def main():
start_time = time.time()
# initialize output dir
init_out_dir()
# check point
if args.clear_checkpoint:
clear_checkpoint()
last_step = get_last_checkpoint_step()
if last_step >= 0:
my_log('\nCheckpoint found: {}\n'.format(last_step))
else:
clear_log()
print_args()
if args.net == 'pixelcnn_xy':
net = PixelCNN(**vars(args))
else:
raise ValueError('Unknown net: {}'.format(args.net))
net.to(args.device)
my_log('{}\n'.format(net))
# parameters of networks
params = list(net.parameters())
params = list(filter(lambda p: p.requires_grad,
params)) # parameters with gradients
nparams = int(sum([np.prod(p.shape) for p in params]))
my_log('Total number of trainable parameters: {}'.format(nparams))
named_params = list(net.named_parameters())
# optimizers
if args.optimizer == 'sgd':
optimizer = torch.optim.SGD(params, lr=args.lr)
elif args.optimizer == 'sgdm':
optimizer = torch.optim.SGD(params, lr=args.lr, momentum=0.9)
elif args.optimizer == 'rmsprop':
optimizer = torch.optim.RMSprop(params, lr=args.lr, alpha=0.99)
elif args.optimizer == 'adam':
optimizer = torch.optim.Adam(params, lr=args.lr, betas=(0.9, 0.999))
elif args.optimizer == 'adam0.5':
optimizer = torch.optim.Adam(params, lr=args.lr, betas=(0.5, 0.999))
else:
raise ValueError('Unknown optimizer: {}'.format(args.optimizer))
# learning rates
if args.lr_schedule:
# 0.92**80 ~ 1e-3
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,
factor=0.92,
patience=100,
threshold=1e-4,
min_lr=1e-6)
# read last step
if last_step >= 0:
state = torch.load('{}_save/{}.state'.format(args.out_filename,
last_step))
ignore_param(state['net'], net)
net.load_state_dict(state['net'])
if state.get('optimizer'):
optimizer.load_state_dict(state['optimizer'])
if args.lr_schedule and state.get('scheduler'):
scheduler.load_state_dict(state['scheduler'])
init_time = time.time() - start_time
my_log('init_time = {:.3f}'.format(init_time))
# start training
my_log('Training...')
sample_time = 0
train_time = 0
start_time = time.time()
for step in range(last_step + 1, args.max_step + 1):
optimizer.zero_grad() # clear last step
sample_start_time = time.time()
with torch.no_grad():
sample, x_hat = net.sample(
args.batch_size
) # sample from networks with batch_size = 10**3 (default)
assert not sample.requires_grad
assert not x_hat.requires_grad
sample_time += time.time() - sample_start_time
train_start_time = time.time()
# log probabilities
log_prob = net.log_prob(sample, args.batch_size)
# 0.998**9000 ~ 1e-8
beta = args.beta * (1 - args.beta_anneal**step
) # anneal process to avoid mode collapse
with torch.no_grad():
energy, vortices = xy.energy(sample, args.ham, args.lattice,
args.boundary)
loss = log_prob + beta * energy # construct loss function(free energy）from configurations
assert not energy.requires_grad
assert not loss.requires_grad
loss_reinforce = torch.mean((loss - loss.mean()) * log_prob)
loss_reinforce.backward() # back propagation
if args.clip_grad:
nn.utils.clip_grad_norm_(params, args.clip_grad)
optimizer.step()
if args.lr_schedule:
scheduler.step(loss.mean())
train_time += time.time() - train_start_time
# export physical observables
if args.print_step and step % args.print_step == 0:
free_energy_mean = loss.mean() / beta / (args.L**2
) # free energy density
free_energy_std = loss.std() / beta / (args.L**2)
entropy_mean = -log_prob.mean() / (args.L**2) # entropy density
