-
Notifications
You must be signed in to change notification settings - Fork 1
/
main_covid_prediction.py
557 lines (481 loc) · 20.3 KB
/
main_covid_prediction.py
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
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
"""
This main would predict covid
"""
# %% Import
###############
##### LIB #####
###############
import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import matplotlib.pyplot as plt
import seaborn as sns
import os
import sys
from dataclasses import dataclass, field
from typing import Dict, Any, List, Optional
import datetime
from PIL import Image
# ML:
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.utils.data import Dataset
import torchvision.models as models
#######################
##### LOCAL LIB #######
#######################
## USER DEFINED:
ABS_PATH = "/home/jx/JX_Project/covid-xray-detection" # Define ur absolute path here
DATA_DIRECTORY = "data-latest"
## Custom Files:
def abspath(relative_path):
return os.path.join(ABS_PATH, relative_path)
def abs_data_path(relative_path):
return os.path.join(os.path.join(ABS_PATH, DATA_DIRECTORY), relative_path)
module_path = os.path.abspath(os.path.join('..'))
if module_path not in sys.path:
sys.path.append(abspath("src_code"))
import jx_lib
import jx_pytorch_lib
# %% USER OPTION: ----- ----- ----- ----- ----- ----- ----- ----- ----- #
#######################
##### PREFERENCE ######
#######################
# SELECTED_TARGET = "1LAYER" # <--- select model !!!
SELECTED_TARGET = "CUSTOM-MODEL" # <--- select model !!!
USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET_400 = False # use 400x400 resolution
USE_PREPROCESS_CUSTOM_DATASET = True # True, to use dataset generated by 'tool_data_gen.py' (differential RGB only)
USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET = True # True, to use dataset generated by 'tool_data_gen.py' (differential RGB + Augmentation)
PRINT_SAMPLES = True
OUTPUT_MODEL = False
USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET_POST_HOMO = True
USE_COMPETITION_EVAL = False
# %% LOAD DATASET INFO: ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- #
#######################
##### LOAD DATASET ####
#######################
LUT_HEADER = ["[patient id]", "[filename]", "[class]", "[data source]"]
# import data
if USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET:
TRAIN_DATA_LUT = pd.read_csv(abs_data_path("pre-processed-[train].txt"), sep=" ", header=None, names=LUT_HEADER)
else:
TRAIN_DATA_LUT = pd.read_csv(abs_data_path("train.txt"), sep=" ", header=None, names=LUT_HEADER)
VALID_DATA_LUT = pd.read_csv(abs_data_path("test.txt"), sep=" ", header=None, names=LUT_HEADER)
# convert class to label 'y'
LABEL_TO_INT_LUT = {
"positive": 1,
"negative": 0,
}
INT_TO_LABEL_LUT = {
1: "positive",
0: "negative",
}
def class_to_binary(cls):
return [LABEL_TO_INT_LUT[c] for c in cls]
TRAIN_DATA_LUT["Y"] = class_to_binary(TRAIN_DATA_LUT["[class]"])
VALID_DATA_LUT["Y"] = class_to_binary(VALID_DATA_LUT["[class]"])
# convert filename to absolute path:
def filename_to_abspath(filenames, tag):
return [abs_data_path("{}/{}".format(tag, filename)) for filename in filenames]
if USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET_400:
TRAIN_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=TRAIN_DATA_LUT["[filename]"], tag="train-custom-with-aug-400")
VALID_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=VALID_DATA_LUT["[filename]"], tag="valid-custom-400")
elif USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET_POST_HOMO:
TRAIN_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=TRAIN_DATA_LUT["[filename]"], tag="train-custom-post")
VALID_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=VALID_DATA_LUT["[filename]"], tag="valid-custom-post")
elif USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET:
TRAIN_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=TRAIN_DATA_LUT["[filename]"], tag="train-custom-with-aug")
VALID_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=VALID_DATA_LUT["[filename]"], tag="valid-custom")
elif USE_PREPROCESS_CUSTOM_DATASET:
TRAIN_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=TRAIN_DATA_LUT["[filename]"], tag="train-custom")
VALID_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=VALID_DATA_LUT["[filename]"], tag="valid-custom")
else:
TRAIN_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=TRAIN_DATA_LUT["[filename]"], tag="train")
VALID_DATA_LUT["img_abs_path"] = filename_to_abspath(filenames=VALID_DATA_LUT["[filename]"], tag="test")
# report status:
def report_status(data, tag):
tp, tn = np.sum(data["Y"] == 1), np.sum(data["Y"] == 0)
print("{}: +:{}, -:{}".format(tag, tp, tn))
report_status(data=TRAIN_DATA_LUT, tag="train")
report_status(data=VALID_DATA_LUT, tag="valid")
# Check files:
tick = 0
for path in TRAIN_DATA_LUT["img_abs_path"]:
if os.path.isfile(path):
tick += 1
else:
print("Missing: ", path)
print("Number of Existed Files: ",tick)
# %% BALANCE TRAINING DATASET -------------------------------- ####
"""
Since we notice the imbalance in training dataset, let's try random downsampling.
