def validate(vae_name, params, val_loader, vae_param):
cuda = params["cuda"] and torch.cuda.is_available()
torch.manual_seed(params["seed"])
device = torch.device("cuda" if cuda else "cpu")
model_path = os.path.join(params["vae_dir"], vae_name)
vae_file = os.path.join(model_path, 'vae', 'best.tar')
assert os.path.exists(vae_file), "VAE Checkpoint does not exist."
state = torch.load(vae_file)
print("Loading VAE at epoch {} "
"with test error {}".format(state['epoch'], state['precision']))
print(str(vae_name))
model = VAE(nc=3,
ngf=params["img_size"],
ndf=params["img_size"],
latent_variable_size=vae_param["latent_size"],
cuda=cuda).to(device)
model.load_state_dict(state['state_dict'])
model.eval()
avg_psnr = 0
avg_ms_ssim = 0
index = 0
with torch.no_grad():
val_loader.dataset.load_next_buffer()
for i, data in enumerate(val_loader):
data = data.to(device)
recon_batch, mu, logvar = model(data)
mse = F.mse_loss(recon_batch, data)
psnr = 10 * log10(1 / mse.item())
avg_psnr += psnr
#print("PSNR:", str(psnr))
ms_ssim = msssim(recon_batch, data).item()
avg_ms_ssim += ms_ssim
#print("MS-SSIM:", str(ms_ssim))
if index < 10:
n = min(data.size(0), 10)
comparison = torch.cat([
data[:n],
recon_batch.view(params["batch_size"], 3,
params["img_size"],
params["img_size"])[:n]
])
save_image(comparison.cpu(),
os.path.join(
params["report_dir"],
str(vae_name) + "_" + str(index) + '.png'),
nrow=n)
index += 1
step = len(val_loader.dataset) / params["batch_size"]
avg_ms_ssim /= step
avg_psnr /= step
print("AVG PSNR", str(avg_psnr))
print("AVG MS-SSIM", str(avg_ms_ssim))
print("index", str(index))
print("step", str(step))
return [avg_psnr, avg_ms_ssim]
class MFEC:
def __init__(self, env, args, device='cpu'):
"""
Instantiate an MFEC Agent
----------
env: gym.Env
gym environment to train on
args: args class from argparser
args are from from train.py: see train.py for help with each arg
device: string
'cpu' or 'cuda:0' depending on use_cuda flag from train.py
"""
self.environment_type = args.environment_type
self.env = env
self.actions = range(self.env.action_space.n)
self.frames_to_stack = args.frames_to_stack
self.Q_train_algo = args.Q_train_algo
self.use_Q_max = args.use_Q_max
self.force_knn = args.force_knn
self.weight_neighbors = args.weight_neighbors
self.delta = args.delta
self.device = device
self.rs = np.random.RandomState(args.seed)
# Hyperparameters
self.epsilon = args.initial_epsilon
self.final_epsilon = args.final_epsilon
self.epsilon_decay = args.epsilon_decay
self.gamma = args.gamma
self.lr = args.lr
self.q_lr = args.q_lr
# Autoencoder for state embedding network
self.vae_batch_size = args.vae_batch_size # batch size for training VAE
self.vae_epochs = args.vae_epochs # number of epochs to run VAE
self.embedding_type = args.embedding_type
self.SR_embedding_type = args.SR_embedding_type
self.embedding_size = args.embedding_size
self.in_height = args.in_height
self.in_width = args.in_width
if self.embedding_type == 'VAE':
self.vae_train_frames = args.vae_train_frames
self.vae_loss = VAELoss()
self.vae_print_every = args.vae_print_every
self.load_vae_from = args.load_vae_from
self.vae_weights_file = args.vae_weights_file
self.vae = VAE(self.frames_to_stack, self.embedding_size,
self.in_height, self.in_width)
self.vae = self.vae.to(self.device)
self.optimizer = get_optimizer(args.optimizer,
self.vae.parameters(), self.lr)
elif self.embedding_type == 'random':
self.projection = self.rs.randn(
self.embedding_size, self.in_height * self.in_width *
self.frames_to_stack).astype(np.float32)
elif self.embedding_type == 'SR':
self.SR_train_algo = args.SR_train_algo
self.SR_gamma = args.SR_gamma
self.SR_epochs = args.SR_epochs
self.SR_batch_size = args.SR_batch_size
self.n_hidden = args.n_hidden
self.SR_train_frames = args.SR_train_frames
self.SR_filename = args.SR_filename
if self.SR_embedding_type == 'random':
self.projection = np.random.randn(
self.embedding_size,
self.in_height * self.in_width).astype(np.float32)
if self.SR_train_algo == 'TD':
self.mlp = MLP(self.embedding_size, self.n_hidden)
self.mlp = self.mlp.to(self.device)
self.loss_fn = nn.MSELoss(reduction='mean')
params = self.mlp.parameters()
self.optimizer = get_optimizer(args.optimizer, params,
self.lr)
# QEC
self.max_memory = args.max_memory
self.num_neighbors = args.num_neighbors
self.qec = QEC(self.actions, self.max_memory, self.num_neighbors,
self.use_Q_max, self.force_knn, self.weight_neighbors,
self.delta, self.q_lr)
#self.state = np.empty(self.embedding_size, self.projection.dtype)
#self.action = int
self.memory = []
self.print_every = args.print_every
self.episodes = 0
def choose_action(self, values):
"""
Choose epsilon-greedy policy according to Q-estimates
"""
# Exploration
if self.rs.random_sample() < self.epsilon:
self.action = self.rs.choice(self.actions)
# Exploitation
else:
best_actions = np.argwhere(values == np.max(values)).flatten()
self.action = self.rs.choice(best_actions)
return self.action
def TD_update(self, prev_embedding, prev_action, reward, values, time):
# On-policy value estimate of current state (epsiloln-greedy)
# Expected Sarsa
v_t = (1 -
self.epsilon) * np.max(values) + self.epsilon * np.mean(values)
value = reward + self.gamma * v_t
self.qec.update(prev_embedding, prev_action, value, time - 1)
def MC_update(self):
value = 0.0
for _ in range(len(self.memory)):
experience = self.memory.pop()
value = value * self.gamma + experience["reward"]
self.qec.update(
experience["state"],
experience["action"],
value,
experience["time"],
)
def add_to_memory(self, state_embedding, action, reward, time):
self.memory.append({
"state": state_embedding,
"action": action,
"reward": reward,
"time": time,
})
def run_episode(self):
"""
Train an MFEC agent for a single episode:
Interact with environment
Perform update
"""
self.episodes += 1
RENDER_SPEED = 0.04
RENDER = False
episode_frames = 0
total_reward = 0
total_steps = 0
# Update epsilon
if self.epsilon > self.final_epsilon:
self.epsilon = self.epsilon * self.epsilon_decay
#self.env.seed(random.randint(0, 1000000))
state = self.env.reset()
if self.environment_type == 'fourrooms':
fewest_steps = self.env.shortest_path_length(self.env.state)
done = False
time = 0
while not done:
time += 1
if self.embedding_type == 'random':
state = np.array(state).flatten()
state_embedding = np.dot(self.projection, state)
elif self.embedding_type == 'VAE':
state = torch.tensor(state).permute(2, 0, 1) #(H,W,C)->(C,H,W)
state = state.unsqueeze(0).to(self.device)
with torch.no_grad():
mu, logvar = self.vae.encoder(state)
state_embedding = torch.cat([mu, logvar], 1)
