def load_best_model(best_model_info, device, config_str, model_type, model_name, load_filename=None):
if load_filename is None:
savename = best_model_info['savename']
split = savename.split('/')
split[-1] = 'best_' + split[-1]
load_filename = '/'.join(split)
try:
_iter = checkpoint['_iter']
except:
_iter = checkpoint['_epoch']
print("Loaded checkpoint (trained for {} iterations)".format(_iter))
try:
params = best_model_info[["momentum", "seed", "lr", "batch_size", "augmentation"]]
except:
params = best_model_info[["seed", "lr", "batch_size", "augmentation"]]
params = params.to_dict()
params_alt = {}
for key, value in params.items():
if (type(value) == str):
params_alt[key] = [eval(value)]
else:
params_alt[key] = [value]
exp = Experiment(None, device, config_str, model_type, model_name, params_alt, None, None)
model, criterion, optimizer = exp._get_model(best_model_info['seed'], params)
model = nn.DataParallel(model)
model.load_state_dict(checkpoint['state_dict'])
model.to(device)
return checkpoint, model, criterion,exp
