Adding model
This commit is contained in:
parent
3fa5e8e74a
commit
e7f5533ee6
478
DPI/train.py
478
DPI/train.py
@ -6,11 +6,12 @@ import wandb
|
||||
import random
|
||||
import argparse
|
||||
import numpy as np
|
||||
from collections import OrderedDict
|
||||
|
||||
import utils
|
||||
from utils import ReplayBuffer, FreezeParameters, make_env, preprocess_obs, soft_update_params, save_image
|
||||
from replay_buffer import ReplayBuffer
|
||||
from models import ObservationEncoder, ObservationDecoder, TransitionModel, Actor, ValueModel, RewardModel, ProjectionHead, ContrastiveHead, CLUBSample
|
||||
from logger import Logger
|
||||
from video import VideoRecorder
|
||||
from dmc2gym.wrappers import set_global_var
|
||||
|
||||
@ -40,24 +41,25 @@ def parse_args():
|
||||
parser.add_argument('--resource_files', type=str)
|
||||
parser.add_argument('--eval_resource_files', type=str)
|
||||
parser.add_argument('--img_source', default=None, type=str, choices=['color', 'noise', 'images', 'video', 'none'])
|
||||
parser.add_argument('--total_frames', default=1000, type=int) # 10000
|
||||
parser.add_argument('--total_frames', default=5000, type=int) # 10000
|
||||
parser.add_argument('--high_noise', action='store_true')
|
||||
# replay buffer
|
||||
parser.add_argument('--replay_buffer_capacity', default=50000, type=int) #50000
|
||||
parser.add_argument('--episode_length', default=21, type=int)
|
||||
parser.add_argument('--episode_length', default=51, type=int)
|
||||
# train
|
||||
parser.add_argument('--agent', default='dpi', type=str, choices=['baseline', 'bisim', 'deepmdp', 'db', 'dpi', 'rpc'])
|
||||
parser.add_argument('--init_steps', default=10000, type=int)
|
||||
parser.add_argument('--num_train_steps', default=100000, type=int)
|
||||
parser.add_argument('--batch_size', default=128, type=int) #512
|
||||
parser.add_argument('--state_size', default=30, type=int)
|
||||
parser.add_argument('--hidden_size', default=256, type=int)
|
||||
parser.add_argument('--update_steps', default=1, type=int)
|
||||
parser.add_argument('--batch_size', default=64, type=int) #512
|
||||
parser.add_argument('--state_size', default=50, type=int)
|
||||
parser.add_argument('--hidden_size', default=512, type=int)
|
||||
parser.add_argument('--history_size', default=128, type=int)
|
||||
parser.add_argument('--episode_collection', default=5, type=int)
|
||||
parser.add_argument('--episodes_buffer', default=20, type=int)
|
||||
parser.add_argument('--episodes_buffer', default=5, type=int, help='Initial number of episodes to store in the buffer')
|
||||
parser.add_argument('--num-units', type=int, default=50, help='num hidden units for reward/value/discount models')
|
||||
parser.add_argument('--load_encoder', default=None, type=str)
|
||||
parser.add_argument('--imagine_horizon', default=10, type=str)
|
||||
parser.add_argument('--imagine_horizon', default=15, type=str)
|
||||
parser.add_argument('--grad_clip_norm', type=float, default=100.0, help='Gradient clipping norm')
|
||||
# eval
|
||||
parser.add_argument('--eval_freq', default=10, type=int) # TODO: master had 10000
|
||||
@ -65,8 +67,6 @@ def parse_args():
|
||||
parser.add_argument('--evaluation_interval', default=10000, type=int) # TODO: master had 10000
|
||||
# value
|
||||
parser.add_argument('--value_lr', default=8e-6, type=float)
|
||||
parser.add_argument('--value_beta', default=0.9, type=float)
|
||||
parser.add_argument('--value_tau', default=0.005, type=float)
|
||||
parser.add_argument('--value_target_update_freq', default=100, type=int)
|