"""
train_pos = TRAIN_DATA_LUT[TRAIN_DATA_LUT["Y"] == 1]
train_neg = TRAIN_DATA_LUT[TRAIN_DATA_LUT["Y"] == 0]
N_balanced = min(len(train_pos), len(train_neg))
# shuffle and resample:
train_pos = train_pos[0:N_balanced]
train_neg = train_neg[0:N_balanced]
NEW_TRAIN_DATA_LUT = pd.concat([train_pos, train_neg])
report_status(data=train_pos, tag="new:train_pos")
report_status(data=train_neg, tag="new:train_neg")
report_status(data=NEW_TRAIN_DATA_LUT, tag="new:train")
TRAIN_DATA_LUT = NEW_TRAIN_DATA_LUT
# %% CONFIG: ----- ----- ----- ----- ----- ----- ----- ----- ----- #
@dataclass
class PredictorConfiguration:
MODEL_TAG : str = "default"
OUT_DIR : str = ""
OUT_DIR_MODELS : str = ""
VERSION : str = "default"
# Settings:
TOTAL_NUM_EPOCHS : int = 5
LEARNING_RATE : float = 0.001
BATCH_SIZE : int = 1000
LOSS_FUNC : nn = nn.NLLLoss()
OPTIMIZER : optim = None
# early stopping:
EARLY_STOPPING_DECLINE_CRITERION : int = 5
# %% MODEL:
# @Credit to https://jarvislabs.ai/blogs/resnet tutorial on resnet34 from scratch
class BasicBlock(nn.Module):
expansion = 1
def __init__(self, inplanes, planes, stride=1, downsample=None):
super().__init__()
self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
self.bn1 = nn.BatchNorm2d(planes)
self.relu = nn.ReLU(inplace=True)
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.downsample = downsample
self.stride = stride
def forward(self, x):
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
if self.downsample is not None:
identity = self.downsample(x)
out += identity
out = self.relu(out)
return out
class ResNet(nn.Module):
def __init__(self, block, layers, num_classes=1000):
super().__init__()
self.inplanes = 64
self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3, bias=False)
self.bn1 = nn.BatchNorm2d(self.inplanes)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
self.layer1 = self._make_layer(block, 64, layers[0])
self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.fc = nn.Linear(512 , num_classes)
def _make_layer(self, block, planes, blocks, stride=1):
downsample = None
if stride != 1 or self.inplanes != planes:
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes, 1, stride, bias=False),
nn.BatchNorm2d(planes),
)
layers = []
layers.append(block(self.inplanes, planes, stride, downsample))
self.inplanes = planes
for _ in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def forward(self, x):
x = self.conv1(x) # 224x224
x = self.bn1(x)
x = self.relu(x)
x = self.maxpool(x) # 112x112
x = self.layer1(x) # 56x56
x = self.layer2(x) # 28x28
x = self.layer3(x) # 14x14
x = self.layer4(x) # 7x7
x = self.avgpool(x) # 1x1
x = torch.flatten(x, 1) # remove 1 X 1 grid and make vector of tensor shape
x = self.fc(x)
return x
# %% INIT: ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- #
#############################
##### MODEL DEFINITION ######
#############################
### MODEL ###
MODEL_DICT = {
"CUSTOM-MODEL": { # <--- name your model
"model":
nn.Sequential(
# Feature Extraction:
# ResNet(BasicBlock, [3,4,6,3], num_classes=2), # ResNet base v6
# ResNet(BasicBlock, [0,1,1,1], num_classes=2), # ResNet reduced v8 - ResNet10 - ablation
ResNet(BasicBlock, [1,1,1,1], num_classes=2), # ResNet reduced v8 - ResNet10
# ResNet(BasicBlock, [1,2,3,2], num_classes=2), # ResNet reduced v7
# Classifier:
nn.Softmax(dim=1),
),
"config":
PredictorConfiguration(