energy_mean = (energy / (args.L**2)).mean() # energy density
energy_std = (energy / (args.L**2)).std()
vortices = vortices.mean() / args.L**2 # vortices density
# heat_capacity=(((energy/ (args.L**2))**2).mean()- ((energy/ (args.L**2)).mean())**2) *(beta**2)
# magnetization
# mag = torch.cos(sample).sum(dim=(2,3)).mean(dim=0) # M_x (M_x,M_y)=(cos(theta), sin(theta))
# mag_mean = mag.mean()
# mag_sqr_mean = (mag**2).mean()
# sus_mean = mag_sqr_mean/args.L**2
# log
if step > 0:
sample_time /= args.print_step
train_time /= args.print_step
used_time = time.time() - start_time
# hyperparameters in training
my_log(
'step = {}, lr = {:.3g}, loss={:.8g}, beta = {:.8g}, sample_time = {:.3f}, train_time = {:.3f}, used_time = {:.3f}'
.format(
step,
optimizer.param_groups[0]['lr'],
loss.mean(),
beta,
sample_time,
train_time,
used_time,
))
# observables
my_log(
'F = {:.8g}, F_std = {:.8g}, E = {:.8g}, E_std={:.8g}, v={:.8g}'
.format(
free_energy_mean.item(),
free_energy_std.item(),
energy_mean.item(),
energy_std.item(),
vortices.item(),
))
sample_time = 0
train_time = 0
# save sample
if args.save_sample and step % args.save_step == 0:
# save traning state
# state = {
# 'sample': sample,
# 'x_hat': x_hat,
# 'log_prob': log_prob,
# 'energy': energy,
# 'loss': loss,
# }
# torch.save(state, '{}_save/{}.sample'.format(
# args.out_filename, step))
# Recognize the Phase Transition
# helicity
with torch.no_grad():
correlations = helicity(sample)
helicity_modulus = -((energy / args.L**2).mean()) - (
args.beta * correlations**2 / args.L**2).mean()
my_log('Rho={:.8g}'.format(helicity_modulus.item()))
# save configurations
sample_array = sample.cpu().numpy()
np.savetxt(
'{}_save/sample{}.txt'.format(args.out_filename, step),
sample_array.reshape(args.batch_size, -1))
# save observables
np.savetxt(
'{}_save/results{}.csv'.format(args.out_filename, step), [
beta,
step,
free_energy_mean,
free_energy_std,
energy_mean,
energy_std,
vortices,
helicity_modulus,
])
# save net
if (args.out_filename and args.save_step
and step % args.save_step == 0):
state = {
'net': net.state_dict(),
'optimizer': optimizer.state_dict(),
}
if args.lr_schedule:
state['scheduler'] = scheduler.state_dict()
torch.save(state,
'{}_save/{}.state'.format(args.out_filename, step))
# visualization in each visual_step
if (args.out_filename and args.visual_step
and step % args.visual_step == 0):
# torchvision.utils.save_image(
# sample,
# '{}_img/{}.png'.format(args.out_filename, step),
# nrow=int(sqrt(sample.shape[0])),
# padding=0,
# normalize=True)
# print sample
if args.print_sample:
x_hat_alpha = x_hat[:, 0, :, :].view(x_hat.shape[0],
-1).cpu().numpy() # alpha
x_hat_std1 = np.std(x_hat_alpha, axis=0).reshape([args.L] * 2)
x_hat_beta = x_hat[:, 1, :, :].view(x_hat.shape[0],
-1).cpu().numpy() # beta
x_hat_std2 = np.std(x_hat_beta, axis=0).reshape([args.L] * 2)
energy_np = energy.cpu().numpy()
energy_count = np.stack(
np.unique(energy_np, return_counts=True)).T
my_log(
'\nsample\n{}\nalpha\n{}\nbeta\n{}\nlog_prob\n{}\nenergy\n{}\nloss\n{}\nalpha_std\n{}\nbeta_std\n{}\nenergy_count\n{}\n'
.format(
sample[:args.print_sample, 0],
x_hat[:args.print_sample, 0],
x_hat[:args.print_sample, 1],
log_prob[:args.print_sample],
energy[:args.print_sample],
loss[:args.print_sample],
x_hat_std1,
x_hat_std2,
energy_count,
))
# print gradient
if args.print_grad:
my_log('grad max_abs min_abs mean std')
for name, param in named_params:
if param.grad is not None:
grad = param.grad
grad_abs = torch.abs(grad)
my_log('{} {:.3g} {:.3g} {:.3g} {:.3g}'.format(
name,
torch.max(grad_abs).item(),
torch.min(grad_abs).item(),
torch.mean(grad).item(),
torch.std(grad).item(),
))
else:
my_log('{} None'.format(name))
my_log('')
def main():
start_time = time.time()
init_out_dir()
if args.clear_checkpoint:
clear_checkpoint()
last_step = get_last_checkpoint_step()
if last_step >= 0:
my_log('\nCheckpoint found: {}\n'.format(last_step))
else:
clear_log()
print_args()
if args.net == 'made':
net = MADE(**vars(args))
elif args.net == 'pixelcnn':
net = PixelCNN(**vars(args))
elif args.net == 'bernoulli':
net = BernoulliMixture(**vars(args))
else:
raise ValueError('Unknown net: {}'.format(args.net))
net.to(args.device)
my_log('{}\n'.format(net))
params = list(net.parameters())
params = list(filter(lambda p: p.requires_grad, params))
nparams = int(sum([np.prod(p.shape) for p in params]))
my_log('Total number of trainable parameters: {}'.format(nparams))
named_params = list(net.named_parameters())
if args.optimizer == 'sgd':
optimizer = torch.optim.SGD(params, lr=args.lr)
elif args.optimizer == 'sgdm':
optimizer = torch.optim.SGD(params, lr=args.lr, momentum=0.9)
elif args.optimizer == 'rmsprop':
optimizer = torch.optim.RMSprop(params, lr=args.lr, alpha=0.99)
elif args.optimizer == 'adam':
optimizer = torch.optim.Adam(params, lr=args.lr, betas=(0.9, 0.999))
elif args.optimizer == 'adam0.5':
optimizer = torch.optim.Adam(params, lr=args.lr, betas=(0.5, 0.999))
else:
raise ValueError('Unknown optimizer: {}'.format(args.optimizer))
if args.lr_schedule:
# 0.92**80 ~ 1e-3
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
optimizer, factor=0.92, patience=100, threshold=1e-4, min_lr=1e-6)
if last_step >= 0:
state = torch.load('{}_save/{}.state'.format(args.out_filename,
last_step))
ignore_param(state['net'], net)
net.load_state_dict(state['net'])
if state.get('optimizer'):
optimizer.load_state_dict(state['optimizer'])
if args.lr_schedule and state.get('scheduler'):
scheduler.load_state_dict(state['scheduler'])
init_time = time.time() - start_time
my_log('init_time = {:.3f}'.format(init_time))
my_log('Training...')
sample_time = 0
train_time = 0
start_time = time.time()
for step in range(last_step + 1, args.max_step + 1):
optimizer.zero_grad()
sample_start_time = time.time()
with torch.no_grad():
sample, x_hat = net.sample(args.batch_size)
assert not sample.requires_grad
assert not x_hat.requires_grad
sample_time += time.time() - sample_start_time
train_start_time = time.time()
log_prob = net.log_prob(sample)
# 0.998**9000 ~ 1e-8
beta = args.beta * (1 - args.beta_anneal**step)
with torch.no_grad():
energy = ising.energy(sample, args.ham, args.lattice,
args.boundary)
loss = log_prob + beta * energy
assert not energy.requires_grad
assert not loss.requires_grad
loss_reinforce = torch.mean((loss - loss.mean()) * log_prob)
loss_reinforce.backward()
if args.clip_grad:
nn.utils.clip_grad_norm_(params, args.clip_grad)
optimizer.step()
if args.lr_schedule:
scheduler.step(loss.mean())
train_time += time.time() - train_start_time
if args.print_step and step % args.print_step == 0:
free_energy_mean = loss.mean() / args.beta / args.L**2
free_energy_std = loss.std() / args.beta / args.L**2
