Skip to content
Snippets Groups Projects
Commit 6a1099c6 authored by John Carter's avatar John Carter
Browse files

small label change

parent 6f53474a
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
MetaAugment/Baseline_JC.ipynb
+ 1
− 1
View file @ 6a1099c6
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
import numpy as np import numpy as np
import torch import torch
import torch.nn as nn import torch.nn as nn
import torch.nn.functional as F import torch.nn.functional as F
import torch.optim as optim import torch.optim as optim
import torch.utils.data as data_utils import torch.utils.data as data_utils
import torchvision import torchvision
import torchvision.datasets as datasets import torchvision.datasets as datasets
from tqdm import trange from tqdm import trange
``` ```
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
"""Define internal NN module that trains on the dataset""" """Define internal NN module that trains on the dataset"""
class LeNet(nn.Module): class LeNet(nn.Module):
def __init__(self, img_height, img_width, num_labels, img_channels): def __init__(self, img_height, img_width, num_labels, img_channels):
super().__init__() super().__init__()
self.conv1 = nn.Conv2d(img_channels, 6, 5) self.conv1 = nn.Conv2d(img_channels, 6, 5)
self.relu1 = nn.ReLU() self.relu1 = nn.ReLU()
self.pool1 = nn.MaxPool2d(2) self.pool1 = nn.MaxPool2d(2)
self.conv2 = nn.Conv2d(6, 16, 5) self.conv2 = nn.Conv2d(6, 16, 5)
self.relu2 = nn.ReLU() self.relu2 = nn.ReLU()
self.pool2 = nn.MaxPool2d(2) self.pool2 = nn.MaxPool2d(2)
self.fc1 = nn.Linear(int((((img_height-4)/2-4)/2)*(((img_width-4)/2-4)/2)*16), 120) self.fc1 = nn.Linear(int((((img_height-4)/2-4)/2)*(((img_width-4)/2-4)/2)*16), 120)
self.relu3 = nn.ReLU() self.relu3 = nn.ReLU()
self.fc2 = nn.Linear(120, 84) self.fc2 = nn.Linear(120, 84)
self.relu4 = nn.ReLU() self.relu4 = nn.ReLU()
self.fc3 = nn.Linear(84, num_labels) self.fc3 = nn.Linear(84, num_labels)
self.relu5 = nn.ReLU() self.relu5 = nn.ReLU()
def forward(self, x): def forward(self, x):
y = self.conv1(x) y = self.conv1(x)
y = self.relu1(y) y = self.relu1(y)
y = self.pool1(y) y = self.pool1(y)
y = self.conv2(y) y = self.conv2(y)
y = self.relu2(y) y = self.relu2(y)
y = self.pool2(y) y = self.pool2(y)
y = y.view(y.shape[0], -1) y = y.view(y.shape[0], -1)
y = self.fc1(y) y = self.fc1(y)
y = self.relu3(y) y = self.relu3(y)
y = self.fc2(y) y = self.fc2(y)
y = self.relu4(y) y = self.relu4(y)
y = self.fc3(y) y = self.fc3(y)
y = self.relu5(y) y = self.relu5(y)
return y return y
``` ```
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
"""Define internal NN module that trains on the dataset""" """Define internal NN module that trains on the dataset"""
class EasyNet(nn.Module): class EasyNet(nn.Module):
def __init__(self, img_height, img_width, num_labels, img_channels): def __init__(self, img_height, img_width, num_labels, img_channels):
super().__init__() super().__init__()
self.fc1 = nn.Linear(img_height*img_width*img_channels, 2048) self.fc1 = nn.Linear(img_height*img_width*img_channels, 2048)
self.relu1 = nn.ReLU() self.relu1 = nn.ReLU()
self.fc2 = nn.Linear(2048, num_labels) self.fc2 = nn.Linear(2048, num_labels)
self.relu2 = nn.ReLU() self.relu2 = nn.ReLU()
def forward(self, x): def forward(self, x):
y = x.view(x.shape[0], -1) y = x.view(x.shape[0], -1)
y = self.fc1(y) y = self.fc1(y)
y = self.relu1(y) y = self.relu1(y)
y = self.fc2(y) y = self.fc2(y)
y = self.relu2(y) y = self.relu2(y)
return y return y
``` ```
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
"""Define internal NN module that trains on the dataset""" """Define internal NN module that trains on the dataset"""
class SimpleNet(nn.Module): class SimpleNet(nn.Module):
