def main():
args = parse_args()
update_config(args.cfg)
fix_random_seed(cfg.SEED)
# cudnn related setting
cudnn.benchmark = cfg.CUDNN.BENCHMARK
cudnn.deterministic = cfg.CUDNN.DETERMINISTIC
cudnn.enabled = cfg.CUDNN.ENABLE
# prepare output directory
output_dir = os.path.join(cfg.ROOT_DIR, cfg.TRAIN.MODEL_DIR)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
# copy config ssad.yaml file to output directory
cfg_file = os.path.join(output_dir, args.cfg.split('/')[-1])
shutil.copyfile(args.cfg, cfg_file)
# data loader
# Notice: we discard the last data
train_dset = TALDataset(cfg, cfg.DATASET.TRAIN_SPLIT)
train_loader = DataLoader(train_dset,
batch_size=cfg.TRAIN.BATCH_SIZE,
shuffle=True,
drop_last=False,
num_workers=cfg.WORKERS,
pin_memory=cfg.PIN_MEMORY)
val_dset = TALDataset(cfg, cfg.DATASET.VAL_SPLIT)
val_loader = DataLoader(val_dset,
batch_size=cfg.TEST.BATCH_SIZE,
shuffle=False,
drop_last=False,
num_workers=cfg.WORKERS,
pin_memory=cfg.PIN_MEMORY)
model = LocNet(cfg)
model.apply(weight_init)
model.cuda()
optimizer = optim.Adam(model.parameters(), lr=cfg.TRAIN.LR)
for epoch in range(cfg.TRAIN.BEGIN_EPOCH, cfg.TRAIN.END_EPOCH + 1):
if epoch in cfg.TRAIN.LR_DECAY_EPOCHS:
decay_lr(optimizer, factor=cfg.TRAIN.LR_DECAY_FACTOR)
loss_train = train(cfg, train_loader, model, optimizer)
print('epoch %d: loss: %f' % (epoch, loss_train))
with open(os.path.join(cfg.ROOT_DIR, cfg.TRAIN.LOG_FILE), 'a') as f:
f.write("epoch %d, loss: %.4f\n" % (epoch, loss_train))
if epoch % cfg.TEST.EVAL_INTERVAL == 0:
# model
weight_file = save_model(cfg,
epoch=epoch,
model=model,
optimizer=optimizer)
out_df_af, out_df_ab = evaluation(val_loader, model, epoch, cfg)
out_df_ab['conf'] = out_df_ab['conf'] * cfg.TEST.CONCAT_AB
out_df = pd.concat([out_df_af, out_df_ab])
post_process(out_df, epoch, cfg, is_soft_nms=False)
def main():
args = parse_args()
update_config(args.cfg)
# create output directory
if cfg.BASIC.CREATE_OUTPUT_DIR:
out_dir = os.path.join(cfg.BASIC.ROOT_DIR, cfg.TRAIN.MODEL_DIR)
if not os.path.exists(out_dir):
os.makedirs(out_dir)
# copy config file
if cfg.BASIC.BACKUP_CODES:
backup_dir = os.path.join(cfg.BASIC.ROOT_DIR, cfg.TRAIN.MODEL_DIR,
'code')
backup_codes(cfg.BASIC.ROOT_DIR, backup_dir, cfg.BASIC.BACKUP_LISTS)
fix_random_seed(cfg.BASIC.SEED)
if cfg.BASIC.SHOW_CFG:
pprint.pprint(cfg)
# cudnn related setting
cudnn.benchmark = cfg.CUDNN.BENCHMARK
cudnn.deterministic = cfg.CUDNN.DETERMINISTIC
cudnn.enabled = cfg.CUDNN.ENABLE
# data loader
train_dset = TALDataset(cfg, cfg.DATASET.TRAIN_SPLIT)
train_loader = DataLoader(train_dset,
batch_size=cfg.TRAIN.BATCH_SIZE,
shuffle=True,
drop_last=False,
num_workers=cfg.BASIC.WORKERS,
pin_memory=cfg.DATASET.PIN_MEMORY)
val_dset = TALDataset(cfg, cfg.DATASET.VAL_SPLIT)
val_loader = DataLoader(val_dset,
batch_size=cfg.TEST.BATCH_SIZE,