state_embedding = state_embedding.squeeze()
state_embedding = state_embedding.cpu().numpy()
elif self.embedding_type == 'SR':
if self.SR_train_algo == 'TD':
state = np.array(state).flatten()
state_embedding = np.dot(self.projection, state)
with torch.no_grad():
state_embedding = self.mlp(
torch.tensor(state_embedding)).cpu().numpy()
elif self.SR_train_algo == 'DP':
s = self.env.state
state_embedding = self.true_SR_dict[s]
state_embedding = state_embedding / np.linalg.norm(state_embedding)
if RENDER:
self.env.render()
time.sleep(RENDER_SPEED)
# Get estimated value of each action
values = [
self.qec.estimate(state_embedding, action)
for action in self.actions
]
action = self.choose_action(values)
state, reward, done, _ = self.env.step(action)
if self.Q_train_algo == 'MC':
self.add_to_memory(state_embedding, action, reward, time)
elif self.Q_train_algo == 'TD':
if time > 1:
self.TD_update(prev_embedding, prev_action, prev_reward,
values, time)
prev_reward = reward
prev_embedding = state_embedding
prev_action = action
total_reward += reward
total_steps += 1
episode_frames += self.env.skip
if self.Q_train_algo == 'MC':
self.MC_update()
if self.episodes % self.print_every == 0:
print("KNN usage:", np.mean(self.qec.knn_usage))
self.qec.knn_usage = []
print("Proportion of replace:", np.mean(self.qec.replace_usage))
self.qec.replace_usage = []
if self.environment_type == 'fourrooms':
n_extra_steps = total_steps - fewest_steps
return n_extra_steps, episode_frames, total_reward
else:
return episode_frames, total_reward
def warmup(self):
"""
Collect 1 million frames from random policy and train VAE
"""
if self.embedding_type == 'VAE':
if self.load_vae_from is not None:
self.vae.load_state_dict(torch.load(self.load_vae_from))
self.vae = self.vae.to(self.device)
else:
# Collect 1 million frames from random policy
print("Generating dataset to train VAE from random policy")
vae_data = []
state = self.env.reset()
total_frames = 0
while total_frames < self.vae_train_frames:
action = random.randint(0, self.env.action_space.n - 1)
state, reward, done, _ = self.env.step(action)
vae_data.append(state)
total_frames += self.env.skip
if done:
state = self.env.reset()
# Dataset, Dataloader for 1 million frames
vae_data = torch.tensor(
vae_data
) # (N x H x W x C) - (1mill/skip X 84 X 84 X frames_to_stack)
vae_data = vae_data.permute(0, 3, 1, 2) # (N x C x H x W)
vae_dataset = TensorDataset(vae_data)
vae_dataloader = DataLoader(vae_dataset,
batch_size=self.vae_batch_size,
shuffle=True)
# Training loop
print("Training VAE")
self.vae.train()
for epoch in range(self.vae_epochs):
train_loss = 0
for batch_idx, batch in enumerate(vae_dataloader):
batch = batch[0].to(self.device)
self.optimizer.zero_grad()
recon_batch, mu, logvar = self.vae(batch)
loss = self.vae_loss(recon_batch, batch, mu, logvar)
train_loss += loss.item()
loss.backward()
self.optimizer.step()
if batch_idx % self.vae_print_every == 0:
msg = 'VAE Epoch: {} [{}/{}]\tLoss: {:.6f}'.format(
epoch, batch_idx * len(batch),
len(vae_dataloader.dataset),
loss.item() / len(batch))
print(msg)
print('====> Epoch {} Average loss: {:.4f}'.format(
epoch, train_loss / len(vae_dataloader.dataset)))
if self.vae_weights_file is not None:
torch.save(self.vae.state_dict(),
self.vae_weights_file)
self.vae.eval()