def __init__(self, img_height, img_width, num_labels, img_channels): def __init__(self, img_height, img_width, num_labels, img_channels):
super().__init__() super().__init__()
self.fc1 = nn.Linear(img_height*img_width*img_channels, num_labels) self.fc1 = nn.Linear(img_height*img_width*img_channels, num_labels)
self.relu1 = nn.ReLU() self.relu1 = nn.ReLU()
def forward(self, x): def forward(self, x):
y = x.view(x.shape[0], -1) y = x.view(x.shape[0], -1)
y = self.fc1(y) y = self.fc1(y)
y = self.relu1(y) y = self.relu1(y)
return y return y
``` ```
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
"""Make toy dataset""" """Make toy dataset"""
def create_toy(train_dataset, test_dataset, batch_size, n_samples): def create_toy(train_dataset, test_dataset, batch_size, n_samples):
# shuffle and take first n_samples %age of training dataset # shuffle and take first n_samples %age of training dataset
shuffle_order_train = np.random.RandomState(seed=100).permutation(len(train_dataset)) shuffle_order_train = np.random.RandomState(seed=100).permutation(len(train_dataset))
shuffled_train_dataset = torch.utils.data.Subset(train_dataset, shuffle_order_train) shuffled_train_dataset = torch.utils.data.Subset(train_dataset, shuffle_order_train)
indices_train = torch.arange(int(n_samples*len(train_dataset))) indices_train = torch.arange(int(n_samples*len(train_dataset)))
reduced_train_dataset = data_utils.Subset(shuffled_train_dataset, indices_train) reduced_train_dataset = data_utils.Subset(shuffled_train_dataset, indices_train)
# shuffle and take first n_samples %age of test dataset # shuffle and take first n_samples %age of test dataset
shuffle_order_test = np.random.RandomState(seed=1000).permutation(len(test_dataset)) shuffle_order_test = np.random.RandomState(seed=1000).permutation(len(test_dataset))
shuffled_test_dataset = torch.utils.data.Subset(test_dataset, shuffle_order_test) shuffled_test_dataset = torch.utils.data.Subset(test_dataset, shuffle_order_test)
indices_test = torch.arange(int(n_samples*len(test_dataset))) indices_test = torch.arange(int(n_samples*len(test_dataset)))
reduced_test_dataset = data_utils.Subset(shuffled_test_dataset, indices_test) reduced_test_dataset = data_utils.Subset(shuffled_test_dataset, indices_test)
# push into DataLoader # push into DataLoader
train_loader = torch.utils.data.DataLoader(reduced_train_dataset, batch_size=batch_size) train_loader = torch.utils.data.DataLoader(reduced_train_dataset, batch_size=batch_size)
test_loader = torch.utils.data.DataLoader(reduced_test_dataset, batch_size=batch_size) test_loader = torch.utils.data.DataLoader(reduced_test_dataset, batch_size=batch_size)
return train_loader, test_loader return train_loader, test_loader
``` ```
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
def run_baseline(batch_size=32, learning_rate=1e-1, ds="MNIST", toy_size=0.02, max_epochs=100, early_stop_num=10, early_stop_flag=True, average_validation=[15,25], IsLeNet="LeNet"): def run_baseline(batch_size=32, learning_rate=1e-1, ds="MNIST", toy_size=0.02, max_epochs=100, early_stop_num=10, early_stop_flag=True, average_validation=[15,25], IsLeNet="LeNet"):
# create transformations using above info # create transformations using above info
transform = torchvision.transforms.Compose([ transform = torchvision.transforms.Compose([
torchvision.transforms.ToTensor()]) torchvision.transforms.ToTensor()])
# open data and apply these transformations # open data and apply these transformations
if ds == "MNIST": if ds == "MNIST":
train_dataset = datasets.MNIST(root='./MetaAugment/train', train=True, download=True, transform=transform) train_dataset = datasets.MNIST(root='./MetaAugment/train', train=True, download=True, transform=transform)
