137 KiB
137 KiB
import os
if not os.path.exists('GTSRB'):
!pip install -U -q torch_snippets
!wget -qq https://sid.erda.dk/public/archives/daaeac0d7ce1152aea9b61d9f1e19370/GTSRB_Final_Training_Images.zip
!wget -qq https://sid.erda.dk/public/archives/daaeac0d7ce1152aea9b61d9f1e19370/GTSRB_Final_Test_Images.zip
!unzip -qq GTSRB_Final_Training_Images.zip
!unzip -qq GTSRB_Final_Test_Images.zip
!wget https://raw.githubusercontent.com/georgesung/traffic_sign_classification_german/master/signnames.csv
!rm GTSRB_Final_Training_Images.zip GTSRB_Final_Test_Images.zip
[K |████████████████████████████████| 61kB 2.2MB/s [K |████████████████████████████████| 36.6MB 86kB/s [K |████████████████████████████████| 102kB 14.0MB/s [?25h Building wheel for contextvars (setup.py) ... [?25l[?25hdone --2020-09-17 11:22:17-- https://raw.githubusercontent.com/georgesung/traffic_sign_classification_german/master/signnames.csv Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ... Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 999 [text/plain] Saving to: ‘signnames.csv’ signnames.csv 100%[===================>] 999 --.-KB/s in 0s 2020-09-17 11:22:18 (72.4 MB/s) - ‘signnames.csv’ saved [999/999]
from torch_snippets import *
classIds = pd.read_csv('signnames.csv')
classIds.set_index('ClassId', inplace=True)
classIds = classIds.to_dict()['SignName']
classIds = {f'{k:05d}':v for k,v in classIds.items()}
id2int = {v:ix for ix,(k,v) in enumerate(classIds.items())}
from torchvision import transforms as T
classIds = pd.read_csv('signnames.csv')
classIds.set_index('ClassId', inplace=True)
classIds = classIds.to_dict()['SignName']
classIds = {f'{k:05d}':v for k,v in classIds.items()}
id2int = {v:ix for ix,(k,v) in enumerate(classIds.items())}
from torchvision import transforms as T
trn_tfms = T.Compose([
T.ToPILImage(),
T.Resize(32),
T.CenterCrop(32),
# T.ColorJitter(brightness=(0.8,1.2),
# contrast=(0.8,1.2),
# saturation=(0.8,1.2),
# hue=0.25),
# T.RandomAffine(5, translate=(0.01,0.1)),
T.ToTensor(),
T.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
val_tfms = T.Compose([
T.ToPILImage(),
T.Resize(32),
T.CenterCrop(32),
T.ToTensor(),
T.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
class GTSRB(Dataset):
"""Face Landmarks dataset."""
def __init__(self, files, transform=None):
self.files = files
self.transform = transform
logger.info(len(self))
def __len__(self):
return len(self.files)
def __getitem__(self, ix):
fpath = self.files[ix]
clss = fname(parent(fpath))
img = read(fpath, 1)
return img, classIds[clss]
def choose(self):
return self[randint(len(self))]
def collate_fn(self, batch):
imgs, classes = list(zip(*batch))
if self.transform:
imgs = [self.transform(img)[None] for img in imgs]
classes = [torch.tensor([id2int[clss]]) for clss in classes]
imgs, classes = [torch.cat(i).to(device) for i in [imgs, classes]]
return imgs, classes
device = 'cuda' if torch.cuda.is_available() else 'cpu'
all_files = Glob('GTSRB/Final_Training/Images/*/*.ppm')
np.random.seed(10)
np.random.shuffle(all_files)
from sklearn.model_selection import train_test_split
trn_files, val_files = train_test_split(all_files, random_state=1)
trn_ds = GTSRB(trn_files, transform=trn_tfms)
val_ds = GTSRB(val_files, transform=val_tfms)
trn_dl = DataLoader(trn_ds, 32, shuffle=True, collate_fn=trn_ds.collate_fn)
val_dl = DataLoader(val_ds, 32, shuffle=False, collate_fn=val_ds.collate_fn)
import torchvision.models as models
def convBlock(ni, no):
return nn.Sequential(
nn.Dropout(0.2),
nn.Conv2d(ni, no, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
#nn.BatchNorm2d(no),
nn.MaxPool2d(2),
)
class SignClassifier(nn.Module):
def __init__(self):
super().__init__()
self.model = nn.Sequential(
convBlock(3, 64),
convBlock(64, 64),
convBlock(64, 128),
convBlock(128, 64),
nn.Flatten(),
nn.Linear(256, 256),
nn.Dropout(0.2),
nn.ReLU(inplace=True),
nn.Linear(256, len(id2int))
)
self.loss_fn = nn.CrossEntropyLoss()
def forward(self, x):
return self.model(x)
def compute_metrics(self, preds, targets):
ce_loss = self.loss_fn(preds, targets)
acc = (torch.max(preds, 1)[1] == targets).float().mean()
return ce_loss, acc
def train_batch(model, data, optimizer, criterion):
ims, labels = data
_preds = model(ims)
optimizer.zero_grad()
loss, acc = criterion(_preds, labels)
loss.backward()
optimizer.step()
return loss.item(), acc.item()
@torch.no_grad()
def validate_batch(model, data, criterion):
ims, labels = data
_preds = model(ims)
loss, acc = criterion(_preds, labels)
return loss.item(), acc.item()
model = SignClassifier().to(device)
criterion = model.compute_metrics
optimizer = optim.Adam(model.parameters(), lr=1e-3)
n_epochs = 40
log = Report(n_epochs)
for ex in range(n_epochs):
N = len(trn_dl)
for bx, data in enumerate(trn_dl):
loss, acc = train_batch(model, data, optimizer, criterion)
log.record(ex+(bx+1)/N, trn_loss=loss, trn_acc=acc, end='\r')
N = len(val_dl)
for bx, data in enumerate(val_dl):
loss, acc = validate_batch(model, data, criterion)
log.record(ex+(bx+1)/N, val_loss=loss, val_acc=acc, end='\r')
log.report_avgs(ex+1)
if ex == 10: optimizer = optim.Adam(model.parameters(), lr=1e-4)
log.plot_epochs()
dumpdill(log, 'no-aug-no-bn.log')2020-09-17 11:22:22.120 | INFO | torch_snippets.loader:Glob:172 - 39209 files found at GTSRB/Final_Training/Images/*/*.ppm 2020-09-17 11:22:22.509 | INFO | __main__:__init__:46 - 29406 2020-09-17 11:22:22.510 | INFO | __main__:__init__:46 - 9803
EPOCH: 1.000 trn_loss: 2.410 trn_acc: 0.290 val_loss: 1.641 val_acc: 0.475 (17.64s - 688.04s remaining) EPOCH: 2.000 trn_loss: 1.214 trn_acc: 0.599 val_loss: 0.914 val_acc: 0.695 (34.74s - 660.13s remaining) EPOCH: 2.610 trn_loss: 0.927 trn_acc: 0.625 (42.70s - 611.66s remaining)
from torch_snippets import *
classIds = pd.read_csv('signnames.csv')
classIds.set_index('ClassId', inplace=True)
classIds = classIds.to_dict()['SignName']
classIds = {f'{k:05d}':v for k,v in classIds.items()}
id2int = {v:ix for ix,(k,v) in enumerate(classIds.items())}
from torchvision import transforms as T
classIds = pd.read_csv('signnames.csv')
classIds.set_index('ClassId', inplace=True)
classIds = classIds.to_dict()['SignName']
classIds = {f'{k:05d}':v for k,v in classIds.items()}
id2int = {v:ix for ix,(k,v) in enumerate(classIds.items())}
from torchvision import transforms as T
trn_tfms = T.Compose([
T.ToPILImage(),
T.Resize(32),
T.CenterCrop(32),
# T.ColorJitter(brightness=(0.8,1.2),
# contrast=(0.8,1.2),
# saturation=(0.8,1.2),
# hue=0.25),
# T.RandomAffine(5, translate=(0.01,0.1)),
T.ToTensor(),
T.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
val_tfms = T.Compose([
T.ToPILImage(),
T.Resize(32),
T.CenterCrop(32),
T.ToTensor(),
T.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
class GTSRB(Dataset):
"""Face Landmarks dataset."""
