, (inputs, targets) in tqdm(enumerate(testloader)):
357
- inputs, targets = inputs.to(device), targets.to(device)
358
- outputs = net(inputs)
359
- loss = criterion(outputs, targets)
360
-
361
- test_loss += loss.item()
362
- _, predicted = outputs.max(1)
363
- total += targets.size(0)
364
- correct += predicted.eq(targets).sum().item()
365
-
366
- iter_prog = iter_prog + 1 # Iterating iteration amount
367
- progress(iter_prog/iterations, desc=f"Testing Epoch {epoch}", total=iterations)
368
-
369
- wandb.log({'epoch': epoch+1, 'loss': test_loss})
370
- wandb.log({"acc": correct/total})
371
-
372
- # progress_bar(batch_idx, len(testloader), 'Loss: %.3f | Acc: %.3f%% (%d/%d)'
373
- # % (test_loss/(batch_idx+1), 100.*correct/total, correct, total))
374
-
375
- # Save checkpoint.
376
- global best_acc
377
- global acc
378
- acc = 100.*correct/total
379
- print(acc)
380
- if acc > best_acc:
381
- best_acc = acc
382
- return best_acc, predicted
383
- else:
384
- return acc, predicted
385
- # if acc > best_acc:
386
- # print('Saving..')
387
- # state = {
388
- # 'net': net.state_dict(),
389
- # 'acc': acc,
390
- # 'epoch': epoch,
391
- # }
392
- # if not os.path.isdir('checkpoint'):
393
- # os.mkdir('checkpoint')
394
- # torch.save(state, './checkpoint/ckpt.pth')
395
- # best_acc = acc
396
-
397
- except Exception as e:
398
- print(f"Error: {e}")
399
- gr.Warning(f"Testing Error: {e}")
400
-
401
-
402
- models_dict = {
403
- #"AlexNet": models.AlexNet(weights=models.AlexNet_Weights.DEFAULT),
404
- #"ConvNext_Small": models.convnext_small(weights=models.ConvNeXt_Small_Weights.DEFAULT),
405
- #"ConvNext_Base": models.convnext_base(weights=models.ConvNeXt_Base_Weights.DEFAULT),
406
- #"ConvNext_Large": models.convnext_large(weights=models.ConvNeXt_Large_Weights.DEFAULT),
407
- "DenseNet": models.densenet121(weights=models.DenseNet121_Weights.DEFAULT),
408
- #"EfficientNet_B0": models.efficientnet_b0(weights=models.EfficientNet_B0_Weights.DEFAULT),
409
- #"GoogLeNet": models.googlenet(weights=models.GoogLeNet_Weights.DEFAULT),
410
- # "InceptionNetV3": models.inception_v3(weights=models.Inception_V3_Weights.DEFAULT),
411
- # "MaxVit": models.maxvit_t(weights=models.MaxVit_T_Weights.DEFAULT),
412
- #"MnasNet0_5": models.mnasnet0_5(weights=models.MNASNet0_5_Weights.DEFAULT),
413
- #"MobileNetV2": models.mobilenet_v2(weights=models.MobileNet_V2_Weights.DEFAULT),
414
- "ResNet18": models.resnet18(weights=models.ResNet18_Weights.DEFAULT),
415
- "ResNet50": models.resnet50(weights=models.ResNet50_Weights.DEFAULT),
416
- #"RegNet_X_400MF": models.regnet_x_400mf(weights=models.RegNet_X_400MF_Weights.DEFAULT),
417
- #"ShuffleNet_V2_X0_5": models.shufflenet_v2_x0_5(weights=models.ShuffleNet_V2_X0_5_Weights.DEFAULT),
418
- #"SqueezeNet": models.squeezenet1_0(weights=models.SqueezeNet1_0_Weights.DEFAULT),
419
- "VGG19": models.vgg19(weights=models.VGG19_Weights.DEFAULT)
420
- }
421
-
422
- # Store dictionary keys into list for dropdown menu choices
423
- names = list(models_dict.keys())
424
-
425
- # Optimizer names
426
- optimizers = ["SGD","Adam"]
427
-
428
- # Scheduler names
429
- schedulers = ["None","CosineAnnealingLR","ReduceLROnPlateau","StepLR"]
430
-
431
- ### GRADIO APP INTERFACE
432
-
433
- def togglepicsettings(choice):
434
- yes=gr.Gallery(visible=True)
435
- no=gr.Gallery(visible=False)
436
- if choice == "Yes":
437
- return yes,no
438
- else:
439
- return no,yes
440
-
441
- def settings(choice):
442
- if choice == "Advanced":
443
- advanced = [
444
- gr.Slider(visible=True),
445
- gr.Slider(visible=True),
446
- gr.Slider(visible=True),
447
- gr.Dropdown(visible=True),
448
- gr.Dropdown(visible=True),
449
- gr.Radio(visible=True)
450
- ]
451
- return advanced
452
- else:
453
- basic = [
454
- gr.Slider(visible=False),
455
- gr.Slider(visible=False),
456
- gr.Slider(visible=False),
457
- gr.Dropdown(visible=False),
458
- gr.Dropdown(visible=False),
459
- gr.Radio(visible=False)
460
- ]
461
- return basic
462
-
463
- def attacks(choice):
464
- if choice == "Yes":
465
- yes = [
466
- gr.Markdown(visible=True),
467
- gr.Radio(visible=True),
468
- gr.Radio(visible=True)
469
- ]
470
- return yes
471
- if choice == "No":
472
- no = [
473
- gr.Markdown(visible=False),
474
- gr.Radio(visible=False),
475
- gr.Radio(visible=False)
476
- ]
477
- return no
478
-
479
- def gaussian(choice):
480
- if choice == "Yes":
481
- yes = [
482
- gr.Slider(visible=True),
483
- gr.Gallery(visible=True),
484
- ]
485
- return yes
486
- else:
487
- no = [
488
- gr.Slider(visible=False),
489
- gr.Gallery(visible=False),
490
- ]
491
- return no
492
- def adversarial(choice):
493
- if choice == "Yes":
494
- yes = gr.Gallery(visible=True)
495
- return yes
496
- else:
497
- no = gr.Gallery(visible=False)
498
-
499
- ## Main app for functionality
500
- with gr.Blocks(css=".caption-label {display:none}") as functionApp:
501
- with gr.Row():
502
- gr.Markdown("# CIFAR-10 Model Training GUI")
503
- with gr.Row():
504
- gr.Markdown("## Parameters")
505
- with gr.Row():
506
- inp = gr.Dropdown(choices=names, label="Training Model", value="ResNet18", info="Choose one of 13 common models provided in the dropdown to use for training.")