test_dataset = datasets.MNIST(root='./MetaAugment/test', train=False, download=True, transform=transform) test_dataset = datasets.MNIST(root='./MetaAugment/test', train=False, download=True, transform=transform)
elif ds == "KMNIST": elif ds == "KMNIST":
train_dataset = datasets.KMNIST(root='./MetaAugment/train', train=True, download=True, transform=transform) train_dataset = datasets.KMNIST(root='./MetaAugment/train', train=True, download=True, transform=transform)
test_dataset = datasets.KMNIST(root='./MetaAugment/test', train=False, download=True, transform=transform) test_dataset = datasets.KMNIST(root='./MetaAugment/test', train=False, download=True, transform=transform)
elif ds == "FashionMNIST": elif ds == "FashionMNIST":
train_dataset = datasets.FashionMNIST(root='./MetaAugment/train', train=True, download=True, transform=transform) train_dataset = datasets.FashionMNIST(root='./MetaAugment/train', train=True, download=True, transform=transform)
test_dataset = datasets.FashionMNIST(root='./MetaAugment/test', train=False, download=True, transform=transform) test_dataset = datasets.FashionMNIST(root='./MetaAugment/test', train=False, download=True, transform=transform)
elif ds == "CIFAR10": elif ds == "CIFAR10":
train_dataset = datasets.CIFAR10(root='./MetaAugment/train', train=True, download=True, transform=transform) train_dataset = datasets.CIFAR10(root='./MetaAugment/train', train=True, download=True, transform=transform)
test_dataset = datasets.CIFAR10(root='./MetaAugment/test', train=False, download=True, transform=transform) test_dataset = datasets.CIFAR10(root='./MetaAugment/test', train=False, download=True, transform=transform)
elif ds == "CIFAR100": elif ds == "CIFAR100":
train_dataset = datasets.CIFAR100(root='./MetaAugment/train', train=True, download=True, transform=transform) train_dataset = datasets.CIFAR100(root='./MetaAugment/train', train=True, download=True, transform=transform)
test_dataset = datasets.CIFAR100(root='./MetaAugment/test', train=False, download=True, transform=transform) test_dataset = datasets.CIFAR100(root='./MetaAugment/test', train=False, download=True, transform=transform)
# check sizes of images # check sizes of images
img_height = len(train_dataset[0][0][0]) img_height = len(train_dataset[0][0][0])
img_width = len(train_dataset[0][0][0][0]) img_width = len(train_dataset[0][0][0][0])
img_channels = len(train_dataset[0][0]) img_channels = len(train_dataset[0][0])
# check output labels # check output labels
if ds == "CIFAR10" or ds == "CIFAR100": if ds == "CIFAR10" or ds == "CIFAR100":
num_labels = (max(train_dataset.targets) - min(train_dataset.targets) + 1) num_labels = (max(train_dataset.targets) - min(train_dataset.targets) + 1)
else: else:
num_labels = (max(train_dataset.targets) - min(train_dataset.targets) + 1).item() num_labels = (max(train_dataset.targets) - min(train_dataset.targets) + 1).item()
# create toy dataset from above uploaded data # create toy dataset from above uploaded data
train_loader, test_loader = create_toy(train_dataset, test_dataset, batch_size, toy_size) train_loader, test_loader = create_toy(train_dataset, test_dataset, batch_size, toy_size)
# create model # create model
device = 'cuda' if torch.cuda.is_available() else 'cpu' device = 'cuda' if torch.cuda.is_available() else 'cpu'
if IsLeNet == "LeNet": if IsLeNet == "LeNet":
model = LeNet(img_height, img_width, num_labels, img_channels).to(device) # added .to(device) model = LeNet(img_height, img_width, num_labels, img_channels).to(device) # added .to(device)
elif IsLeNet == "EasyNet": elif IsLeNet == "EasyNet":