shuffle=False,
drop_last=False,
num_workers=cfg.BASIC.WORKERS,
pin_memory=cfg.DATASET.PIN_MEMORY)
model = LocNet(cfg)
model.apply(weight_init)
model.cuda()
optimizer = optim.Adam(model.parameters(), lr=cfg.TRAIN.LR)
for epoch in range(cfg.TRAIN.BEGIN_EPOCH, cfg.TRAIN.END_EPOCH + 1):
loss_train = train(cfg, train_loader, model, optimizer)
print('epoch %d: loss: %f' % (epoch, loss_train))
with open(os.path.join(cfg.BASIC.ROOT_DIR, cfg.TRAIN.LOG_FILE),
'a') as f:
f.write("epoch %d, loss: %.4f\n" % (epoch, loss_train))
# decay lr
if epoch in cfg.TRAIN.LR_DECAY_EPOCHS:
decay_lr(optimizer, factor=cfg.TRAIN.LR_DECAY_FACTOR)
if epoch in cfg.TEST.EVAL_INTERVAL:
save_model(cfg, epoch=epoch, model=model, optimizer=optimizer)
out_df_ab, out_df_af = evaluation(val_loader, model, epoch, cfg)
out_df_list = [out_df_ab, out_df_af]
final_result_process(out_df_list, epoch, cfg, flag=0)
def main():
args = args_parser()
update_config(args.cfg)
if cfg.BASIC.SHOW_CFG:
pprint.pprint(cfg)
# prepare running environment for the whole project
prepare_env(cfg)
# log
writer = SummaryWriter(
log_dir=os.path.join(cfg.BASIC.CKPT_DIR, cfg.BASIC.LOG_DIR))
# dataloader
train_dset = WtalDataset(cfg, cfg.DATASET.TRAIN_SPLIT)
train_loader = DataLoader(train_dset,
batch_size=cfg.TRAIN.BATCH_SIZE,
shuffle=True,
num_workers=cfg.BASIC.WORKERS,
pin_memory=cfg.BASIC.PIN_MEMORY)
val_dset = WtalDataset(cfg, cfg.DATASET.VAL_SPLIT)
val_loader = DataLoader(val_dset,
batch_size=cfg.TEST.BATCH_SIZE,
shuffle=False,
num_workers=cfg.BASIC.WORKERS,
pin_memory=cfg.BASIC.PIN_MEMORY)
# network
model = LocNet(cfg)
# model.apply(weight_init)
model.cuda()
# weight_file = "/disk3/zt/code/actloc/thumos/17_CAS_CAM_fast_tuning/output/20class_seed_0_save_model/checkpoint_best_cas_0.2701.pth"
weight_file = '/disk3/zt/code/actloc/thumos/20_0.2701_try/output/debug_save_epoch30/checkpoint_best_cas.pth'
from utils.utils import load_weights
model = load_weights(model, weight_file)
# optimizer
optimizer = optim.Adam(model.parameters(),
lr=cfg.TRAIN.LR,
betas=cfg.TRAIN.BETAS,
weight_decay=cfg.TRAIN.WEIGHT_DECAY)
optimizer.load_state_dict(torch.load(weight_file)['optimizer'])
# criterion
criterion = BasNetLoss()
for epoch in range(1, cfg.TRAIN.EPOCH_NUM + 1):
print('Epoch: %d:' % epoch)
loss_average_cas, loss_average_cam, loss_average_consistency, loss_average_norm, loss_average_cas_inv, loss_average_cam_inv = train(
cfg, train_loader, model, optimizer, criterion)
writer.add_scalar('train_loss/cas', loss_average_cas, epoch)
writer.add_scalar('train_loss/cam', loss_average_cam, epoch)
writer.add_scalar('train_loss/consistency', loss_average_consistency,
epoch)
writer.add_scalar('train_loss/norm', loss_average_norm, epoch)
writer.add_scalar('train_loss/cas_inv', loss_average_cas_inv, epoch)