# VERSION="v6-base-model-resnet34", # <--- name your run
# VERSION="v7-reduced-model", # <--- name your run
# VERSION="v8-reduced-model", # <--- name your run
# VERSION="v8-reduced-model-ablation-1", # <--- name your run
VERSION="latest-v8-reduced-model", # <--- name your run
OPTIMIZER=optim.SGD,
LEARNING_RATE=0.01,
BATCH_SIZE=100,
TOTAL_NUM_EPOCHS=200,#50
EARLY_STOPPING_DECLINE_CRITERION=30,# No stopping
),
"transformation":
transforms.Compose([
# same:
# transforms.Resize(320),
transforms.CenterCrop(320),
transforms.ToTensor(),
transforms.Normalize((0.5), (0.5)),
]),
},
}
#############################
##### MODEL AUTOMATION ######
#############################
# select model:
SELECTED_NET_MODEL = MODEL_DICT[SELECTED_TARGET]["model"]
SELECTED_NET_CONFIG = MODEL_DICT[SELECTED_TARGET]["config"]
SELECTED_NET_TRANSFORMATION = MODEL_DICT[SELECTED_TARGET]["transformation"]
# model specific declaration:
SELECTED_NET_CONFIG.MODEL_TAG = SELECTED_TARGET
SELECTED_NET_CONFIG.OPTIMIZER = SELECTED_NET_CONFIG.OPTIMIZER(
SELECTED_NET_MODEL.parameters(), lr=SELECTED_NET_CONFIG.LEARNING_RATE
)
### Directory generation ###
OUT_DIR = abspath("output")
MODEL_OUT_DIR = "{}/{}".format(OUT_DIR, SELECTED_TARGET)
SELECTED_NET_CONFIG.OUT_DIR = "{}/{}".format(MODEL_OUT_DIR, SELECTED_NET_CONFIG.VERSION)
SELECTED_NET_CONFIG.OUT_DIR_MODELS = "{}/{}".format(SELECTED_NET_CONFIG.OUT_DIR, "models")
jx_lib.create_folder(DIR=OUT_DIR)
jx_lib.create_folder(DIR=MODEL_OUT_DIR)
jx_lib.create_folder(DIR=SELECTED_NET_CONFIG.OUT_DIR)
jx_lib.create_folder(DIR=SELECTED_NET_CONFIG.OUT_DIR_MODELS)
# define logger:
def _print(content):
print("[ENGINE] ", content)
with open(os.path.join(SELECTED_NET_CONFIG.OUT_DIR,"log.txt"), "a") as log_file:
log_file.write("\n")
log_file.write("[{}]: {}".format(datetime.datetime.now(), content))
# log model:
_print(" USER PREFERENCE: \n > {}:{}\n > {}:{}\n > {}:{}\n > {}:{}\n".format(
"SELECTED_TARGET", SELECTED_TARGET,
"USE_PREPROCESS_CUSTOM_DATASET", USE_PREPROCESS_CUSTOM_DATASET,
"USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET", USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET,
"PRINT_SAMPLES", PRINT_SAMPLES
))
_print(str(SELECTED_NET_MODEL))
_print(str(SELECTED_NET_CONFIG))
_print(str(SELECTED_NET_TRANSFORMATION))
#%% LOAD NET: ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- #
##########################
##### GPU CUDA ACC. ######
##########################
# check device:
# hardware-acceleration
device = None
if torch.cuda.is_available():
_print("[ALERT] Attempt to use GPU => CUDA:0")
device = torch.device("cuda:0")
else:
_print("[ALERT] GPU not found, use CPU!")
device = torch.device("cpu")
SELECTED_NET_MODEL.to(device)
# %% LOAD DATASET: ----- ----- ----- ----- ----- ----- ----- ----- #####
###########################
##### DATASET LOADER ######
###########################
# define custom dataset methods:
class CTscanDataSet(Dataset):
def __init__(self, list_of_img_dir, transform, labels):
self.list_of_img_dir = list_of_img_dir
self.transform = transform
self.labels = labels
def __len__(self):
return len(self.list_of_img_dir)
def __getitem__(self, idx):
img_loc = self.list_of_img_dir[idx]
img = Image.open(img_loc).convert('RGB')
# arr = np.array(img)
# norm_arr = arr / 255
# new_img = Image.fromarray(norm_arr.astype('float'),'RGB')
img_transformed = self.transform(img)
return (img_transformed, self.labels[idx])
def _report(self):
N_total = len(self.labels)
N_pos = np.sum(self.labels)
N_neg = N_total-N_pos
tag = "BALANCED." if N_pos == N_neg else "UNBALANCED !!!"
