import tensorflow.compat.v1 as tf
import lenet
import utils.read_mnist as read_mnist
import time
import os
import numpy as np
train_x, train_y, test_x, test_y, index2class = read_mnist.read_fashion_mnist(one_hot=True, standard=True)
index_train = np.random.permutation(train_x.shape[0])
train_x, train_y = train_x[index_train], train_y[index_train]
lr_rate = 0.01
batch_size = 64
sess = tf.Session()
model = lenet.LeNet(lr_rate=0.001, regular=0.0005, train=True)
sess.run(tf.global_variables_initializer())
tensorboard_dir = r"fashiontensorboardlog/"
if os.path.exists(tensorboard_dir):
for file in os.listdir(tensorboard_dir):
path_file = os.path.join(tensorboard_dir, file)
os.remove(path_file)
file_writer = tf.summary.FileWriter(tensorboard_dir, sess.graph)
tf.summary.scalar("loss of test data", model.loss)
tf.summary.scalar("accuracy on test data", model.accuracy)
tf.summary.scalar('learning rate ', tf.reduce_mean(model.lr_rate))
merge = tf.summary.merge_all()
print(tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES))
epochs = 100
saver = tf.train.Saver()
for epoch in range(epochs):
# In[ ]:
if args.dataset in ['MNIST', 'CIFAR10']:
n_channels = {
'MNIST': 1,
'CIFAR10': 3,
}[args.dataset]
size = {
'MNIST': 28,
'CIFAR10': 32,
}[args.dataset]
actor = {
'linear': nn.Linear(n_channels * size**2, n_classes),
'lenet': lenet.LeNet(3, n_classes, size),
'resnet': resnet.ResNet(depth=18, n_classes=n_classes),
}[args.actor]
elif args.dataset in ['covtype']:
n_features = train_x.size(1)
actor = {
'linear': nn.Linear(n_features, n_classes),
'mlp': mlp.MLP([n_features, 60, 60, 80, n_classes], th.tanh)
}[args.actor]
if args.w > 0:
assert n_classes == 2
w = th.tensor([1 -
args.w, args.w]) if args.w else th.full(n_classes, 1.0 /
n_classes)
cross_entropy = loss.CrossEntropyLoss(w)
"""
Created on 2017.6.11
@author: tfygg
"""
import lenet
import lenetseq
import utils
# Net
net = lenet.LeNet()
print(net)
for index, param in enumerate(net.parameters()):
print(list(param.data))
print(type(param.data), param.size())
print index, "-->", param
print(net.state_dict())
print(net.state_dict().keys())
for key in net.state_dict():
print key, 'corresponds to', list(net.state_dict()[key])
#NetSeq
netSeq = lenetseq.LeNetSeq()
print(netSeq)
utils.initNetParams(netSeq)
def main(args):
logfilename = 'convergence_mnist_{}_kfac{}_gpu{}_bs{}_{}_lr{}_sr{}_wp{}.log'.format(
args.model, args.kfac_update_freq, hvd.size(), args.batch_size,
args.kfac_name, args.base_lr, args.sparse_ratio, args.warmup_epochs)
if hvd.rank() == 0:
wandb.init(project='kfac',
entity='hkust-distributedml',
name=logfilename,
config=args)
logfile = './logs/' + logfilename
#logfile = './logs/sparse_cifar10_{}_kfac{}_gpu{}_bs{}.log'.format(args.model, args.kfac_update_freq, hvd.size(), args.batch_size)
#logfile = './logs/cifar10_{}_kfac{}_gpu{}_bs{}.log'.format(args.model, args.kfac_update_freq, hvd.size(), args.batch_size)
hdlr = logging.FileHandler(logfile)
hdlr.setFormatter(formatter)
logger.addHandler(hdlr)
logger.info(args)
torch.manual_seed(args.seed)
verbose = True if hvd.rank() == 0 else False
args.verbose = 1 if hvd.rank() == 0 else 0
if args.cuda:
torch.cuda.set_device(hvd.local_rank())
torch.cuda.manual_seed(args.seed)
torch.backends.cudnn.benchmark = True
args.log_dir = os.path.join(
args.log_dir, "mnist_{}_kfac{}_gpu_{}_{}".format(
args.model, args.kfac_update_freq, hvd.size(),
datetime.datetime.now().strftime('%Y-%m-%d_%H-%M-%S')))
#os.makedirs(args.log_dir, exist_ok=True)
#log_writer = SummaryWriter(args.log_dir) if verbose else None
log_writer = None
# Horovod: limit # of CPU threads to be used per worker.
torch.set_num_threads(1)
kwargs = {'num_workers': 8, 'pin_memory': True} if args.cuda else {}
transform_train = []
if args.model.lower() == 'lenet':
transform_train.append(transforms.Resize(32))
transform_train.extend(
[transforms.ToTensor(),
transforms.Normalize((0.1307, ), (0.3081, ))])
transform_train = transforms.Compose(transform_train)
transform_test = transform_train
download = True if hvd.local_rank() == 0 else False
if not download: hvd.allreduce(torch.tensor(1), name="barrier")
train_dataset = torchvision.datasets.MNIST(root=self.dir,
train=True,
download=download,
transform=transform_train)
test_dataset = torchvision.datasets.MNIST(root=args.dir,
train=False,
download=download,
transform=transform_test)
if download: hvd.allreduce(torch.tensor(1), name="barrier")
# Horovod: use DistributedSampler to partition the training data.
train_sampler = torch.utils.data.distributed.DistributedSampler(
train_dataset, num_replicas=hvd.size(), rank=hvd.rank())
#train_loader = torch.utils.data.DataLoader(train_dataset,
train_loader = MultiEpochsDataLoader(train_dataset,
batch_size=args.batch_size *
args.batches_per_allreduce,
sampler=train_sampler,
**kwargs)
# Horovod: use DistributedSampler to partition the test data.
test_sampler = torch.utils.data.distributed.DistributedSampler(
test_dataset, num_replicas=hvd.size(), rank=hvd.rank())
test_loader = torch.utils.data.DataLoader(test_dataset,
batch_size=args.test_batch_size,
sampler=test_sampler,
**kwargs)
if args.model.lower() == "fcn5net":
model = fcn.FCN5Net()
elif args.model.lower() == "lenet":
model = lenet.LeNet()
if args.cuda:
model.cuda()
#if verbose:
# summary(model, (3, 32, 32))
criterion = nn.CrossEntropyLoss()
args.base_lr = args.base_lr * hvd.size()
use_kfac = True if args.kfac_update_freq > 0 else False
optimizer = optim.SGD(model.parameters(),
lr=args.base_lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
if use_kfac:
KFAC = kfac.get_kfac_module(args.kfac_name)
preconditioner = KFAC(
model,
lr=args.base_lr,
factor_decay=args.stat_decay,
damping=args.damping,
kl_clip=args.kl_clip,
fac_update_freq=args.kfac_cov_update_freq,
kfac_update_freq=args.kfac_update_freq,
diag_blocks=args.diag_blocks,
diag_warmup=args.diag_warmup,
distribute_layer_factors=args.distribute_layer_factors,
sparse_ratio=args.sparse_ratio)
kfac_param_scheduler = kfac.KFACParamScheduler(
preconditioner,
damping_alpha=args.damping_alpha,
damping_schedule=args.damping_schedule,
update_freq_alpha=args.kfac_update_freq_alpha,
update_freq_schedule=args.kfac_update_freq_schedule)
# KFAC guarentees grads are equal across ranks before opt.step() is called
# so if we do not use kfac we need to wrap the optimizer with horovod
compression = hvd.Compression.fp16 if args.fp16_allreduce else hvd.Compression.none
optimizer = hvd.DistributedOptimizer(
optimizer,
named_parameters=model.named_parameters(),
compression=compression,
op=hvd.Average,
backward_passes_per_step=args.batches_per_allreduce)
hvd.broadcast_optimizer_state(optimizer, root_rank=0)
hvd.broadcast_parameters(model.state_dict(), root_rank=0)
lrs = create_lr_schedule(hvd.size(), args.warmup_epochs, args.lr_decay)
lr_scheduler = [LambdaLR(optimizer, lrs)]
if use_kfac:
lr_scheduler.append(LambdaLR(preconditioner, lrs))
def train(epoch):
model.train()
train_sampler.set_epoch(epoch)
train_loss = Metric('train_loss')
train_accuracy = Metric('train_accuracy')
if STEP_FIRST:
for scheduler in lr_scheduler:
scheduler.step()
if use_kfac:
kfac_param_scheduler.step(epoch)
# with tqdm(total=len(train_loader),
# desc='Epoch {:3d}/{:3d}'.format(epoch + 1, args.epochs),
# disable=not verbose) as t:
display = 20
avg_time = 0.0
io_time = 0.0
if True:
for batch_idx, (data, target) in enumerate(train_loader):
stime = time.time()
if args.cuda:
data, target = data.cuda(non_blocking=True), target.cuda(
non_blocking=True)
io_time += time.time() - stime
optimizer.zero_grad()
for i in range(0, len(data), args.batch_size):
data_batch = data[i:i + args.batch_size]
target_batch = target[i:i + args.batch_size]
output = model(data_batch)
loss = criterion(output, target_batch)
with torch.no_grad():
train_loss.update(loss)
train_accuracy.update(accuracy(output, target_batch))
loss.div_(math.ceil(float(len(data)) / args.batch_size))
loss.backward()
optimizer.synchronize()
if use_kfac:
preconditioner.step(epoch=epoch)
with optimizer.skip_synchronize():
optimizer.step()
#t.set_postfix_str("loss: {:.4f}, acc: {:.2f}%".format(
#train_loss.avg.item(), 100*train_accuracy.avg.item()))
#t.update(1)
avg_time += (time.time() - stime)
if batch_idx > 0 and batch_idx % display == 0:
if args.verbose:
logger.info(
"[%d][%d] train loss: %.4f, acc: %.3f, time: %.3f [io: %.3f], speed: %.3f images/s"
% (epoch, batch_idx, train_loss.avg.item(), 100 *
train_accuracy.avg.item(), avg_time / display,
io_time / display, args.batch_size /
(avg_time / display)))
avg_time = 0.0
io_time = 0.0
if hvd.rank() == 0:
wandb.log({"loss": loss, "epoch": epoch})
if args.verbose:
logger.info("[%d] epoch train loss: %.4f, acc: %.3f" %
(epoch, train_loss.avg.item(),
100 * train_accuracy.avg.item()))
if not STEP_FIRST:
for scheduler in lr_scheduler:
scheduler.step()
if use_kfac:
kfac_param_scheduler.step(epoch)
if log_writer:
log_writer.add_scalar('train/loss', train_loss.avg, epoch)
log_writer.add_scalar('train/accuracy', train_accuracy.avg, epoch)
def test(epoch):
model.eval()
test_loss = Metric('val_loss')
test_accuracy = Metric('val_accuracy')
#with tqdm(total=len(test_loader),
# bar_format='{l_bar}{bar}|{postfix}',
# desc=' '.format(epoch + 1, args.epochs),
# disable=not verbose) as t:
if True:
with torch.no_grad():
for i, (data, target) in enumerate(test_loader):
if args.cuda:
data, target = data.cuda(), target.cuda()
output = model(data)
test_loss.update(criterion(output, target))
test_accuracy.update(accuracy(output, target))
if args.verbose:
logger.info("[%d][0] evaluation loss: %.4f, acc: %.3f" %
(epoch, test_loss.avg.item(),
100 * test_accuracy.avg.item()))
if hvd.rank() == 0:
wandb.log({
"val top-1 acc": test_accuracy.avg.item(),
"epoch": epoch
})
#t.update(1)
#if i + 1 == len(test_loader):
# t.set_postfix_str("\b\b test_loss: {:.4f}, test_acc: {:.2f}%".format(
# test_loss.avg.item(), 100*test_accuracy.avg.item()),
# refresh=False)
if log_writer:
log_writer.add_scalar('test/loss', test_loss.avg, epoch)
log_writer.add_scalar('test/accuracy', test_accuracy.avg, epoch)
start = time.time()
for epoch in range(args.epochs):
if args.verbose:
logger.info("[%d] epoch train starts" % (epoch))
train(epoch)
test(epoch)
if verbose:
logger.info("Training time: %s",
str(datetime.timedelta(seconds=time.time() - start)))
pass
from keras.preprocessing.image import ImageDataGenerator
import os,sys
import numpy as np
model_path = os.path.join('..','models','keras','models')
sys.path.append(model_path)
import lenet
from imagenet_utils import preprocess_input
from keras.preprocessing import image as keras_image
weights_path = os.path.join('..', 'models', 'keras', 'weights', 'weights_basic.h5')
#model.load_weights(weights_path)
#model = lenet.LeNet(weights_path=weights_path)
model = lenet.LeNet(weights_path)
model.compile(loss='categorical_crossentropy',
optimizer='rmsprop',
metrics=['accuracy'])
# this is the augmentation configuration we will use for training
train_datagen = ImageDataGenerator(
rescale=1./255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
# this is the augmentation configuration we will use for testing:
# only rescaling
test_datagen = ImageDataGenerator(rescale=1./255)
# this is a generator that will read pictures found in
MODEL = 'lenet_E' + str(EPOCHS) + '_D' + str(dropout) + '_R' + str(
rate)
model_dir = '/home/valentin/Desktop/MLProject/models/' + MODEL + '/'
save_dir = '/home/valentin/Desktop/MLProject/results/' + MODEL + '/'
if not os.path.exists(save_dir):
os.makedirs(save_dir)
#placeholder for model
x = tf.placeholder(tf.float32, (None, 32, 32, 1))
y = tf.placeholder(tf.int32, (None))
one_hot_y = tf.one_hot(y, 10)
keep_prob = tf.placeholder(
tf.float32) # dropout (keep probability)
# load Alexnet
logits = lenet.LeNet(x, keep_prob)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(
logits=logits, labels=one_hot_y)
loss_operation = tf.reduce_mean(cross_entropy)
optimizer = tf.train.AdamOptimizer(learning_rate=rate)
training_operation = optimizer.minimize(loss_operation)
correct_prediction = tf.equal(tf.argmax(logits, 1),
tf.argmax(one_hot_y, 1))
accuracy_operation = tf.reduce_mean(
tf.cast(correct_prediction, tf.float32))
train_prediction = tf.nn.softmax(logits)
output = tf.argmax(logits, 1)
# evaluate model accuracy with validation set
import lenet
import simplenet
import resnet
import resnet_with_compression
import mobilenet
import mobilenetv2
import densenet
import dpn
import preact_resnet
cifar10_networks = {
'lenet': lenet.LeNet(),
'simplenet9': simplenet.SimpleNet9(),
'simplenet9_thin': simplenet.SimpleNet9_thin(),
'simplenet9_mobile': simplenet.SimpleNet9_mobile(),
'simplenet7': simplenet.SimpleNet7(),
'simplenet7_thin': simplenet.SimpleNet7_thin(),
'resnet18NNFC1': resnet_with_compression.ResNet18NNFC1(),
'resnet18EH0': resnet_with_compression.ResNet18EH(layer=0, quantizer=20),
'resnet18EH1': resnet_with_compression.ResNet18EH(layer=1, quantizer=6),
'resnet18EH2': resnet_with_compression.ResNet18EH(layer=2, quantizer=5),
'resnet18EH3': resnet_with_compression.ResNet18EH(layer=3, quantizer=3),
'resnet18EH4': resnet_with_compression.ResNet18EH(layer=4, quantizer=10),
'resnet18JPEG90': resnet_with_compression.ResNet18JPEG(quantizer=90),
'resnet18JPEG87': resnet_with_compression.ResNet18JPEG(quantizer=87),
'resnet18AVC': resnet_with_compression.ResNet18AVC(layer=2, quantizer=24),
'resnet18': resnet.ResNet18(),
'resnet101': resnet.ResNet101(),
'mobilenetslimplus': mobilenet.MobileNetSlimPlus(),
'mobilenetslim': mobilenet.MobileNetSlim(),
'mobilenet': mobilenet.MobileNet(),
def main():
global args, best_prec1
args = parser.parse_args()
torch.manual_seed(args.seed)
# Check the save_dir exists or not
if not os.path.exists(args.save_dir):
os.makedirs(args.save_dir)
#model = torch.nn.DataParallel(resnet.__dict__[args.arch]())
if args.arch == 'lenet':
model = lenet.LeNet(num_classes=int(args.data[5:]))
else:
model = resnet.__dict__[args.arch](num_classes=int(args.data[5:]))
origpar = sum(param.numel() for param in model.parameters())
print('Original weight count:', origpar)
torch.cuda.set_device(args.device)
criterion = nn.CrossEntropyLoss().cuda()
writer = SummaryWriter(args.save_dir)
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
if args.permute or args.get_permute:
if args.get_permute:
permute = torch.load(args.get_permute)['permute']
elif args.resume:
permute = torch.load(args.resume)['permute']
else:
permute = RowColPermute(32, 32)
train_transforms = [transforms.ToTensor(), permute, normalize]
val_transforms = [transforms.ToTensor(), permute, normalize]
else:
permute = None
train_transforms = [transforms.RandomHorizontalFlip(), transforms.RandomCrop(32, 4), transforms.ToTensor(), normalize]
val_transforms = [transforms.ToTensor(), normalize]
cifar = datasets.CIFAR100 if args.data == 'cifar100' else datasets.CIFAR10
train_loader = torch.utils.data.DataLoader(
cifar(root='./data', train=True, transform=transforms.Compose(train_transforms), download=True),
batch_size=args.batch_size, shuffle=True,
num_workers=args.workers, pin_memory=True)
val_loader = torch.utils.data.DataLoader(
cifar(root='./data', train=False, transform=transforms.Compose(val_transforms)),
batch_size=128, shuffle=False,
num_workers=args.workers, pin_memory=True)
# define optimizer
if args.half:
model.half()
criterion.half()
if args.darts:
model, original = Supernet.metamorphosize(model), model
X, _ = next(iter(train_loader))
arch_kwargs = {'perturb': args.perturb,
'verbose': not args.resume,
'warm_start': not args.from_scratch}
patchlist = (['conv'] if args.patch_conv else []) \
+ (['pool'] if args.patch_pool else []) \
+ (['shortcut'] if args.patch_skip else [])
model.patch_darts(X[:1], named_modules=((n, m) for n, m in model.named_modules() if any(patch in n for patch in patchlist)), **arch_kwargs)
print('Model weight count:', sum(p.numel() for p in model.model_weights()))
print('Arch param count:', sum(p.numel() for p in model.arch_params()))
else:
model, original = Chrysalis.metamorphosize(model), model
if args.patch_skip or args.patch_conv or args.patch_pool:
X, _ = next(iter(train_loader))
arch_kwargs = {key: getattr(args, key) for key in [
'kmatrix_depth',
'max_kernel_size',
'global_biasing',
'channel_gating',
'base',
'perturb',
]}
arch_kwargs['verbose'] = not args.resume
arch_kwargs['warm_start'] = not args.from_scratch
if args.patch_skip:
model.patch_skip(X[:1], named_modules=((n, m) for n, m in model.named_modules() if 'shortcut' in n), **arch_kwargs)
if args.patch_pool:
model.patch_pool(X[:1], named_modules=((n, m) for n, m in model.named_modules() if 'pool' in n), **arch_kwargs)
if args.patch_conv:
model.patch_conv(X[:1], **arch_kwargs)
print('Model weight count:', sum(p.numel() for p in model.model_weights()))
print('Arch param count:', sum(p.numel() for p in model.arch_params()))
else:
args.arch_lr = 0.0
# optionally resume from a checkpoint
if args.resume:
if os.path.isfile(args.resume):
print("=> loading checkpoint '{}'".format(args.resume))
checkpoint = torch.load(args.resume)
args.start_epoch = checkpoint['epoch']
best_prec1 = checkpoint['best_prec1']
model.load_state_dict(checkpoint['state_dict'])
print("=> loaded checkpoint '{}' (epoch {})"
.format(args.evaluate, checkpoint['epoch']))
else:
print("=> no checkpoint found at '{}'".format(args.resume))
if args.offline:
model.load_arch(args.offline)
args.arch_lr = 0.0
if args.darts:
model.discretize()
for name, module in model.named_modules():
if hasattr(module, 'discrete'):
print(name, '\t', module.discrete)
cudnn.benchmark = True
model.cuda()
momentum = partial(torch.optim.SGD, momentum=args.momentum)
opts = [momentum(model.model_weights(), lr=args.lr, weight_decay=args.weight_decay)]
if args.arch_lr:
arch_opt = torch.optim.Adam if args.arch_adam else momentum
opts.append(arch_opt(model.arch_params(), lr=args.arch_lr, weight_decay=0.0 if args.arch_adam else args.weight_decay))
optimizer = MixedOptimizer(opts)
def weight_sched(epoch):
if args.arch in ['lenet']:
return 0.1 if epoch >= int(0.75 * args.epochs) else 0.5 if epoch >= int(0.5 * args.epochs) else 1.0
if epoch < 1 and args.arch in ['resnet1202', 'resnet110']:
return 0.1
return 0.1 ** (epoch >= int(0.5 * args.epochs)) * 0.1 ** (epoch >= int(0.75 * args.epochs))
def arch_sched(epoch):
return 0.0 if epoch < args.warmup_epochs or epoch > args.epochs-args.cooldown_epochs else weight_sched(epoch)
sched_groups = [weight_sched if g['params'][0] in set(model.model_weights()) else arch_sched for g in optimizer.param_groups]
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=sched_groups, last_epoch=args.start_epoch-1)
def metrics(epoch):
if args.darts:
return
for label, name, patched in [
('skip', 'shortcut', args.patch_skip),
('pool', 'pool', args.patch_pool),
('conv', 'conv', args.patch_conv),
]:
if patched:
mods = [m for n, m in model.named_modules() if name in n and hasattr(m, 'distance_from')]
for metric, metric_kwargs in [
('euclidean', {}),
('frobenius', {'approx': 16}),
('averaged', {'approx': 16, 'samples': 10}),
]:
writer.add_scalar('/'.join([label, metric+'-dist']),
sum(m.distance_from(label, metric=metric, relative=True, **metric_kwargs) for m in mods) / len(mods),
epoch)
if not metric == 'averaged':
writer.add_scalar('/'.join([label, metric+'-norm']),
sum(getattr(m, metric)(**metric_kwargs) for m in mods) / len(mods),
epoch)
writer.add_scalar(label+'/weight-norm', sum(m.weight.data.norm() for m in mods) / len(mods), epoch)
if args.evaluate:
validate(val_loader, model, criterion)
return
with open(os.path.join(args.save_dir, 'args.json'), 'w') as f:
json.dump(vars(args), f, indent=4)
for epoch in range(args.start_epoch, args.epochs):
writer.add_scalar('hyper/lr', weight_sched(epoch) * args.lr, epoch)
writer.add_scalar('hyper/arch', arch_sched(epoch) * args.arch_lr, epoch)
metrics(epoch)
model.set_arch_requires_grad(arch_sched(epoch) * args.arch_lr > 0.0)
# train for one epoch
print('current lr {:.5e}'.format(optimizer.param_groups[0]['lr']))
acc, loss = train(train_loader, model, criterion, optimizer, epoch)
writer.add_scalar('train/acc', acc, epoch)
writer.add_scalar('train/loss', loss, epoch)
lr_scheduler.step()
# evaluate on validation set
prec1, loss = validate(val_loader, model, criterion)
writer.add_scalar('valid/acc', prec1, epoch)
writer.add_scalar('valid/loss', loss, epoch)
# remember best prec@1 and save checkpoint
is_best = prec1 > best_prec1
best_prec1 = max(prec1, best_prec1)
model.train()
if epoch > 0 and epoch % args.save_every == 0:
save_checkpoint({
'epoch': epoch + 1,
'state_dict': model.state_dict(),
'best_prec1': best_prec1,
'permute': permute,
}, is_best, filename=os.path.join(args.save_dir, 'checkpoint.th'))
if epoch > 0 and epoch+1 == args.warmup_epochs:
save_checkpoint({
'epoch': epoch + 1,
'state_dict': model.state_dict(),
'best_prec1': best_prec1,
'permute': permute,
}, is_best, filename=os.path.join(args.save_dir, 'warmup.th'))
if epoch > 0 or epoch+1 == args.epochs-args.cooldown_epochs:
save_checkpoint({
'epoch': epoch + 1,
'state_dict': model.state_dict(),
'best_prec1': best_prec1,
'permute': permute,
}, is_best, filename=os.path.join(args.save_dir, 'cooldown.th'))
save_checkpoint({
'state_dict': model.state_dict(),
'best_prec1': best_prec1,
'permute': permute,
}, is_best, filename=os.path.join(args.save_dir, 'model.th'))
model.save_arch(os.path.join(args.save_dir, 'arch.th'))
metrics(args.epochs)
writer.flush()
with open(os.path.join(args.save_dir, 'results.json'), 'w') as f:
json.dump({'final validation accuracy': prec1,
'best validation accuracy': best_prec1,
}, f, indent=4)