def __init__(self, files, transform=None):
self.files = files
self.transform = transform
logger.info(len(self))
def __len__(self):
return len(self.files)
def __getitem__(self, ix):
fpath = self.files[ix]
clss = fname(parent(fpath))
img = read(fpath, 1)
return img, classIds[clss]
def choose(self):
return self[randint(len(self))]
def collate_fn(self, batch):
imgs, classes = list(zip(*batch))
if self.transform:
imgs = [self.transform(img)[None] for img in imgs]
classes = [torch.tensor([id2int[clss]]) for clss in classes]
imgs, classes = [torch.cat(i).to(device) for i in [imgs, classes]]
return imgs, classes
device = 'cuda' if torch.cuda.is_available() else 'cpu'
all_files = Glob('GTSRB/Final_Training/Images/*/*.ppm')
np.random.seed(10)
np.random.shuffle(all_files)
from sklearn.model_selection import train_test_split
trn_files, val_files = train_test_split(all_files, random_state=1)
trn_ds = GTSRB(trn_files, transform=trn_tfms)
val_ds = GTSRB(val_files, transform=val_tfms)
trn_dl = DataLoader(trn_ds, 32, shuffle=True, collate_fn=trn_ds.collate_fn)
val_dl = DataLoader(val_ds, 32, shuffle=False, collate_fn=val_ds.collate_fn)
import torchvision.models as models
def convBlock(ni, no):
return nn.Sequential(
nn.Dropout(0.2),
nn.Conv2d(ni, no, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.BatchNorm2d(no),
nn.MaxPool2d(2),
)
class SignClassifier(nn.Module):
def __init__(self):
super().__init__()
self.model = nn.Sequential(
convBlock(3, 64),
convBlock(64, 64),
convBlock(64, 128),
convBlock(128, 64),
nn.Flatten(),
nn.Linear(256, 256),
nn.Dropout(0.2),
nn.ReLU(inplace=True),
nn.Linear(256, len(id2int))
)
self.loss_fn = nn.CrossEntropyLoss()
def forward(self, x):
return self.model(x)
def compute_metrics(self, preds, targets):
ce_loss = self.loss_fn(preds, targets)
acc = (torch.max(preds, 1)[1] == targets).float().mean()
return ce_loss, acc
def train_batch(model, data, optimizer, criterion):
ims, labels = data
_preds = model(ims)
optimizer.zero_grad()
loss, acc = criterion(_preds, labels)
loss.backward()
optimizer.step()
return loss.item(), acc.item()
@torch.no_grad()
def validate_batch(model, data, criterion):
ims, labels = data
_preds = model(ims)
loss, acc = criterion(_preds, labels)
return loss.item(), acc.item()
model = SignClassifier().to(device)
criterion = model.compute_metrics
optimizer = optim.Adam(model.parameters(), lr=1e-3)
n_epochs = 40
log = Report(n_epochs)
for ex in range(n_epochs):
N = len(trn_dl)
for bx, data in enumerate(trn_dl):
loss, acc = train_batch(model, data, optimizer, criterion)
log.record(ex+(bx+1)/N, trn_loss=loss, trn_acc=acc, end='\r')
N = len(val_dl)
for bx, data in enumerate(val_dl):
loss, acc = validate_batch(model, data, criterion)
log.record(ex+(bx+1)/N, val_loss=loss, val_acc=acc, end='\r')
log.report_avgs(ex+1)
if ex == 10: optimizer = optim.Adam(model.parameters(), lr=1e-4)
log.plot_epochs()
dumpdill(log, 'no-aug-yes-bn.log')from torch_snippets import *
classIds = pd.read_csv('signnames.csv')
classIds.set_index('ClassId', inplace=True)
classIds = classIds.to_dict()['SignName']
classIds = {f'{k:05d}':v for k,v in classIds.items()}
id2int = {v:ix for ix,(k,v) in enumerate(classIds.items())}
from torchvision import transforms as T
classIds = pd.read_csv('signnames.csv')
classIds.set_index('ClassId', inplace=True)
classIds = classIds.to_dict()['SignName']
classIds = {f'{k:05d}':v for k,v in classIds.items()}
id2int = {v:ix for ix,(k,v) in enumerate(classIds.items())}
from torchvision import transforms as T
trn_tfms = T.Compose([
T.ToPILImage(),
T.Resize(32),
T.CenterCrop(32),
T.ColorJitter(brightness=(0.8,1.2),
contrast=(0.8,1.2),
saturation=(0.8,1.2),
hue=0.25),
T.RandomAffine(5, translate=(0.01,0.1)),
T.ToTensor(),
T.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
val_tfms = T.Compose([
T.ToPILImage(),
T.Resize(32),
T.CenterCrop(32),
T.ToTensor(),
T.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
class GTSRB(Dataset):
"""Face Landmarks dataset."""
