plot_data.append('val_acc1', val_acc1)
# also evaluate training data
tr_loss, tr_acc1, tr_acc5 = performance_eval(primal_model.net, train_loader4eval, loss_func,
args.cuda, class_offset=class_offset)
print('###Training loss:{:.4e}, top-1 accuracy:{:.5f}, top-5 accuracy:{:.5f}'.format(tr_loss, tr_acc1,
tr_acc5))
plot_data.append('tr_loss', tr_loss)
plot_data.append('tr_acc1', tr_acc1)
print(
'***Validation loss:{:.4e}, top-1 accuracy:{:.5f}, top-5 accuracy:{:.5f}, current energy:{:.4e}'.format(
val_loss, val_acc1,
val_acc5, cur_energy))
# save current model
model_snapshot(primal_model, os.path.join(args.logdir, 'primal_model_latest.pkl'))
plot_data.dump(os.path.join(args.logdir, 'plot_data.pkl'))
if args.save_interval > 0 and (epoch + 1) % args.save_interval == 0:
model_snapshot(primal_model, os.path.join(args.logdir, 'primal_model_epoch{}.pkl'.format(epoch)))
elapse_time = time.time() - t_begin
reduced_energy = energy_residue_ub - (cur_energy - args.budget)
if reduced_energy <= 0:
print('No energy reduced! Considering improve rho_z.')
else:
speed_energy = elapse_time / reduced_energy
eta = speed_energy * energy_residue_ub - elapse_time
print("Elapsed {:.2f}s, ets {:.2f}s".format(elapse_time, eta))
if cur_energy <= args.budget:
1] # get the index of the max log-probability
correct = pred.cpu().eq(indx_target).sum()
acc = correct * 1.0 / len(data)
print(
'Train Epoch: {} [{}/{}] Loss: {:.6f} Acc: {:.4f} lr: {:.2e}'
.format(epoch, batch_idx * len(data),
len(train_loader.dataset), loss.data[0], acc,
optimizer.param_groups[0]['lr']))
elapse_time = time.time() - t_begin
speed_epoch = elapse_time / (epoch + 1)
speed_batch = speed_epoch / len(train_loader)
eta = speed_epoch * args.epochs - elapse_time
print("Elapsed {:.2f}s, {:.2f} s/epoch, {:.2f} s/batch, ets {:.2f}s".
format(elapse_time, speed_epoch, speed_batch, eta))
misc.model_snapshot(model, os.path.join(args.logdir, 'latest.pth'))
if epoch % args.test_interval == 0:
with torch.no_grad():
model.eval()
test_loss = 0
correct = 0
for data, target in test_loader:
indx_target = target.clone()
if args.cuda:
data, target = data.cuda(), target.cuda(
) #.long().squeeze()
data, target = Variable(data), Variable(target)
output = model(data)
test_loss += F.cross_entropy(output, target).item()
pred = output.data.max(1)[
sum([
weight_bits_eval[i] * num_nnz_eval[i]
for i in range(n_conv_layers)
]) / full_model_size))
print('compression rate={:.4e} (train)'.format(
sum([weight_bits[i] * num_nnz[i]
for i in range(n_conv_layers)]) / full_model_size))
print('======================================================')
# remember best prec@1 and save checkpoint
if prec1 > best_acc:
print('find accuracy {:4f} > {:.4f}'.format(prec1, best_acc))
new_file = os.path.join(args.logdir,
'model_best-{}.pkl'.format(epoch))
misc.model_snapshot(model_eval,
new_file,
old_file=old_file,
verbose=True)
best_acc = prec1
old_file = new_file
if epoch > 0 and args.save_every > 0 and epoch % args.save_every == 0:
model_snapshot(
model,
os.path.join(args.logdir, 'model_epoch{}.pt'.format(epoch)))
# save the lastest model
model_snapshot(model, os.path.join(args.logdir, 'model_latest.pt'))
# print(tr_loss_hist)
# print(ts_loss_hist)
# print(residual_hist)