||||
parser.add_argument('--td_lambda', default=0.95, type=int)
|
||||
# actor
|
||||
@ -78,13 +78,13 @@ def parse_args():
|
||||
# world/encoder/decoder
|
||||
parser.add_argument('--encoder_type', default='pixel', type=str, choices=['pixel', 'pixelCarla096', 'pixelCarla098', 'identity'])
|
||||
parser.add_argument('--world_model_lr', default=1e-5, type=float)
|
||||
parser.add_argument('--encoder_tau', default=0.005, type=float)
|
||||
parser.add_argument('--encoder_tau', default=0.001 , type=float)
|
||||
parser.add_argument('--decoder_type', default='pixel', type=str, choices=['pixel', 'identity', 'contrastive', 'reward', 'inverse', 'reconstruction'])
|
||||
parser.add_argument('--num_layers', default=4, type=int)
|
||||
parser.add_argument('--num_filters', default=32, type=int)
|
||||
parser.add_argument('--aug', action='store_true')
|
||||
# sac
|
||||
parser.add_argument('--discount', default=0.99, type=float)
|
||||
parser.add_argument('--discount', default=0.95, type=float)
|
||||
# misc
|
||||
parser.add_argument('--seed', default=1, type=int)
|
||||
parser.add_argument('--logging_freq', default=100, type=int)
|
||||
@ -131,6 +131,7 @@ class DPI:
|
||||
self.env = utils.FrameStack(self.env, k=self.args.frame_stack)
|
||||
self.env = utils.ActionRepeat(self.env, self.args.action_repeat)
|
||||
self.env = utils.NormalizeActions(self.env)
|
||||
self.env = utils.TimeLimit(self.env, 1000 / args.action_repeat)
|
||||
|
||||
# create replay buffer
|
||||
self.data_buffer = ReplayBuffer(size=self.args.replay_buffer_capacity,
|
||||
@ -250,7 +251,7 @@ class DPI:
|
||||
self.obs_decoder.load_state_dict(torch.load(os.path.join(saved_model_dir, 'obs_decoder.pt')))
|
||||
self.transition_model.load_state_dict(torch.load(os.path.join(saved_model_dir, 'transition_model.pt')))
|
||||
|
||||
def collect_sequences(self, episodes, random=True, actor_model=None, encoder_model=None):
|
||||
def collect_random_sequences(self, episodes):
|
||||
obs = self.env.reset()
|
||||
done = False
|
||||
|
||||
@ -259,219 +260,104 @@ class DPI:
|
||||
self.global_episodes += 1
|
||||
epi_reward = 0
|
||||
while not done:
|
||||
if random:
|
||||
action = self.env.action_space.sample()
|
||||
else:
|
||||
with torch.no_grad():
|
||||
obs = torch.tensor(obs.copy(), dtype=torch.float32).unsqueeze(0)
|
||||
obs_processed = preprocess_obs(obs).to(device)
|
||||
state = self.obs_encoder(obs_processed)["distribution"].sample()
|
||||
action = self.actor_model(state).cpu().numpy().squeeze()
|
||||
#action = self.env.action_space.sample()
|
||||
|
||||
next_obs, rew, done, _ = self.env.step(action)
|
||||
self.data_buffer.add(obs, action, next_obs, rew, done, self.global_episodes)
|
||||
obs = next_obs
|
||||
epi_reward += rew
|
||||
obs = self.env.reset()
|
||||
done=False
|
||||
all_rews.append(epi_reward)
|
||||
return all_rews
|
||||
|
||||
def collect_sequences(self, episodes, actor_model):
|
||||
obs = self.env.reset()
|
||||
done = False
|
||||
all_rews = []
|
||||
for episode_count in tqdm.tqdm(range(episodes), desc='Collecting episodes'):
|
||||
self.global_episodes += 1
|
||||
epi_reward = 0
|
||||
while not done:
|
||||
with torch.no_grad():
|
||||
obs = torch.tensor(obs.copy(), dtype=torch.float32).to(device).unsqueeze(0)
|
||||
state = self.get_features(obs)["distribution"].rsample()
|
||||
action = self.actor_model(state)
|
||||
action = actor_model.add_exploration(action).cpu().numpy()[0]
|
||||
print(action)
|
||||
obs = obs.cpu().numpy()[0]
|
||||
next_obs, rew, done, _ = self.env.step(action)
|
||||
self.data_buffer.add(obs, action, next_obs, rew, done, self.global_episodes)
|
||||
obs = next_obs
|
||||
epi_reward += rew
|
||||
obs = self.env.reset()
|
||||
done=False
|
||||
all_rews.append(epi_reward)
|
||||
return all_rews
|
||||
|
||||
def train(self, step, total_steps):
|
||||
counter = 0
|
||||
import matplotlib.pyplot as plt
|
||||
fig, ax = plt.subplots()
|
||||
while step < total_steps:
|
||||
episodic_rews = self.collect_random_sequences(self.args.episodes_buffer)
|
||||
global_step = self.data_buffer.steps
|
||||
|
||||
# logger
|
||||
logs = OrderedDict()
|
||||
|
||||
while global_step < total_steps:
|
||||
step += 1
|
||||
for update_steps in range(self.args.update_steps):
|
||||
model_loss, actor_loss, value_loss = self.update((step-1)*args.update_steps + update_steps)
|
||||
episodic_rews = self.collect_sequences(self.args.episode_collection, actor_model=self.actor_model, encoder_model=self.obs_encoder)
|
||||
|
||||
logs.update({
|
||||
'model_loss' : model_loss,
|
||||
'actor_loss': actor_loss,
|
||||
'value_loss': value_loss,
|
||||
'train_avg_reward':np.mean(episodic_rews),
|
||||
'train_max_reward': np.max(episodic_rews),
|
||||
'train_min_reward': np.min(episodic_rews),
|
||||
'train_std_reward':np.std(episodic_rews),
|
||||
})
|
||||
|
||||
print("########## Global Step: ", global_step, " ##########")
|
||||
for key, value in logs.items():
|
||||
print(key, " : ", value)
|
||||
|
||||
print(global_step)
|
||||
if global_step % 3150 == 0 and self.data_buffer.steps!=0: #self.args.evaluation_interval == 0:
|
||||
print("Saving model")
|
||||
path = os.path.dirname(os.path.realpath(__file__)) + "/saved_models/models.pth"
|
||||
self.save_models(path)
|
||||
self.evaluate()
|
||||
|
||||
global_step = self.data_buffer.steps
|
||||
|
||||
# collect experience
|
||||
if step !=0:
|
||||
encoder = self.obs_encoder
|
||||
actor = self.actor_model
|
||||
all_rews = self.collect_sequences(self.args.episode_collection, random=False, actor_model=actor, encoder_model=encoder)
|
||||
else:
|
||||
all_rews = self.collect_sequences(self.args.episodes_buffer, random=True)
|
||||
all_rews = self.collect_sequences(self.args.episode_collection, actor_model=actor, encoder_model=encoder)
|
||||
|
||||
# collect sequences
|
||||
non_zero_indices = np.nonzero(self.data_buffer.episode_count)[0]
|
||||
current_obs = self.data_buffer.observations[non_zero_indices]
|
||||
next_obs = self.data_buffer.next_observations[non_zero_indices]
|
||||
actions_raw = self.data_buffer.actions[non_zero_indices]
|
||||
rewards = self.data_buffer.rewards[non_zero_indices]
|
||||
self.terms = np.where(self.data_buffer.terminals[non_zero_indices]!=0)[0]
|
||||
def update(self, step):
|
||||
last_observations, current_observations, next_observations, actions, next_actions, rewards = self.select_one_batch()
|
||||
|
||||
# Group by episodes
|
||||
current_obs = self.grouped_arrays(current_obs)
|
||||
next_obs = self.grouped_arrays(next_obs)
|
||||
actions_raw = self.grouped_arrays(actions_raw)
|
||||
rewards_ = self.grouped_arrays(rewards)
|
||||
|
||||
# Train encoder
|
||||
if step == 0:
|
||||
step += 1
|
||||
world_loss, enc_loss, rew_loss, dec_loss, ub_loss, lb_loss = self.world_model_losses(last_observations,
|
||||
current_observations,
|
||||
next_observations,
|
||||
actions,
|
||||
next_actions,
|
||||
rewards)
|
||||
self.world_model_opt.zero_grad()
|
||||
world_loss.backward()
|
||||
nn.utils.clip_grad_norm_(self.world_model_parameters, self.args.grad_clip_norm)
|
||||
self.world_model_opt.step()
|
||||
|
||||
update_steps = 1 if step > 1 else 1
|
||||
#for _ in range(self.args.collection_interval // self.args.episode_length+1):
|
||||
for _ in range(update_steps):
|
||||
counter += 1
|
||||
|
||||
# Select random chunks of episodes
|
||||
if current_obs.shape[0] < self.args.batch_size:
|
||||
random_episode_number = np.random.randint(0, current_obs.shape[0], self.args.batch_size)
|
||||
else:
|
||||
random_episode_number = random.sample(range(current_obs.shape[0]), self.args.batch_size)
|
||||
|
||||
if current_obs[0].shape[0]-self.args.episode_length < self.args.batch_size:
|
||||
init_index = np.random.randint(0, current_obs[0].shape[0]-self.args.episode_length-2, self.args.batch_size)
|
||||
else:
|
||||
init_index = np.asarray(random.sample(range(current_obs[0].shape[0]-self.args.episode_length), self.args.batch_size))
|
||||
random.shuffle(random_episode_number)
|
||||
random.shuffle(init_index)
|
||||
last_observations = self.select_first_k(current_obs, init_index, random_episode_number)[:-1]
|
||||
current_observations = self.select_first_k(current_obs, init_index, random_episode_number)[1:]
|
||||
next_observations = self.select_first_k(next_obs, init_index, random_episode_number)[:-1]
|
||||
actions = self.select_first_k(actions_raw, init_index, random_episode_number)[:-1].to(device)
|
||||
next_actions = self.select_first_k(actions_raw, init_index, random_episode_number)[1:].to(device)
|
||||
rewards = self.select_first_k(rewards_, init_index, random_episode_number)[1:].to(device)
|
||||
|
||||
# Preprocessing
|
||||
last_observations = preprocess_obs(last_observations).to(device)
|
||||
current_observations = preprocess_obs(current_observations).to(device)
|
||||
next_observations = preprocess_obs(next_observations).to(device)
|
||||
|
||||
# Initialize transition model states
|
||||
self.transition_model.init_states(self.args.batch_size, device) # (N,128)
|
||||
self.history = self.transition_model.prev_history # (N,128)
|
||||
|
||||
past_world_model_loss = 0
|
||||
past_action_loss = 0
|
||||
past_value_loss = 0
|
||||
for i in range(self.args.episode_length-1):
|
||||
if i > 0:
|
||||
# Encode observations and next_observations
|
||||
self.last_states_dict = self.get_features(last_observations[i])
|
||||
self.current_states_dict = self.get_features(current_observations[i])
|
||||
self.next_states_dict = self.get_features(next_observations[i], momentum=True)
|
||||
self.action = actions[i] # (N,6)
|
||||
self.next_action = next_actions[i] # (N,6)
|
||||
history = self.transition_model.prev_history
|
||||
|
||||
# Encode negative observations fro upper bound loss
|
||||
idx = torch.randperm(current_observations[i].shape[0]) # random permutation on batch
|
||||
random_time_index = torch.randint(0, current_observations.shape[0]-2, (1,)).item() # random time index
|
||||
negative_current_observations = current_observations[random_time_index][idx]
|
||||
self.negative_current_states_dict = self.get_features(negative_current_observations)
|
||||
|
||||
# Predict current state from past state with transition model
|
||||
last_states_sample = self.last_states_dict["sample"]
|
||||
predicted_current_state_dict = self.transition_model.imagine_step(last_states_sample, self.action, self.history)
|
||||
self.history = predicted_current_state_dict["history"]
|
||||
|
||||
# Calculate upper bound loss
|
||||
likeli_loss, ub_loss = self._upper_bound_minimization(self.last_states_dict["sample"].detach(),
|
||||
self.current_states_dict["sample"].detach(),
|
||||
self.negative_current_states_dict["sample"].detach(),
|
||||
predicted_current_state_dict["sample"].detach(),
|
||||
)
|
||||
|
||||
# Calculate encoder loss
|
||||
encoder_loss = self._past_encoder_loss(self.current_states_dict, predicted_current_state_dict)
|
||||
|
||||
# decoder loss
|
||||
horizon = np.minimum(self.args.imagine_horizon, self.args.episode_length-1-i)
|
||||
nxt_obs = next_observations[i:i+horizon].reshape(-1,9,84,84)
|
||||
next_states_encodings = self.get_features(nxt_obs)["sample"].view(horizon,self.args.batch_size, -1)
|
||||
obs_dist = self.obs_decoder(next_states_encodings)
|
||||
decoder_loss = -torch.mean(obs_dist.log_prob(next_observations[i:i+horizon]))
|
||||
|
||||
# contrastive projection
|
||||
vec_anchor = predicted_current_state_dict["sample"].detach()
|
||||
vec_positive = self.next_states_dict["sample"].detach()
|
||||
z_anchor = self.prjoection_head(vec_anchor, self.action)
|
||||
z_positive = self.prjoection_head_momentum(vec_positive, next_actions[i]).detach()
|
||||
|
||||
# contrastive loss
|
||||
logits = self.contrastive_head(z_anchor, z_positive)
|
||||
labels = torch.arange(logits.shape[0]).long().to(device)
|
||||
lb_loss = F.cross_entropy(logits, labels)
|
||||
|
||||
# reward loss
|
||||
reward_dist = self.reward_model(self.current_states_dict["sample"])
|
||||
reward_loss = -torch.mean(reward_dist.log_prob(rewards[i]))
|
||||
|
||||
# world model loss
|
||||
world_model_loss = (10*encoder_loss + 10*ub_loss + 1e-1*lb_loss + reward_loss + 1e-3*decoder_loss + past_world_model_loss) * 1e-3
|
||||
past_world_model_loss = world_model_loss.item()
|
||||
|
||||
# actor loss
|
||||
with FreezeParameters(self.world_model_modules):
|
||||
imagine_horizon = self.args.imagine_horizon #np.minimum(self.args.imagine_horizon, self.args.episode_length-1-i)
|
||||
action = self.actor_model(self.current_states_dict["sample"])
|
||||
imagined_rollout = self.transition_model.imagine_rollout(self.current_states_dict["sample"],
|
||||
action, self.history,
|
||||
imagine_horizon)
|
||||
|
||||
with FreezeParameters(self.world_model_modules + self.value_modules):
|
||||
imag_rewards = self.reward_model(imagined_rollout["sample"]).mean
|
||||
imag_values = self.target_value_model(imagined_rollout["sample"]).mean
|
||||
|
||||
discounts = self.args.discount * torch.ones_like(imag_rewards).detach()
|
||||
|
||||
self.returns = self._compute_lambda_return(imag_rewards[:-1],
|
||||
imag_values[:-1],
|
||||
discounts[:-1] ,
|
||||
self.args.td_lambda,
|
||||
imag_values[-1])
|
||||
|
||||
discounts = torch.cat([torch.ones_like(discounts[:1]), discounts[1:-1]], 0)
|
||||
self.discounts = torch.cumprod(discounts, 0).detach()
|
||||
actor_loss = -torch.mean(self.discounts * self.returns) + past_action_loss
|
||||
past_action_loss = actor_loss.item()
|
||||
|
||||
# value loss
|
||||
with torch.no_grad():
|
||||
value_feat = imagined_rollout["sample"][:-1].detach()
|
||||
value_targ = self.returns.detach()
|
||||
|
||||
value_dist = self.value_model(value_feat)
|
||||
|
||||
value_loss = -torch.mean(self.discounts * value_dist.log_prob(value_targ).unsqueeze(-1)) + past_value_loss
|
||||
past_value_loss = value_loss.item()
|
||||
|
||||
# update target value
|
||||
if step % self.args.value_target_update_freq == 0:
|
||||
self.target_value_model = copy.deepcopy(self.value_model)
|
||||
|
||||
# counter for reward
|
||||
#count = np.arange((counter-1) * (self.args.batch_size), (counter) * (self.args.batch_size))
|
||||
count = (counter-1) * (self.args.batch_size)
|
||||
if step % self.args.logging_freq:
|
||||