def __init__(self, files, transform=None):
self.files = files
self.transform = transform
logger.info(len(self))
def __len__(self):
return len(self.files)
def __getitem__(self, ix):
fpath = self.files[ix]
clss = fname(parent(fpath))
img = read(fpath, 1)
return img, classIds[clss]
def choose(self):
return self[randint(len(self))]
def collate_fn(self, batch):
imgs, classes = list(zip(*batch))
if self.transform:
imgs = [self.transform(img)[None] for img in imgs]
classes = [torch.tensor([id2int[clss]]) for clss in classes]
imgs, classes = [torch.cat(i).to(device) for i in [imgs, classes]]
return imgs, classes
device = 'cuda' if torch.cuda.is_available() else 'cpu'
all_files = Glob('GTSRB/Final_Training/Images/*/*.ppm')
np.random.seed(10)
np.random.shuffle(all_files)
from sklearn.model_selection import train_test_split
trn_files, val_files = train_test_split(all_files, random_state=1)
trn_ds = GTSRB(trn_files, transform=trn_tfms)
val_ds = GTSRB(val_files, transform=val_tfms)
trn_dl = DataLoader(trn_ds, 32, shuffle=True, collate_fn=trn_ds.collate_fn)
val_dl = DataLoader(val_ds, 32, shuffle=False, collate_fn=val_ds.collate_fn)
import torchvision.models as models
def convBlock(ni, no):
return nn.Sequential(
nn.Dropout(0.2),
nn.Conv2d(ni, no, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.BatchNorm2d(no),
nn.MaxPool2d(2),
)
class SignClassifier(nn.Module):
def __init__(self):
super().__init__()
self.model = nn.Sequential(
convBlock(3, 64),
convBlock(64, 64),
convBlock(64, 128),
convBlock(128, 64),
nn.Flatten(),
nn.Linear(256, 256),
nn.Dropout(0.2),
nn.ReLU(inplace=True),
nn.Linear(256, len(id2int))
)
self.loss_fn = nn.CrossEntropyLoss()
def forward(self, x):
return self.model(x)
def compute_metrics(self, preds, targets):
ce_loss = self.loss_fn(preds, targets)
acc = (torch.max(preds, 1)[1] == targets).float().mean()
return ce_loss, acc
def train_batch(model, data, optimizer, criterion):
ims, labels = data
_preds = model(ims)
optimizer.zero_grad()
loss, acc = criterion(_preds, labels)
loss.backward()
optimizer.step()
return loss.item(), acc.item()
@torch.no_grad()
def validate_batch(model, data, criterion):
ims, labels = data
_preds = model(ims)
loss, acc = criterion(_preds, labels)
return loss.item(), acc.item()
model = SignClassifier().to(device)
criterion = model.compute_metrics
optimizer = optim.Adam(model.parameters(), lr=1e-3)
n_epochs = 40
log = Report(n_epochs)
for ex in range(n_epochs):
N = len(trn_dl)
for bx, data in enumerate(trn_dl):
loss, acc = train_batch(model, data, optimizer, criterion)
log.record(ex+(bx+1)/N, trn_loss=loss, trn_acc=acc, end='\r')
N = len(val_dl)
for bx, data in enumerate(val_dl):
loss, acc = validate_batch(model, data, criterion)
log.record(ex+(bx+1)/N, val_loss=loss, val_acc=acc, end='\r')
log.report_avgs(ex+1)
if ex == 10: optimizer = optim.Adam(model.parameters(), lr=1e-4)
log.plot_epochs()
dumpdill(log, '40-yes-aug-yes-bn.log')
from google.colab import files