model = EasyNet(img_height, img_width, num_labels, img_channels).to(device) # added .to(device) model = EasyNet(img_height, img_width, num_labels, img_channels).to(device) # added .to(device)
else: else:
model = SimpleNet(img_height, img_width, num_labels, img_channels).to(device) # added .to(device) model = SimpleNet(img_height, img_width, num_labels, img_channels).to(device) # added .to(device)
sgd = optim.SGD(model.parameters(), lr=learning_rate) sgd = optim.SGD(model.parameters(), lr=learning_rate)
cost = nn.CrossEntropyLoss() cost = nn.CrossEntropyLoss()
# set variables for best validation accuracy and early stop count # set variables for best validation accuracy and early stop count
best_acc = 0 best_acc = 0
early_stop_cnt = 0 early_stop_cnt = 0
total_val = 0 total_val = 0
# train model and check validation accuracy each epoch # train model and check validation accuracy each epoch
for _epoch in range(max_epochs): for _epoch in range(max_epochs):
# train model # train model
model.train() model.train()
for idx, (train_x, train_label) in enumerate(train_loader): for idx, (train_x, train_label) in enumerate(train_loader):
train_x, train_label = train_x.to(device), train_label.to(device) train_x, train_label = train_x.to(device), train_label.to(device)
label_np = np.zeros((train_label.shape[0], num_labels)) label_np = np.zeros((train_label.shape[0], num_labels))
sgd.zero_grad() sgd.zero_grad()
predict_y = model(train_x.float()) predict_y = model(train_x.float())
loss = cost(predict_y, train_label.long()) loss = cost(predict_y, train_label.long())
loss.backward() loss.backward()
sgd.step() sgd.step()
# check validation accuracy on validation set # check validation accuracy on validation set
correct = 0 correct = 0
_sum = 0 _sum = 0
model.eval() model.eval()
for idx, (test_x, test_label) in enumerate(test_loader): for idx, (test_x, test_label) in enumerate(test_loader):
test_x, test_label = test_x.to(device), test_label.to(device) # new code test_x, test_label = test_x.to(device), test_label.to(device) # new code
predict_y = model(test_x.float()).detach() predict_y = model(test_x.float()).detach()
#predict_ys = np.argmax(predict_y, axis=-1) #predict_ys = np.argmax(predict_y, axis=-1)
predict_ys = torch.argmax(predict_y, axis=-1) # changed np to torch predict_ys = torch.argmax(predict_y, axis=-1) # changed np to torch
#label_np = test_label.numpy() #label_np = test_label.numpy()
_ = predict_ys == test_label _ = predict_ys == test_label
#correct += np.sum(_.numpy(), axis=-1) #correct += np.sum(_.numpy(), axis=-1)
correct += np.sum(_.cpu().numpy(), axis=-1) # added .cpu() correct += np.sum(_.cpu().numpy(), axis=-1) # added .cpu()
_sum += _.shape[0] _sum += _.shape[0]
acc = correct / _sum acc = correct / _sum
# update the total validation # update the total validation
if average_validation[0] <= _epoch <= average_validation[1]: if average_validation[0] <= _epoch <= average_validation[1]:
total_val += acc total_val += acc
# update best validation accuracy if it was higher, otherwise increase early stop count # update best validation accuracy if it was higher, otherwise increase early stop count
if acc > best_acc: if acc > best_acc:
best_acc = acc best_acc = acc
early_stop_cnt = 0 early_stop_cnt = 0
else: else:
early_stop_cnt += 1 early_stop_cnt += 1
# exit if validation gets worse over 10 runs and using early stopping # exit if validation gets worse over 10 runs and using early stopping
if early_stop_cnt >= early_stop_num and early_stop_flag: if early_stop_cnt >= early_stop_num and early_stop_flag:
return best_acc return best_acc
# exit if using fixed epoch length # exit if using fixed epoch length
if _epoch >= average_validation[1] and not early_stop_flag: if _epoch >= average_validation[1] and not early_stop_flag:
return total_val / (average_validation[1] - average_validation[0] + 1) return total_val / (average_validation[1] - average_validation[0] + 1)
``` ```
%% Cell type:code id: tags: %% Cell type:code id: tags:
``` python ``` python
%%time %%time
batch_size = 32 # size of batch the inner NN is trained with batch_size = 32 # size of batch the inner NN is trained with
learning_rate = 1e-1 # fix learning rate learning_rate = 1e-1 # fix learning rate
ds = "FashionMNIST" # pick dataset (MNIST, KMNIST, FashionMNIST, CIFAR10,...) ds = "FashionMNIST" # pick dataset (MNIST, KMNIST, FashionMNIST, CIFAR10,...)
toy_size = 1 # total propeortion of training and test set we use toy_size = 1 # total propeortion of training and test set we use
max_epochs = 100 # max number of epochs that is run if early stopping is not hit max_epochs = 100 # max number of epochs that is run if early stopping is not hit
early_stop_num = 10 # max number of worse validation scores before early stopping is triggered early_stop_num = 10 # max number of worse validation scores before early stopping is triggered
early_stop_flag = True # implement early stopping or not early_stop_flag = True # implement early stopping or not
average_validation = [15,25] # if not implementing early stopping, what epochs are we averaging over average_validation = [15,25] # if not implementing early stopping, what epochs are we averaging over
num_iterations = 10 # how many iterations are we averaging over num_iterations = 10 # how many iterations are we averaging over
IsLeNet = "SimpleNet" # using LeNet or EasyNet or SimpleNet IsLeNet = "SimpleNet" # using LeNet or EasyNet or SimpleNet
# run using early stopping # run using early stopping
best_accuracies_1 = [] best_accuracies_1 = []
for baselines in trange(num_iterations): for baselines in trange(num_iterations):
best_acc = run_baseline(batch_size, learning_rate, ds, toy_size, max_epochs, early_stop_num, early_stop_flag, average_validation, IsLeNet) best_acc = run_baseline(batch_size, learning_rate, ds, toy_size, max_epochs, early_stop_num, early_stop_flag, average_validation, IsLeNet)
best_accuracies_1.append(best_acc) best_accuracies_1.append(best_acc)
if baselines % 10 == 0: if baselines % 10 == 0:
print("{}\tBest accuracy: {:.2f}%".format(baselines, best_acc*100)) print("{}\tBest accuracy: {:.2f}%".format(baselines, best_acc*100))
print("Average best accuracy: {:.2f}%\n".format(np.mean(best_accuracies_1)*100)) print("Average best accuracy: {:.2f}%\n".format(np.mean(best_accuracies_1)*100))
file = open(f"{ds}_v1.txt", "w") file = open(f"{ds}_v1.txt", "w")
content = ','.join(str(e) for e in best_accuracies_1) content = ','.join(str(e) for e in best_accuracies_1)
file.write(content) file.write(content)
file.close() file.close()
# run using average validation losses # run using average validation losses
early_stop_flag = False early_stop_flag = False
best_accuracies_2 = [] best_accuracies_2 = []
for baselines in trange(num_iterations): for baselines in trange(num_iterations):
best_acc = run_baseline(batch_size, learning_rate, ds, toy_size, max_epochs, early_stop_num, early_stop_flag, average_validation, IsLeNet) best_acc = run_baseline(batch_size, learning_rate, ds, toy_size, max_epochs, early_stop_num, early_stop_flag, average_validation, IsLeNet)
best_accuracies_2.append(best_acc) best_accuracies_2.append(best_acc)
if baselines % 10 == 0: if baselines % 10 == 0:
print("{}\tAverage accuracy: {:.2f}%".format(baselines, best_acc*100)) print("{}\tBest accuracy: {:.2f}%".format(baselines, best_acc*100))
print("Average average accuracy: {:.2f}%\n".format(np.mean(best_accuracies_2)*100)) print("Average average accuracy: {:.2f}%\n".format(np.mean(best_accuracies_2)*100))
