def load_checkpoint(context):
args = context['args']
model = context['model']
if args.resume == '':
return
t_saver = Saver(model_dir=args.resume)
print('==> loading checkpoint from {}'.format(args.resume))
if args.evaluate:
checkpoint = t_saver.load_best()
else:
checkpoint = t_saver.load_latest()
if checkpoint:
best_metric = checkpoint['best_metric']
context['best_metric'] = best_metric
model.load_state_dict(checkpoint['model_state_dict'])
if 'step' in checkpoint:
step = checkpoint['step']
else:
step = 0
if args.step != -1:
step = args.step
print("==> loaded checkpoint {} (step {}, best_metric {})".format(args.resume,
step, best_metric))
context['step'] = step
else:
raise RuntimeError("==> no checkpoint at: {}".format(args.resume))
def main():
global step, best_prec1, svm_layer, rgbCNN, flowCNN, crit, optimizer, saver, writer
global train_dataset, test_dataset, train_loader, test_loader, vocab
print('prepare dataset...')
(train_dataset, train_loader), (test_dataset,
test_loader) = prepare_dataset()
# prepare model
svm_layer, rgbCNN, flowCNN, crit, optimizer = prepare_model()
if args.resume != '':
t_saver = Saver(model_dir=args.resume, max_to_keep=5)
print("=> loading checkpoint '{}'".format(args.resume))
if args.evaluate:
checkpoint = t_saver.load_best()
# checkpoint = t_saver.load_latest()
else:
checkpoint = t_saver.load_latest()
if checkpoint is not None:
# torch.load(args.resume)
best_prec1 = checkpoint['best_prec1']
svm_layer.load_state_dict(checkpoint['model_state_dict'])
if 'step' in checkpoint:
step = checkpoint['step']
else:
step = 0
if args.step != -1:
step = args.step
print("=> loaded checkpoint '{}' (step {})".format(
args.resume, checkpoint['step']))
else:
raise RuntimeError("=> no checkpoint found at '{}'".format(
args.resume))
if args.evaluate:
validate(True)
return
print('prepare logger...')
# writer, saver = prepare_logger()
print('start training...')
with torch.no_grad():
feature_Matrix, labels = train()
np.save('train_feature_list_hmdb.npy', np.array(feature_Matrix))
np.save('train_labels_list_hmdb.npy', np.array(labels))
print('done!')