writer.add_scalar('World Loss/World Loss', world_model_loss.detach().item(), self.data_buffer.steps)
|
||||
writer.add_scalar('Main Models Loss/Encoder Loss', encoder_loss.detach().item(), self.data_buffer.steps)
|
||||
writer.add_scalar('Main Models Loss/Decoder Loss', decoder_loss, self.data_buffer.steps)
|
||||
writer.add_scalar('Actor Critic Loss/Actor Loss', actor_loss.detach().item(), self.data_buffer.steps)
|
||||
writer.add_scalar('Actor Critic Loss/Value Loss', value_loss.detach().item(), self.data_buffer.steps)
|
||||
writer.add_scalar('Actor Critic Loss/Reward Loss', reward_loss.detach().item(), self.data_buffer.steps)
|
||||
writer.add_scalar('Bound Loss/Upper Bound Loss', ub_loss.detach().item(), self.data_buffer.steps)
|
||||
writer.add_scalar('Bound Loss/Lower Bound Loss', lb_loss.detach().item(), self.data_buffer.steps)
|
||||
step += 1
|
||||
|
||||
print(world_model_loss, actor_loss, value_loss)
|
||||
|
||||
# update actor model
|
||||
actor_loss = self.actor_model_losses()
|
||||
self.actor_opt.zero_grad()
|
||||
actor_loss.backward()
|
||||
nn.utils.clip_grad_norm_(self.actor_model.parameters(), self.args.grad_clip_norm)
|
||||
self.actor_opt.step()
|
||||
|
||||
# update world model
|
||||
self.world_model_opt.zero_grad()
|
||||
world_model_loss.backward()
|
||||
nn.utils.clip_grad_norm_(self.world_model_parameters, self.args.grad_clip_norm)
|
||||
self.world_model_opt.step()
|
||||
|
||||
# update value model
|
||||
value_loss = self.value_model_losses()
|
||||
self.value_opt.zero_grad()
|
||||
value_loss.backward()
|
||||
nn.utils.clip_grad_norm_(self.value_model.parameters(), self.args.grad_clip_norm)
|
||||
@ -481,16 +367,170 @@ class DPI:
|
||||
soft_update_params(self.obs_encoder, self.obs_encoder_momentum, self.args.encoder_tau)
|
||||
soft_update_params(self.prjoection_head, self.prjoection_head_momentum, self.args.encoder_tau)
|
||||
|
||||
rew_len = np.arange(count, count+self.args.episode_collection) if count != 0 else np.arange(0, self.args.batch_size)
|
||||
for j in range(len(all_rews)):
|
||||
writer.add_scalar('Rewards/Rewards', all_rews[j], rew_len[j])
|
||||
# update target value networks
|
||||
if step % self.args.value_target_update_freq == 0:
|
||||
self.target_value_model = copy.deepcopy(self.value_model)
|
||||
|
||||
if step % self.args.logging_freq:
|
||||
writer.add_scalar('World Loss/World Loss', world_loss.detach().item(), step)
|
||||
writer.add_scalar('Main Models Loss/Encoder Loss', enc_loss.detach().item(), step)
|
||||
writer.add_scalar('Main Models Loss/Decoder Loss', dec_loss, step)
|
||||
writer.add_scalar('Actor Critic Loss/Actor Loss', actor_loss.detach().item(), step)
|
||||
writer.add_scalar('Actor Critic Loss/Value Loss', value_loss.detach().item(), step)
|
||||
writer.add_scalar('Actor Critic Loss/Reward Loss', rew_loss.detach().item(), step)
|
||||
writer.add_scalar('Bound Loss/Upper Bound Loss', ub_loss.detach().item(), step)
|
||||
writer.add_scalar('Bound Loss/Lower Bound Loss', lb_loss.detach().item(), step)
|
||||
|
||||
return world_loss.item(), actor_loss.item(), value_loss.item()
|
||||
|
||||
def world_model_losses(self, last_obs, curr_obs, nxt_obs, actions, nxt_actions, rewards):
|
||||
self.last_state_feat = self.get_features(last_obs)
|
||||
self.curr_state_feat = self.get_features(curr_obs)
|
||||
self.nxt_state_feat = self.get_features(nxt_obs)
|
||||
|
||||
# states
|
||||
self.last_state_enc = self.last_state_feat["sample"]
|
||||
self.curr_state_enc = self.curr_state_feat["sample"]
|
||||
self.nxt_state_enc = self.nxt_state_feat["sample"]