file = open(f"{ds}_v2.txt", "w") file = open(f"{ds}_v2.txt", "w")
content = ','.join(str(e) for e in best_accuracies_2) content = ','.join(str(e) for e in best_accuracies_2)
file.write(content) file.write(content)
file.close() file.close()
``` ```
%% Output %% Output
0%| | 0/10 [00:00<?, ?it/s] 0%| | 0/10 [00:00<?, ?it/s]
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw/train-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw/train-images-idx3-ubyte.gz
Extracting ./MetaAugment/train/MNIST/raw/train-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw Extracting ./MetaAugment/train/MNIST/raw/train-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw/train-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw/train-labels-idx1-ubyte.gz
Extracting ./MetaAugment/train/MNIST/raw/train-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw Extracting ./MetaAugment/train/MNIST/raw/train-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw/t10k-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw/t10k-images-idx3-ubyte.gz
Extracting ./MetaAugment/train/MNIST/raw/t10k-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw Extracting ./MetaAugment/train/MNIST/raw/t10k-images-idx3-ubyte.gz to ./MetaAugment/train/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw/t10k-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw/t10k-labels-idx1-ubyte.gz
Extracting ./MetaAugment/train/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw Extracting ./MetaAugment/train/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./MetaAugment/train/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw/train-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw/train-images-idx3-ubyte.gz
Extracting ./MetaAugment/test/MNIST/raw/train-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw Extracting ./MetaAugment/test/MNIST/raw/train-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw/train-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw/train-labels-idx1-ubyte.gz
Extracting ./MetaAugment/test/MNIST/raw/train-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw Extracting ./MetaAugment/test/MNIST/raw/train-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw/t10k-images-idx3-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw/t10k-images-idx3-ubyte.gz
Extracting ./MetaAugment/test/MNIST/raw/t10k-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw Extracting ./MetaAugment/test/MNIST/raw/t10k-images-idx3-ubyte.gz to ./MetaAugment/test/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw/t10k-labels-idx1-ubyte.gz Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw/t10k-labels-idx1-ubyte.gz
Extracting ./MetaAugment/test/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw Extracting ./MetaAugment/test/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./MetaAugment/test/MNIST/raw
10%|█ | 1/10 [00:17<02:34, 17.17s/it] 10%|█ | 1/10 [00:17<02:34, 17.17s/it]
0 Best accuracy: 62.00% 0 Best accuracy: 62.00%
100%|██████████| 10/10 [01:13<00:00, 7.40s/it] 100%|██████████| 10/10 [01:13<00:00, 7.40s/it]
Average best accuracy: 78.95% Average best accuracy: 78.95%
10%|█ | 1/10 [00:07<01:03, 7.02s/it] 10%|█ | 1/10 [00:07<01:03, 7.02s/it]
0 Average accuracy: 84.05% 0 Average accuracy: 84.05%
100%|██████████| 10/10 [01:09<00:00, 6.95s/it] 100%|██████████| 10/10 [01:09<00:00, 6.95s/it]
Average average accuracy: 81.36% Average average accuracy: 81.36%
CPU times: user 2min 12s, sys: 3.91 s, total: 2min 16s CPU times: user 2min 12s, sys: 3.91 s, total: 2min 16s
Wall time: 2min 23s Wall time: 2min 23s
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment