MeCo/zero-cost-nas/nasbench2_train.py
HamsterMimi 189df25fd3 upload
2023-05-04 13:09:03 +08:00

184 lines
7.0 KiB
Python

# Copyright 2021 Samsung Electronics Co., Ltd.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# =============================================================================
import argparse
import pickle
import time
import torch
import torch.optim as optim
import torch.nn.functional as F
from ignite.engine import Events, create_supervised_trainer, create_supervised_evaluator
from ignite.metrics import Accuracy, Loss
from ignite.contrib.handlers import ProgressBar
from foresight.models import *
from foresight.pruners import *
from foresight.dataset import *
def get_num_classes(args):
return 100 if args.dataset == 'cifar100' else 10 if args.dataset == 'cifar10' else 120
def setup_experiment(net, args):
optimiser = optim.SGD(
net.parameters(),
lr=0.1,
momentum=0.9,
weight_decay=0.0005,
nesterov=True)
lr_scheduler = optim.lr_scheduler.CosineAnnealingLR(optimiser, T_max=args.epochs, eta_min=0, last_epoch=-1)
train_loader, val_loader = get_cifar_dataloaders(args.batch_size, args.batch_size, args.dataset, args.num_data_workers, resize=args.img_size)
return optimiser, lr_scheduler, train_loader, val_loader
def parse_arguments():
parser = argparse.ArgumentParser(description='EcoNAS Training Pipeline for NAS-Bench-201')
parser.add_argument('--api_loc', default='data/NAS-Bench-201-v1_0-e61699.pth',
type=str, help='path to API')
parser.add_argument('--outdir', default='./',
type=str, help='output directory')
parser.add_argument('--outfname', default='test',
type=str, help='output filename')
parser.add_argument('--batch_size', default=2048, type=int)
parser.add_argument('--epochs', default=40, type=int)
parser.add_argument('--init_channels', default=4, type=int)
parser.add_argument('--img_size', default=8, type=int)
parser.add_argument('--dataset', type=str, default='cifar10', help='dataset to use [cifar10, cifar100, ImageNet16-120]')
parser.add_argument('--gpu', type=int, default=0, help='GPU index to work on')
parser.add_argument('--num_data_workers', type=int, default=2, help='number of workers for dataloaders')
parser.add_argument('--dataload', type=str, default='random', help='random or grasp supported')
parser.add_argument('--dataload_info', type=int, default=1, help='number of batches to use for random dataload or number of samples per class for grasp dataload')
parser.add_argument('--start', type=int, default=5, help='start index')
parser.add_argument('--end', type=int, default=10, help='end index')
parser.add_argument('--write_freq', type=int, default=5, help='frequency of write to file')
parser.add_argument('--logmeasures', action="store_true", default=False, help='add extra logging for predictive measures')
args = parser.parse_args()
args.device = torch.device("cuda:"+str(args.gpu) if torch.cuda.is_available() else "cpu")
return args
def train_nb2():
args = parse_arguments()
archs = pickle.load(open(args.api_loc,'rb'))
pre='cf' if 'cifar' in args.dataset else 'im'
if args.outfname == 'test':
fn = f'nb2_train_{pre}{get_num_classes(args)}_r{args.img_size}_c{args.init_channels}_e{args.epochs}.p'
else:
fn = f'{args.outfname}.p'
op = os.path.join(args.outdir,fn)
print('outfile =',op)
cached_res = []
#loop over nasbench2 archs
for i, arch_str in enumerate(archs):
if i < args.start:
continue
if i >= args.end:
break
res = {'idx':i, 'arch_str':arch_str, 'logmeasures':[]}
net = nasbench2.get_model_from_arch_str(arch_str, get_num_classes(args), init_channels=args.init_channels)
net.to(args.device)
optimiser, lr_scheduler, train_loader, val_loader = setup_experiment(net, args)
#start training
criterion = F.cross_entropy
trainer = create_supervised_trainer(net, optimiser, criterion, args.device)
evaluator = create_supervised_evaluator(net, {
'accuracy': Accuracy(),
'loss': Loss(criterion)
}, args.device)
pbar = ProgressBar()
pbar.attach(trainer)
@trainer.on(Events.EPOCH_COMPLETED)
def log_epoch(engine):
#change LR
lr_scheduler.step()
#run evaluator
evaluator.run(val_loader)
#metrics
metrics = evaluator.state.metrics
avg_accuracy = metrics['accuracy']
avg_loss = metrics['loss']
pbar.log_message(f"Validation Results - Epoch: {engine.state.epoch} Avg accuracy: {round(avg_accuracy*100,2)}% Val loss: {round(avg_loss,2)} Train loss: {round(engine.state.output,2)}")
measures = {}
if args.logmeasures:
measures = predictive.find_measures(net,
train_loader,
(args.dataload, args.dataload_info, get_num_classes(args)),
args.device)
measures['train_loss'] = engine.state.output
measures['val_loss'] = avg_loss
measures['val_acc'] = avg_accuracy
measures['epoch'] = engine.state.epoch
res['logmeasures'].append(measures)
#at epoch zero
#run evaluator
evaluator.run(val_loader)
#metrics
metrics = evaluator.state.metrics
avg_accuracy = metrics['accuracy']
avg_loss = metrics['loss']
measures = {}
if args.logmeasures:
measures = predictive.find_measures(net,
train_loader,
(args.dataload, args.dataload_info, get_num_classes(args)),
args.device)
measures['train_loss'] = 0
measures['val_loss'] = avg_loss
measures['val_acc'] = avg_accuracy
measures['epoch'] = 0
res['logmeasures'].append(measures)
#run training
stime = time.time()
trainer.run(train_loader, args.epochs)
etime = time.time()
res['time'] = etime-stime
#print(res)
cached_res.append(res)
#write to file
if i % args.write_freq == 0 or i == args.end-1 or i == args.start + 10:
print(f'writing {len(cached_res)} results to {op}')
pf=open(op, 'ab')
for cr in cached_res:
pickle.dump(cr, pf)
pf.close()
cached_res = []
if __name__ == '__main__':
train_nb2()