elif self.embedding_type == 'SR':
if self.SR_embedding_type == 'random':
if self.SR_train_algo == 'TD':
total_frames = 0
transitions = []
while total_frames < self.SR_train_frames:
observation = self.env.reset()
s_t = self.env.state # will not work on Atari
done = False
while not done:
action = np.random.randint(self.env.action_space.n)
observation, reward, done, _ = self.env.step(
action)
s_tp1 = self.env.state # will not work on Atari
transitions.append((s_t, s_tp1))
total_frames += self.env.skip
s_t = s_tp1
# Dataset, Dataloader
dataset = SRDataset(self.env, self.projection, transitions)
dataloader = DataLoader(dataset,
batch_size=self.SR_batch_size,
shuffle=True)
train_losses = []
#Training loop
for epoch in range(self.SR_epochs):
for batch_idx, batch in enumerate(dataloader):
self.optimizer.zero_grad()
e_t, e_tp1 = batch
e_t = e_t.to(self.device)
e_tp1 = e_tp1.to(self.device)
mhat_t = self.mlp(e_t)
mhat_tp1 = self.mlp(e_tp1)
target = e_t + self.gamma * mhat_tp1.detach()
loss = self.loss_fn(mhat_t, target)
loss.backward()
self.optimizer.step()
train_losses.append(loss.item())
print("Epoch:", epoch, "Average loss",
np.mean(train_losses))
emb_reps = np.zeros(
[self.env.n_states, self.embedding_size])
SR_reps = np.zeros(
[self.env.n_states, self.embedding_size])
labels = []
room_size = self.env.room_size
for i, (state,
obs) in enumerate(self.env.state_dict.items()):
emb = np.dot(self.projection, obs.flatten())
emb_reps[i, :] = emb
with torch.no_grad():
emb = torch.tensor(emb).to(self.device)
SR = self.mlp(emb).cpu().numpy()
SR_reps[i, :] = SR
if state[0] < room_size + 1 and state[
1] < room_size + 1:
label = 0
elif state[0] > room_size + 1 and state[
1] < room_size + 1:
label = 1
elif state[0] < room_size + 1 and state[
1] > room_size + 1:
label = 2
elif state[0] > room_size + 1 and state[
1] > room_size + 1:
label = 3
else:
label = 4
labels.append(label)
np.save('%s_SR_reps.npy' % (self.SR_filename), SR_reps)
np.save('%s_emb_reps.npy' % (self.SR_filename), emb_reps)
np.save('%s_labels.npy' % (self.SR_filename), labels)
elif self.SR_train_algo == 'MC':
pass
elif self.SR_train_algo == 'DP':
# Use this to ensure same order every time
idx_to_state = {
i: state
for i, state in enumerate(self.env.state_dict.keys())
}
state_to_idx = {v: k for k, v in idx_to_state.items()}
T = np.zeros([self.env.n_states, self.env.n_states])
for i, s in idx_to_state.items():
for a in range(4):
self.env.state = s
_, _, _, _ = self.env.step(a)
s_tp1 = self.env.state
T[state_to_idx[s], state_to_idx[s_tp1]] += 0.25
true_SR = np.eye(self.env.n_states)
done = False
t = 0
while not done:
t += 1
new_SR = true_SR + (self.SR_gamma**t) * (np.matmul(
true_SR, T))
done = np.max(np.abs(true_SR - new_SR)) < 1e-10
true_SR = new_SR
self.true_SR_dict = {}
for s, obs in self.env.state_dict.items():
idx = state_to_idx[s]
self.true_SR_dict[s] = true_SR[idx, :]
else:
pass # random projection doesn't require warmup
class VAE_TRAINER():
def __init__(self, params):
self.params = params
self.loss_function = {
'ms-ssim': ms_ssim_loss,
'mse': mse_loss,
'mix': mix_loss
}[params["loss"]]
# Choose device
self.cuda = params["cuda"] and torch.cuda.is_available()
torch.manual_seed(params["seed"])
# Fix numeric divergence due to bug in Cudnn
torch.backends.cudnn.benchmark = True