entropy_mean = -log_prob.mean() / args.L**2
energy_mean = energy.mean() / args.L**2
mag = sample.mean(dim=0)
mag_mean = mag.mean()
mag_sqr_mean = (mag**2).mean()
if step > 0:
sample_time /= args.print_step
train_time /= args.print_step
used_time = time.time() - start_time
my_log(
'step = {}, F = {:.8g}, F_std = {:.8g}, S = {:.8g}, E = {:.8g}, M = {:.8g}, Q = {:.8g}, lr = {:.3g}, beta = {:.8g}, sample_time = {:.3f}, train_time = {:.3f}, used_time = {:.3f}'
.format(
step,
free_energy_mean.item(),
free_energy_std.item(),
entropy_mean.item(),
energy_mean.item(),
mag_mean.item(),
mag_sqr_mean.item(),
optimizer.param_groups[0]['lr'],
beta,
sample_time,
train_time,
used_time,
))
sample_time = 0
train_time = 0
if args.save_sample:
state = {
'sample': sample,
'x_hat': x_hat,
'log_prob': log_prob,
'energy': energy,
'loss': loss,
}
torch.save(state, '{}_save/{}.sample'.format(
args.out_filename, step))
if (args.out_filename and args.save_step
and step % args.save_step == 0):
state = {
'net': net.state_dict(),
'optimizer': optimizer.state_dict(),
}
if args.lr_schedule:
state['scheduler'] = scheduler.state_dict()
torch.save(state, '{}_save/{}.state'.format(
args.out_filename, step))
if (args.out_filename and args.visual_step
and step % args.visual_step == 0):
torchvision.utils.save_image(
sample,
'{}_img/{}.png'.format(args.out_filename, step),
nrow=int(sqrt(sample.shape[0])),
padding=0,
normalize=True)
if args.print_sample:
x_hat_np = x_hat.view(x_hat.shape[0], -1).cpu().numpy()
x_hat_std = np.std(x_hat_np, axis=0).reshape([args.L] * 2)
x_hat_cov = np.cov(x_hat_np.T)
x_hat_cov_diag = np.diag(x_hat_cov)
x_hat_corr = x_hat_cov / (
sqrt(x_hat_cov_diag[:, None] * x_hat_cov_diag[None, :]) +
args.epsilon)
x_hat_corr = np.tril(x_hat_corr, -1)
x_hat_corr = np.max(np.abs(x_hat_corr), axis=1)
x_hat_corr = x_hat_corr.reshape([args.L] * 2)
energy_np = energy.cpu().numpy()
energy_count = np.stack(
np.unique(energy_np, return_counts=True)).T
my_log(
'\nsample\n{}\nx_hat\n{}\nlog_prob\n{}\nenergy\n{}\nloss\n{}\nx_hat_std\n{}\nx_hat_corr\n{}\nenergy_count\n{}\n'
.format(
sample[:args.print_sample, 0],
x_hat[:args.print_sample, 0],
log_prob[:args.print_sample],
energy[:args.print_sample],
loss[:args.print_sample],
x_hat_std,
x_hat_corr,
energy_count,
))
if args.print_grad:
my_log('grad max_abs min_abs mean std')
for name, param in named_params:
if param.grad is not None:
grad = param.grad
grad_abs = torch.abs(grad)
my_log('{} {:.3g} {:.3g} {:.3g} {:.3g}'.format(
name,
torch.max(grad_abs).item(),
torch.min(grad_abs).item(),
torch.mean(grad).item(),
torch.std(grad).item(),
))
else:
my_log('{} None'.format(name))
my_log('')
return x_ufloat
if __name__ == '__main__':
if args.net == 'made':
net = MADE(**vars(args))
elif args.net == 'pixelcnn':
net = PixelCNN(**vars(args))
else:
raise ValueError('Unknown net: {}'.format(args.net))
net.to(args.device)
print('{}\n'.format(net))
print(state_filename)
state = torch.load(state_filename, map_location=args.device)
ignore_param(state['net'], net)
net.load_state_dict(state['net'])
F_sum = 0
F_sqr_sum = 0
S_sum = 0
S_sqr_sum = 0
E_sum = 0
E_sqr_sum = 0
start_time = time.time()
for step in range(args.max_step):
with torch.no_grad():
sample, x_hat = net.sample(args.batch_size)
log_prob = net._log_prob(sample, x_hat)
energy = ising.energy(sample, args.model, args.lattice,
args.boundary) / args.L**2