files.download('40-yes-aug-yes-bn.log')2020-09-17 09:54:30.795 | INFO | torch_snippets.loader:Glob:172 - 39209 files found at GTSRB/Final_Training/Images/*/*.ppm 2020-09-17 09:54:30.808 | INFO | __main__:__init__:46 - 29406 2020-09-17 09:54:30.809 | INFO | __main__:__init__:46 - 9803
EPOCH: 1.000 trn_loss: 1.840 trn_acc: 0.462 val_loss: 0.752 val_acc: 0.762 (62.07s - 2420.80s remaining) EPOCH: 2.000 trn_loss: 0.679 trn_acc: 0.780 val_loss: 0.432 val_acc: 0.858 (125.57s - 2385.74s remaining) EPOCH: 3.000 trn_loss: 0.480 trn_acc: 0.845 val_loss: 0.369 val_acc: 0.878 (189.63s - 2338.78s remaining) EPOCH: 4.000 trn_loss: 0.397 trn_acc: 0.871 val_loss: 0.299 val_acc: 0.902 (253.21s - 2278.87s remaining) EPOCH: 5.000 trn_loss: 0.341 trn_acc: 0.889 val_loss: 0.267 val_acc: 0.911 (316.45s - 2215.16s remaining) EPOCH: 6.000 trn_loss: 0.297 trn_acc: 0.905 val_loss: 0.246 val_acc: 0.920 (379.64s - 2151.27s remaining) EPOCH: 7.000 trn_loss: 0.269 trn_acc: 0.910 val_loss: 0.218 val_acc: 0.930 (442.87s - 2087.82s remaining) EPOCH: 8.000 trn_loss: 0.250 trn_acc: 0.918 val_loss: 0.198 val_acc: 0.934 (506.33s - 2025.32s remaining) EPOCH: 9.000 trn_loss: 0.230 trn_acc: 0.923 val_loss: 0.198 val_acc: 0.938 (569.49s - 1961.58s remaining) EPOCH: 10.000 trn_loss: 0.211 trn_acc: 0.930 val_loss: 0.173 val_acc: 0.943 (632.87s - 1898.60s remaining) EPOCH: 11.000 trn_loss: 0.202 trn_acc: 0.933 val_loss: 0.186 val_acc: 0.943 (696.65s - 1836.64s remaining) EPOCH: 12.000 trn_loss: 0.145 trn_acc: 0.952 val_loss: 0.130 val_acc: 0.958 (759.59s - 1772.37s remaining) EPOCH: 13.000 trn_loss: 0.130 trn_acc: 0.956 val_loss: 0.116 val_acc: 0.963 (822.65s - 1708.59s remaining) EPOCH: 14.000 trn_loss: 0.119 trn_acc: 0.959 val_loss: 0.113 val_acc: 0.962 (885.85s - 1645.14s remaining) EPOCH: 15.000 trn_loss: 0.113 trn_acc: 0.963 val_loss: 0.115 val_acc: 0.965 (948.54s - 1580.91s remaining) EPOCH: 16.000 trn_loss: 0.109 trn_acc: 0.963 val_loss: 0.106 val_acc: 0.964 (1011.32s - 1516.98s remaining) EPOCH: 17.000 trn_loss: 0.104 trn_acc: 0.964 val_loss: 0.105 val_acc: 0.966 (1074.53s - 1453.78s remaining) EPOCH: 18.000 trn_loss: 0.097 trn_acc: 0.967 val_loss: 0.098 val_acc: 0.968 (1138.27s - 1391.22s remaining) EPOCH: 19.000 trn_loss: 0.100 trn_acc: 0.967 val_loss: 0.106 val_acc: 0.966 (1201.71s - 1328.20s remaining) EPOCH: 20.000 trn_loss: 0.098 trn_acc: 0.966 val_loss: 0.095 val_acc: 0.969 (1265.57s - 1265.57s remaining) EPOCH: 21.000 trn_loss: 0.095 trn_acc: 0.968 val_loss: 0.098 val_acc: 0.969 (1329.09s - 1202.51s remaining) EPOCH: 22.000 trn_loss: 0.094 trn_acc: 0.967 val_loss: 0.093 val_acc: 0.972 (1393.45s - 1140.09s remaining) EPOCH: 23.000 trn_loss: 0.090 trn_acc: 0.970 val_loss: 0.096 val_acc: 0.971 (1457.31s - 1077.14s remaining) EPOCH: 24.000 trn_loss: 0.085 trn_acc: 0.971 val_loss: 0.092 val_acc: 0.972 (1521.21s - 1014.14s remaining) EPOCH: 25.000 trn_loss: 0.089 trn_acc: 0.970 val_loss: 0.096 val_acc: 0.968 (1585.12s - 