self.device = torch.device("cuda" if self.cuda else "cpu")
# Prepare data transformations
red_size = params["img_size"]
transform_train = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize((red_size, red_size)),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
])
transform_val = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize((red_size, red_size)),
transforms.ToTensor(),
])
# Initialize Data loaders
op_dataset = RolloutObservationDataset(params["path_data"],
transform_train,
train=True)
val_dataset = RolloutObservationDataset(params["path_data"],
transform_val,
train=False)
self.train_loader = torch.utils.data.DataLoader(
op_dataset,
batch_size=params["batch_size"],
shuffle=True,
num_workers=0)
self.eval_loader = torch.utils.data.DataLoader(
val_dataset,
batch_size=params["batch_size"],
shuffle=False,
num_workers=0)
# Initialize model and hyperparams
self.model = VAE(nc=3,
ngf=64,
ndf=64,
latent_variable_size=params["latent_size"],
cuda=self.cuda).to(self.device)
self.optimizer = optim.Adam(self.model.parameters())
self.init_vae_model()
self.visualize = params["visualize"]
if self.visualize:
self.plotter = VisdomLinePlotter(env_name=params['env'])
self.img_plotter = VisdomImagePlotter(env_name=params['env'])
self.alpha = params["alpha"] if params["alpha"] else 1.0
def train(self, epoch):
self.model.train()
# dataset_train.load_next_buffer()
mse_loss = 0
ssim_loss = 0
train_loss = 0
# Train step
for batch_idx, data in enumerate(self.train_loader):
data = data.to(self.device)
self.optimizer.zero_grad()
recon_batch, mu, logvar = self.model(data)
loss, mse, ssim = self.loss_function(recon_batch, data, mu, logvar,
self.alpha)
loss.backward()
train_loss += loss.item()
ssim_loss += ssim
mse_loss += mse
self.optimizer.step()
if batch_idx % params["log_interval"] == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
epoch, batch_idx * len(data),
len(self.train_loader.dataset),
100. * batch_idx / len(self.train_loader), loss.item()))
print('MSE: {} , SSIM: {:.4f}'.format(mse, ssim))
step = len(self.train_loader.dataset) / float(
self.params["batch_size"])
mean_train_loss = train_loss / step
mean_ssim_loss = ssim_loss / step
mean_mse_loss = mse_loss / step
print('-- Epoch: {} Average loss: {:.4f}'.format(
epoch, mean_train_loss))
print('-- Average MSE: {:.5f} Average SSIM: {:.4f}'.format(
mean_mse_loss, mean_ssim_loss))
if self.visualize:
self.plotter.plot('loss', 'train', 'VAE Train Loss', epoch,
mean_train_loss)
return
def eval(self):
self.model.eval()
# dataset_test.load_next_buffer()
eval_loss = 0
mse_loss = 0
ssim_loss = 0
vis = True
with torch.no_grad():
# Eval step
for data in self.eval_loader:
data = data.to(self.device)
recon_batch, mu, logvar = self.model(data)
loss, mse, ssim = self.loss_function(recon_batch, data, mu,
logvar, self.alpha)
eval_loss += loss.item()
ssim_loss += ssim
mse_loss += mse
if vis:
org_title = "Epoch: " + str(epoch)
comparison1 = torch.cat([
data[:4],
recon_batch.view(params["batch_size"], 3,
params["img_size"],
params["img_size"])[:4]
])
if self.visualize:
self.img_plotter.plot(comparison1, org_title)
vis = False
step = len(self.eval_loader.dataset) / float(params["batch_size"])
mean_eval_loss = eval_loss / step
mean_ssim_loss = ssim_loss / step
mean_mse_loss = mse_loss / step