writer.add_scalar('train_loss/cam_inv', loss_average_cam_inv, epoch)
if cfg.BASIC.VERBOSE:
print(
'loss: cas %f, cam %f, consistency %f, norm %f, cas_inv %f, cam_inv %f'
% (loss_average_cas, loss_average_cam,
loss_average_consistency, loss_average_norm,
loss_average_cas_inv, loss_average_cam_inv))
# decay learning rate
if epoch in cfg.TRAIN.LR_DECAY_EPOCHS:
decay_lr(optimizer, factor=cfg.TRAIN.LR_DECAY_FACTOR)
if epoch % cfg.TEST.EVAL_INTERVAL == 0:
_, _, test_acc_cas, test_acc_cam = evaluate(
cfg, val_loader, model, epoch)
if cfg.BASIC.VERBOSE:
print('test_acc, cas %f, cam %f' %
(test_acc_cas, test_acc_cam))
writer.add_scalar('test_acc/cas', test_acc_cas, epoch)
writer.add_scalar('test_acc/cam', test_acc_cam, epoch)
writer.close()
def main():
args = args_parser()
update_config(args.cfg)
if cfg.BASIC.SHOW_CFG:
pprint.pprint(cfg)
# prepare running environment for the whole project
prepare_env(cfg)
# log
writer = SummaryWriter(
log_dir=os.path.join(cfg.BASIC.CKPT_DIR, cfg.BASIC.LOG_DIR))
# dataloader
train_dset = WtalDataset(cfg, cfg.DATASET.TRAIN_SPLIT)
train_loader = DataLoader(train_dset,
batch_size=cfg.TRAIN.BATCH_SIZE,
shuffle=True,
num_workers=cfg.BASIC.WORKERS,
pin_memory=cfg.BASIC.PIN_MEMORY)
val_dset = WtalDataset(cfg, cfg.DATASET.VAL_SPLIT)
val_loader = DataLoader(val_dset,
batch_size=cfg.TEST.BATCH_SIZE,
shuffle=False,
num_workers=cfg.BASIC.WORKERS,
pin_memory=cfg.BASIC.PIN_MEMORY)
# network
model = LocNet(cfg)
model.cuda()
# optimizer
optimizer = optim.Adam(model.parameters(),
lr=cfg.TRAIN.LR,
betas=cfg.TRAIN.BETAS,
weight_decay=cfg.TRAIN.WEIGHT_DECAY)
# criterion
criterion = BasNetLoss()
for epoch in range(1, cfg.TRAIN.EPOCH_NUM + 1):
print('Epoch: %d:' % epoch)
loss_average_cas, loss_average_cam, loss_average_consistency, loss_average_norm, loss_average_cam_inv = train(
cfg, train_loader, model, optimizer, criterion)
writer.add_scalar('train_loss/cas', loss_average_cas, epoch)
writer.add_scalar('train_loss/cam', loss_average_cam, epoch)
writer.add_scalar('train_loss/consistency', loss_average_consistency,
epoch)
writer.add_scalar('train_loss/norm', loss_average_norm, epoch)
writer.add_scalar('train_loss/cam_inv', loss_average_cam_inv, epoch)
if cfg.BASIC.VERBOSE:
print(
'loss: cas %f, cam %f, consistency %f, norm %f, cam_inv %f' %
(loss_average_cas, loss_average_cam, loss_average_consistency,
loss_average_norm, loss_average_cam_inv))
# decay learning rate
if epoch in cfg.TRAIN.LR_DECAY_EPOCHS:
decay_lr(optimizer, factor=cfg.TRAIN.LR_DECAY_FACTOR)
if epoch % cfg.TEST.EVAL_INTERVAL == 0:
_, test_acc_cas = evaluate(cfg, val_loader, model, epoch)
if cfg.BASIC.VERBOSE:
print('test_acc, cas %f' % (test_acc_cas))
writer.add_scalar('test_acc/cas', test_acc_cas, epoch)
save_best_model(cfg,
epoch=epoch,
model=model,
optimizer=optimizer,
name='cas')
writer.close()