return "+: {1}/{0} ({3:.2f}%) -: {2}/{0} ({4:.2f}%) [{5}]".format(
N_total, N_pos, N_neg, N_pos/N_total*100, N_neg/N_total*100, tag
)
# load image:
img_dataset_train = CTscanDataSet(
list_of_img_dir=TRAIN_DATA_LUT["img_abs_path"],
transform=SELECTED_NET_TRANSFORMATION, labels=TRAIN_DATA_LUT["Y"]
)
img_dataset_valid = CTscanDataSet(
list_of_img_dir=VALID_DATA_LUT["img_abs_path"],
transform=SELECTED_NET_TRANSFORMATION, labels=VALID_DATA_LUT["Y"]
)
# Prep. dataloader
train_dataloader = torch.utils.data.DataLoader(
img_dataset_train,
batch_size=SELECTED_NET_CONFIG.BATCH_SIZE, shuffle=True
)
valid_dataloader = torch.utils.data.DataLoader(
img_dataset_valid,
batch_size=SELECTED_NET_CONFIG.BATCH_SIZE, shuffle=True
)
_print("=== Dataset Loaded:")
_print("> Train Dataset: {}".format(train_dataloader.dataset._report()))
_print("> Valid Dataset: {}".format(valid_dataloader.dataset._report()))
# %% Load Competition dataset:
#############################
##### COMPETITION DATASET ###
#############################
if USE_COMPETITION_EVAL:
N_TEST = 400
if USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET_400:
COMPETITION_PATH = [ abs_data_path("competition_test-custom-400/{}.png".format(i+1)) for i in range(N_TEST) ]
elif USE_PREPROCESS_CUSTOM_DATASET or USE_PREPROCESS_AUGMENTED_CUSTOM_DATASET:
COMPETITION_PATH = [ abs_data_path("competition_test-custom/{}.png".format(i+1)) for i in range(N_TEST) ]
else:
COMPETITION_PATH = [ abs_data_path("competition_test/{}.png".format(i+1)) for i in range(N_TEST) ]
COMPETITION_Y = np.zeros(N_TEST)
img_dataset_competition = CTscanDataSet(
list_of_img_dir=COMPETITION_PATH,
transform=SELECTED_NET_TRANSFORMATION, labels=COMPETITION_Y
)
competition_dataloader = torch.utils.data.DataLoader(
img_dataset_competition,
batch_size=SELECTED_NET_CONFIG.BATCH_SIZE, shuffle=False
)
_print("=== Dataset Loaded:")
_print("> Competition Dataset: {}".format(competition_dataloader.dataset._report()))
else:
competition_dataloader = None
# %% PRINT SAMPLE: ----- ----- ----- ----- ----- ----- ---
########################################
##### CONSTRUCT SAMPLE IMAGE PLOT ######
########################################
def plot_sample_from_dataloader(dataloader, tag:str, N_COLS = 4, N_MAX=20):
N_MAX = min(SELECTED_NET_CONFIG.BATCH_SIZE, N_MAX)
N_COLS = min(N_COLS, N_MAX)
N_ROWS = int(np.ceil(N_MAX/N_COLS)) * 2
fig, axes = plt.subplots(
figsize=(N_COLS * 8, N_ROWS * 8),
ncols=N_COLS, nrows=N_ROWS
)
_print("=== Print Sample Data ({}) [n_display:{} / batch_size:{}]".format(
tag, N_MAX, SELECTED_NET_CONFIG.BATCH_SIZE))
# get one batch:
images, labels = next(iter(dataloader))
for i in range(N_MAX):
print("\r >[{}/{}]".format(i+1,N_MAX), end='')
# Plot img:
id_ = i * 2
ax = axes[int(id_/N_COLS), id_%N_COLS]
# show remapped image, since the range was distorted by normalization
ax.imshow((np.dstack((images[i][0], images[i][1], images[i][2])) + 1)/2, vmin=0, vmax=1)
ax.set_title(
"{}".format(INT_TO_LABEL_LUT[int(labels[i])]),
color="red" if int(labels[i]) else "blue"
)
# Plot Hist:
id_ += 1
ax = axes[int(id_/N_COLS), id_%N_COLS]
ax.hist(np.ravel(images[i][0]), bins=256, color='r', alpha = 0.5, range=[-1, 1])
ax.hist(np.ravel(images[i][1]), bins=256, color='g', alpha = 0.5, range=[-1, 1])
ax.hist(np.ravel(images[i][2]), bins=256, color='b', alpha = 0.5, range=[-1, 1])
ax.legend(['R', 'G', 'B'])
ax.set_title("<- Histogram")
fig.savefig("{}/plot_{}.png".format(SELECTED_NET_CONFIG.OUT_DIR, tag), bbox_inches = 'tight')
if PRINT_SAMPLES:
plot_sample_from_dataloader(train_dataloader, tag="training-sample")
plot_sample_from_dataloader(valid_dataloader, tag="validation-sample")
if USE_COMPETITION_EVAL:
plot_sample_from_dataloader(competition_dataloader, tag="competition-sample")
# %% DEFINE EVALUATION WITH COMPETITION TEST DATASET -------------------------------- %%
################################################################
##### DEFINE EVALUATION CALLBACK FOR COMPETITION PREDICTION ####
################################################################
from sklearn.metrics import confusion_matrix, classification_report
def evaluate_net(net, dataloader):
_print("> Evaluation Begin ...")
y_true = []
y_pred = []
for X, y in dataloader:
if device != None:
X = X.to(device)
y = y.to(device)
# Predict:
y_prediction = net(X)
# record:
y_true.extend(y.cpu().detach().numpy())
y_pred.extend(y_prediction.argmax(dim=1).cpu().detach().numpy())
_print("> Evaluation Complete")
return y_true, y_pred
def eval_competition(net, dataloader, tag):
# eval:
y_true, y_pred = evaluate_net(net=net, dataloader=dataloader)
_print("> {3} +:{1}/{0} -:{2}/{0}".format(N_TEST, np.sum(y_pred), N_TEST-np.sum(y_pred), tag))
OUT_FILE_PATH = "{}/y_pred[{}].txt".format(SELECTED_NET_CONFIG.OUT_DIR, tag)
with open(OUT_FILE_PATH, "w") as file_out:
file_out.write("\n".join(["{}".format(yi) for yi in y_pred]))
if not USE_COMPETITION_EVAL:
eval_competition = None
# %% TRAIN: ----- ----- ----- ----- ----- ----- ---
#####################
###### M A I N ######
#####################
# Reload:
import importlib
importlib.reload(jx_pytorch_lib)
importlib.reload(jx_lib)
from jx_pytorch_lib import ProgressReport, VerboseLevel, CNN_MODEL_TRAINER
# run:
report, best_net = CNN_MODEL_TRAINER.train_and_monitor(
device=device,
train_dataset=train_dataloader,
test_dataset=valid_dataloader,
optimizer=SELECTED_NET_CONFIG.OPTIMIZER,
loss_func=SELECTED_NET_CONFIG.LOSS_FUNC,
net=SELECTED_NET_MODEL,
num_epochs=SELECTED_NET_CONFIG.TOTAL_NUM_EPOCHS,
model_output_path=SELECTED_NET_CONFIG.OUT_DIR_MODELS,
target_names=LABEL_TO_INT_LUT,
early_stopping_n_epochs_consecutive_decline=SELECTED_NET_CONFIG.EARLY_STOPPING_DECLINE_CRITERION,
eval_func_competition=eval_competition,
eval_data_competition=competition_dataloader,
# max_data_samples=20,
verbose_level= VerboseLevel.HIGH,
_print=_print,
# save_model=OUTPUT_MODEL,
)
report.output_progress_plot(
OUT_DIR=SELECTED_NET_CONFIG.OUT_DIR,
tag=SELECTED_NET_CONFIG.VERSION,
verbose_level=VerboseLevel.HIGH
)
report.save(
OUT_DIR=SELECTED_NET_CONFIG.OUT_DIR,
tag=SELECTED_NET_CONFIG.VERSION,
)
#eval:
eval_competition(net=best_net, dataloader=competition_dataloader, tag="best")
eval_competition(net=SELECTED_NET_MODEL, dataloader=competition_dataloader, tag="final")
# %%