print('-- Eval set loss: {:.4f}'.format(mean_eval_loss))
print('-- Eval MSE: {:.5f} Eval SSIM: {:.4f}'.format(
mean_mse_loss, mean_ssim_loss))
if self.visualize:
self.plotter.plot('loss', 'eval', 'VAE Eval Loss', epoch,
mean_eval_loss)
self.plotter.plot('loss', 'mse train', 'VAE MSE Loss', epoch,
mean_mse_loss)
self.plotter.plot('loss', 'ssim train', 'VAE MSE Loss', epoch,
mean_ssim_loss)
return mean_eval_loss
def init_vae_model(self):
self.vae_dir = os.path.join(self.params["logdir"], 'vae')
check_dir(self.vae_dir, 'samples')
if not self.params["noreload"]: # and os.path.exists(reload_file):
reload_file = os.path.join(self.params["vae_location"], 'best.tar')
state = torch.load(reload_file)
print("Reloading model at epoch {}"
", with eval error {}".format(state['epoch'],
state['precision']))
self.model.load_state_dict(state['state_dict'])
self.optimizer.load_state_dict(state['optimizer'])
def checkpoint(self, cur_best, eval_loss):
# Save the best and last checkpoint
best_filename = os.path.join(self.vae_dir, 'best.tar')
filename = os.path.join(self.vae_dir, 'checkpoint.tar')
is_best = not cur_best or eval_loss < cur_best
if is_best:
cur_best = eval_loss
save_checkpoint(
{
'epoch': epoch,
'state_dict': self.model.state_dict(),
'precision': eval_loss,
'optimizer': self.optimizer.state_dict()
}, is_best, filename, best_filename)
return cur_best
def plot(self, train, eval, epochs):
plt.plot(epochs, train, label="train loss")
plt.plot(epochs, eval, label="eval loss")
plt.legend()
plt.grid()
plt.savefig(self.params["logdir"] + "/vae_training_curve.png")
plt.close()
qsms = (qsms.to(device, dtype=torch.float) + trans) * scale
masks = masks.to(device, dtype=torch.float)
qsms = qsms * masks
recon_loss, kl_loss = vae_train(model=vae3d, optimizer=optimizer, x=qsms, mask=masks)
recon_loss_sum += recon_loss
kl_loss_sum += kl_loss
gen_iterations += 1
time.sleep(1)
scheduler.step(epoch)
# validation phase
vae3d.eval()
loss_total = 0
idx = 0
with torch.no_grad(): # to solve memory exploration issue
for idx, (rdfs, masks, weights, qsms) in enumerate(valLoader):
idx += 1
qsms = (qsms.to(device, dtype=torch.float) + trans) * scale
masks = masks.to(device, dtype=torch.float)
qsms = qsms * masks
x_mu, x_var, z_mu, z_logvar = vae3d(qsms)
x_factor = torch.prod(torch.tensor(x_mu.size()))
z_factor = torch.prod(torch.tensor(z_mu.size()))
# recon_loss = 0.5 * torch.sum((x_mu*masks - qsms*masks)**2 / x_var + torch.log(x_var)*masks)
recon_loss = 0.5 * torch.sum((x_mu - qsms)**2 / x_var + torch.log(x_var)) / x_factor
# recon_loss = torch.sum((x_mu - qsms)**2) / x_factor
writer.add_scalar(tag='loss/test',
scalar_value=loss,
global_step=i)
likelihood_x[i * batch_size:(i + 1) * batch_size] = logsumexp(
losses, axis=1) - np.log(number_samples)
return np.mean(likelihood_x)
if __name__ == '__main__':
from loaders.load_funtions import load_MNIST
from models.VAE import VAE
import pathlib
_, loader, _, dataset_type = load_MNIST('../datasets/')
output_dit = pathlib.Path('../outputs/')
input_shape = (1, 28, 28)
model = VAE(dimension_latent_space=50,
input_shape=input_shape,
dataset_type=dataset_type)
model.load_state_dict(
torch.load('../outputs/trained/mnist_bin_standard_50/model',
map_location='cpu'))
model.eval()
print(model.calculate_likelihood(loader, 100))