507
- username = gr.Textbox(label="Weights and Biases", info="Enter your username or team name from the Weights and Biases API.")
508
- epochs_sldr = gr.Slider(label="Number of Epochs", minimum=1, maximum=100, step=1, value=1, info="How many times the model will see the entire dataset during trianing.")
509
- with gr.Column():
510
- setting_radio = gr.Radio(["Basic", "Advanced"], label="Settings", value="Basic")
511
- btn = gr.Button("Run")
512
- with gr.Row():
513
- train_sldr = gr.Slider(visible=False, label="Training Batch Size", minimum=1, maximum=1000, step=1, value=128, info="The number of training samples processed before the model's internal parameters are updated.")
514
- test_sldr = gr.Slider(visible=False, label="Testing Batch Size", minimum=1, maximum=1000, step=1, value=100, info="The number of testing samples processed at once during the evaluation phase.")
515
- learning_rate_sldr = gr.Slider(visible=False, label="Learning Rate", minimum=0.0001, maximum=0.1, step=0.0001, value=0.001, info="The learning rate of the optimization program.")
516
- optimizer = gr.Dropdown(visible=False, label="Optimizer", choices=optimizers, value="SGD", info="The optimization algorithm used to minimize the loss function during training.")
517
- scheduler = gr.Dropdown(visible=False, label="Scheduler", choices=schedulers, value="CosineAnnealingLR", info="The scheduler used to iteratively alter learning rate.")
518
- use_attacks = gr.Radio(["Yes", "No"], visible=False, label="Use Attacking Methods?", value="No")
519
- setting_radio.change(fn=settings, inputs=setting_radio, outputs=[train_sldr, test_sldr, learning_rate_sldr, optimizer, scheduler, use_attacks])
520
- with gr.Row():
521
- attack_method = gr.Markdown("## Attacking Methods", visible=False)
522
- with gr.Row():
523
- use_sigma = gr.Radio(["Yes","No"], visible=False, label="Use Gaussian Noise?", value="No")
524
- sigma_sldr = gr.Slider(visible=False, label="Gaussian Noise", minimum=0, maximum=1, value=0, step=0.1, info="The sigma value of the gaussian noise eqaution. A value of 0 disables gaussian noise.")
525
- adv_attack = gr.Radio(["Yes","No"], visible=False, label="Use Adversarial Attacks?", value="No")
526
- with gr.Row():
527
- gr.Markdown("## Training Results")
528
- with gr.Row():
529
- accuracy = gr.Textbox(label = "Accuracy", info="The validation accuracy of the trained model (accuracy evaluated on testing data).")
530
- with gr.Column():
531
- showpics = gr.Radio(["Yes","No"], visible = True, label = "Show all pictures?", value = "No")
532
- pics = gr.Gallery(preview=False, selected_index=0, object_fit='contain', label="Testing Images")
533
- allpics = gr.Gallery(preview=True, selected_index=0, object_fit='contain', label="Full Testing Images",visible = False)
534
- showpics.change(fn=togglepicsettings, inputs=[showpics], outputs = [allpics, pics])
535
- with gr.Row():
536
- gaussian_pics = gr.Gallery(visible=False, preview=False, selected_index=0, object_fit='contain', label="Gaussian Noise")
537
- attack_pics = gr.Gallery(visible=False, preview=False, selected_index=0, object_fit='contain', label="Adversarial Attack")
538
- use_attacks.change(fn=attacks, inputs=use_attacks, outputs=[attack_method, use_sigma, adv_attack])
539
- use_sigma.change(fn=gaussian, inputs=use_sigma, outputs=[sigma_sldr, gaussian_pics])
540
- adv_attack.change(fn=adversarial, inputs=adv_attack, outputs=attack_pics)
541
- btn.click(fn=main, inputs=[inp, epochs_sldr, train_sldr, test_sldr, learning_rate_sldr, optimizer, sigma_sldr, adv_attack, username, scheduler], outputs=[accuracy, pics, allpics, gaussian_pics, attack_pics])
542
-
543
- ## Documentation app (implemented as second tab)
544
-
545
- markdown_file_path = 'documentation.md'
546
- with open(markdown_file_path, 'r') as file:
547
- markdown_content = file.read()
548
-
549
- with gr.Blocks() as documentationApp:
550
- with gr.Row():
551
- gr.Markdown("# CIFAR-10 Training Interface Documentation")
552
- with gr.Row():
553
- gr.Markdown(markdown_content) # Can be collapesed in VSCode to hide paragraphs from view. Vscode can also wrap text.
554
-
555
- ### LAUNCH APP
556
-
557
- if __name__ == '__main__':
558
- mainApp = gr.TabbedInterface([functionApp, documentationApp], ["Welcome", "Documentation"], theme=theme)
559
- mainApp.queue()
560
- mainApp.launch()