951.07s remaining) EPOCH: 26.000 trn_loss: 0.085 trn_acc: 0.971 val_loss: 0.090 val_acc: 0.972 (1649.27s - 888.07s remaining) EPOCH: 27.000 trn_loss: 0.080 trn_acc: 0.972 val_loss: 0.083 val_acc: 0.974 (1713.44s - 824.99s remaining) EPOCH: 28.000 trn_loss: 0.080 trn_acc: 0.973 val_loss: 0.095 val_acc: 0.972 (1777.52s - 761.79s remaining) EPOCH: 29.000 trn_loss: 0.080 trn_acc: 0.972 val_loss: 0.081 val_acc: 0.975 (1841.47s - 698.49s remaining) EPOCH: 30.000 trn_loss: 0.081 trn_acc: 0.972 val_loss: 0.088 val_acc: 0.974 (1905.06s - 635.02s remaining) EPOCH: 31.000 trn_loss: 0.075 trn_acc: 0.975 val_loss: 0.084 val_acc: 0.972 (1968.86s - 571.60s remaining) EPOCH: 32.000 trn_loss: 0.081 trn_acc: 0.973 val_loss: 0.089 val_acc: 0.972 (2032.45s - 508.11s remaining) EPOCH: 33.000 trn_loss: 0.076 trn_acc: 0.974 val_loss: 0.089 val_acc: 0.971 (2095.91s - 444.59s remaining) EPOCH: 34.000 trn_loss: 0.072 trn_acc: 0.975 val_loss: 0.083 val_acc: 0.975 (2159.69s - 381.12s remaining) EPOCH: 35.000 trn_loss: 0.072 trn_acc: 0.975 val_loss: 0.087 val_acc: 0.974 (2223.42s - 317.63s remaining) EPOCH: 36.000 trn_loss: 0.070 trn_acc: 0.976 val_loss: 0.079 val_acc: 0.977 (2287.19s - 254.13s remaining) EPOCH: 37.000 trn_loss: 0.071 trn_acc: 0.975 val_loss: 0.081 val_acc: 0.975 (2351.40s - 190.65s remaining) EPOCH: 38.000 trn_loss: 0.071 trn_acc: 0.976 val_loss: 0.074 val_acc: 0.975 (2415.64s - 127.14s remaining) EPOCH: 38.106 trn_loss: 0.100 trn_acc: 0.938 (2421.79s - 120.40s remaining)
for f in Glob('*.log'):
log = loaddill(f)
print()
log.report_avgs(20)
log.plot_epochs(['trn_acc', 'val_acc'], title=f.replace(',','\n').replace('.log',''))
print()
line()2020-09-17 09:35:05.018 | INFO | torch_snippets.loader:Glob:172 - 4 files found at *.log 0%| | 0/20 [00:00<?, ?it/s]/usr/local/lib/python3.6/dist-packages/numpy/core/fromnumeric.py:3335: RuntimeWarning: Mean of empty slice. out=out, **kwargs) /usr/local/lib/python3.6/dist-packages/numpy/core/_methods.py:161: RuntimeWarning: invalid value encountered in double_scalars ret = ret.dtype.type(ret / rcount) 100%|██████████| 20/20 [00:00<00:00, 351.47it/s]
EPOCH: 20.000 trn_loss: 0.031 trn_acc: 0.990 val_loss: 0.077 val_acc: 0.978 (8455.42s - 0.00s remaining)
100%|██████████| 20/20 [00:00<00:00, 358.36it/s]
================================================================== EPOCH: 20.000 trn_loss: 0.391 trn_acc: 0.869 val_loss: 0.323 val_acc: 0.889 (9196.88s - 0.00s remaining)
100%|██████████| 20/20 [00:00<00:00, 372.78it/s]
================================================================== EPOCH: 20.000 trn_loss: 0.094 trn_acc: 0.969 val_loss: 0.093 val_acc: 0.969 (8106.35s - 0.00s remaining)
100%|██████████| 20/20 [00:00<00:00, 330.28it/s]
================================================================== EPOCH: 20.000 trn_loss: 0.146 trn_acc: 0.952 val_loss: 0.202 val_acc: 0.938 (7348.78s - 0.00s remaining)
==================================================================