fix: remove concurrency, refactor (7ac12bd9) · Commits · Václav Novák / Signature Verification

main.py

+1 −14

Original line number	Diff line number	Diff line
		@@ -15,7 +15,6 @@ from src.evaluation import evaluate
		NUM_WORKERS = 0 if platform.system() == "Windows" else 6
		CUDA_CAPABLE = torch.cuda.is_available()
		MAX_SECONDS_PER_TRIAL = 7200 if len(sys.argv) < 2 else int(sys.argv[1])
		STORAGE_PATH = "results.log" if len(sys.argv) < 3 else sys.argv[2]

		###############################################################################

		@@ -55,19 +54,7 @@ def objective(trial):
		###############################################################################


		if platform.system() != "Windows":
		storage = optuna.storages.JournalStorage(
		optuna.storages.JournalFileStorage(STORAGE_PATH),
		)

		else:
		lock_obj = optuna.storages.JournalFileOpenLock(STORAGE_PATH)
		storage = optuna.storages.JournalStorage(
		optuna.storages.JournalFileStorage(STORAGE_PATH, lock_obj=lock_obj),
		)


		if __name__ == '__main__':
		study = optuna.create_study(direction="maximize", sampler=optuna.samplers.TPESampler(), storage=storage)
		study = optuna.create_study(direction="maximize", sampler=optuna.samplers.TPESampler())

		study.optimize(objective)

src/builders.py

+2 −7

Original line number	Diff line number	Diff line
		@@ -3,7 +3,6 @@ from torch.utils.data import DataLoader, Dataset
		from typing import Iterator, Tuple
		from optuna import Trial

		import torchvision.transforms as transforms
		import torch
		import torch.nn as nn

		@@ -14,15 +13,11 @@ def build_dataset(dataset_dir: str, is_train: bool, image_input_size: int) -> Si

		@param dataset_dir: str, path to dataset
		@param is_train: bool, indicates whether the dataset should load test or train data
		@param image_input_size: int, dimension of image taken from dataset
		@param image_input_size: int, dimension to which data should be scaled

		@return: SiameseNetworkDataset, dataset
		"""
		return SiameseNetworkDataset(dataset_dir, is_train,
		transform=transforms.Compose([transforms.Resize((image_input_size,
		image_input_size)),
		transforms.ToTensor()
		]))
		return SiameseNetworkDataset(dataset_dir, is_train, image_input_size)


		def build_dataloader(dataset: Dataset, batch_size: int, shuffle: bool, num_workers: int) -> DataLoader:

src/data_handlers.py

+16 −8

Original line number	Diff line number	Diff line
		@@ -3,9 +3,11 @@ import csv

		from PIL import Image
		from platform import system
		from typing import Tuple, List, Optional
		from typing import Tuple, List
		from torchvision.transforms import Compose
		from torch.utils.data import Dataset
		import torchvision.transforms as transforms



		class SiameseNetworkDataset(Dataset):
		@@ -14,18 +16,19 @@ class SiameseNetworkDataset(Dataset):
		return two images instead of one.
		"""

		def __init__(self, data_dir: str, is_train: bool = False, transform: Optional[Compose] = None):
		def __init__(self, data_dir: str, is_train: bool,
		image_input_size: int):
		"""
		Constructor.

		@param data_dir: str, path to directory with dataset, has to contain index files with paths to images
		(e.g. train_unix.csv, test_unix.csv...)
		@param is_train: bool, indicates wherether to load train or test data
		@param transform: Compose, transforms to be applied on each image
		@image_input_size: int, dimension to which data should be scaled
		"""
		self.data_dir = data_dir
		self.variations = self.get_variations(is_train)
		self.transform = transform
		self.image_input_size = image_input_size

		def get_variations(self, is_train) -> List[Tuple[str, str, str]]:
		"""
		@@ -53,12 +56,17 @@ class SiameseNetworkDataset(Dataset):
		img0 = Image.open(image_variation[0])
		img1 = Image.open(image_variation[1])

		# preprocessing
		img0 = img0.convert("L").point(lambda p: 255 if p > 242 else p)
		img1 = img1.convert("L").point(lambda p: 255 if p > 242 else p)

		if self.transform is not None:
		img0 = self.transform(img0)
		img1 = self.transform(img1)
		current_transform = transforms.Compose([transforms.Resize((self.image_input_size,
		self.image_input_size)),
		transforms.ToTensor()])

		img0 = current_transform(img0)
		img1 = current_transform(img1)


		return img0, img1, int(image_variation[2])