# param.requires_grad = False
model.to(device)
logisticReg.to(device)
optimizer_model = torch.optim.SGD(chain(model.parameters(),
logisticReg.parameters()),
lr=lr,
momentum=0.9,
weight_decay=0.001)
model.train()
logisticReg.train()
# --------------------------------------------------------------------------------------
# Resume training if start is False
# --------------------------------------------------------------------------------------
if not start:
reporter = Reporter(ckpt_root=os.path.join(ROOT_DIR, 'ckpt'),
exp=exp_name,
monitor='acc')
last_model_filename = reporter.select_last(run=run_name).selected_ckpt
last_epoch = int(reporter.select_last(run=run_name).last_epoch)
loss0 = reporter.select_last(run=run_name).last_loss
loss0 = float(loss0[:-4])
model.load_state_dict(
torch.load(last_model_filename)['model_state_dict'])
reporter = Reporter(ckpt_root=os.path.join(ROOT_DIR, 'ckpt'),
exp=exp_name,
monitor='acc')
last_model_filename = reporter.select_last(run=run_name +
'_lr').selected_ckpt
logisticReg.load_state_dict(
torch.load(last_model_filename)['model_state_dict'])
# Model Definitions
# --------------------------------------------------------------------------------------
model = HashSetNet(base_model_architecture=model_type,
num_clusters=num_clusters,
vset_dim=vlad_dim,
vlad_v2=vlad_v2,
pooling=pooling)
logisticReg = LogisticReg()
model.to(device)
logisticReg.to(device)
# --------------------------------------------------------------------------------------
# Loading the model
# --------------------------------------------------------------------------------------
# reporter.monitor = 'auc' or 'acc' ????????????
reporter = Reporter(ckpt_root=os.path.join(ROOT_DIR, 'ckpt'),
exp=exp_name,
monitor='acc')
best_model_filename = reporter.select_best(run=run_name).selected_ckpt
# print(best_model_filename)
model.load_state_dict(torch.load(best_model_filename)['model_state_dict'])
reporter = Reporter(ckpt_root=os.path.join(ROOT_DIR, 'ckpt'),
exp=exp_name,
monitor='acc')
best_model_filename = reporter.select_best(run=run_name +
'_lr').selected_ckpt
# print(best_model_filename)
logisticReg.load_state_dict(
torch.load(best_model_filename)['model_state_dict'])
# model.eval()
# logisticReg.eval()
tot_loss, tot_acc = 0, 0