|
||||
|
||||
# actions
|
||||
actions = actions
|
||||
nxt_actions = nxt_actions
|
||||
|
||||
# rewards
|
||||
rewards = rewards
|
||||
|
||||
# predict next states
|
||||
self.transition_model.init_states(self.args.batch_size, device) # (N,128)
|
||||
self.observed_rollout = self.transition_model.observe_rollout(self.last_state_enc, actions, self.transition_model.prev_history)
|
||||
self.pred_curr_state_dist = self.transition_model.get_dist(self.observed_rollout["mean"], self.observed_rollout["std"])
|
||||
self.pred_curr_state_enc = self.pred_curr_state_dist.mean
|
||||
|
||||
#print(torch.nn.MSELoss()(self.curr_state_enc, self.pred_curr_state_enc))
|
||||
#print(torch.distributions.kl_divergence(self.curr_state_feat["distribution"], self.pred_curr_state_dist).mean(),0)
|
||||
|
||||
|
||||
# encoder loss
|
||||
enc_loss = torch.nn.MSELoss()(self.curr_state_enc, self.pred_curr_state_enc)
|
||||
#self._encoder_loss(self.curr_state_feat["distribution"], self.pred_curr_state_dist)
|
||||
|
||||
# reward loss
|
||||
rew_dist = self.reward_model(self.curr_state_enc)
|
||||
rew_loss = -torch.mean(rew_dist.log_prob(rewards.unsqueeze(-1)))
|
||||
|
||||
# decoder loss
|
||||
dec_dist = self.obs_decoder(self.nxt_state_enc)
|
||||
dec_loss = -torch.mean(dec_dist.log_prob(nxt_obs))
|
||||
|
||||
# upper bound loss
|
||||
likelihood_loss, ub_loss = self._upper_bound_minimization(self.curr_state_enc,
|
||||
self.pred_curr_state_enc)
|
||||
|
||||
# lower bound loss
|
||||
# contrastive projection
|
||||
vec_anchor = self.pred_curr_state_enc
|
||||
vec_positive = self.nxt_state_enc
|
||||
z_anchor = self.prjoection_head(vec_anchor, nxt_actions)
|
||||
z_positive = self.prjoection_head_momentum(vec_positive, nxt_actions)
|
||||
|
||||
# contrastive loss
|
||||
past_lb_loss = 0
|
||||
for i in range(z_anchor.shape[0]):
|
||||
logits = self.contrastive_head(z_anchor[i], z_positive[i])
|
||||
labels = torch.arange(logits.shape[0]).long().to(device)
|
||||
lb_loss = F.cross_entropy(logits, labels) + past_lb_loss
|
||||
past_lb_loss = lb_loss.detach().item()
|
||||
lb_loss = lb_loss/(z_anchor.shape[0])
|
||||
|
||||
world_loss = enc_loss + rew_loss + dec_loss * 1e-4 + ub_loss * 10 + lb_loss
|
||||
|
||||
return world_loss, enc_loss , rew_loss, dec_loss * 1e-4, ub_loss * 10, lb_loss
|
||||
|
||||
def actor_model_losses(self):
|
||||
with torch.no_grad():
|
||||
curr_state_enc = self.transition_model.seq_to_batch(self.curr_state_feat, "sample")["sample"]
|
||||
curr_state_hist = self.transition_model.seq_to_batch(self.observed_rollout, "history")["sample"]
|
||||
|
||||
with FreezeParameters(self.world_model_modules):
|
||||
imagine_horizon = self.args.imagine_horizon
|
||||
action = self.actor_model(curr_state_enc)
|
||||
self.imagined_rollout = self.transition_model.imagine_rollout(curr_state_enc,
|
||||
action, curr_state_hist,
|
||||
imagine_horizon)
|
||||
|
||||
with FreezeParameters(self.world_model_modules + self.value_modules):
|
||||
imag_rewards = self.reward_model(self.imagined_rollout["sample"]).mean
|
||||
imag_values = self.target_value_model(self.imagined_rollout["sample"]).mean
|
||||
discounts = self.args.discount * torch.ones_like(imag_rewards).detach()
|
||||
self.returns = self._compute_lambda_return(imag_rewards[:-1],
|
||||
imag_values[:-1],
|
||||
discounts[:-1] ,
|
||||
self.args.td_lambda,
|
||||
imag_values[-1])
|
||||
|
||||
discounts = torch.cat([torch.ones_like(discounts[:1]), discounts[1:-1]], 0)
|
||||
self.discounts = torch.cumprod(discounts, 0).detach()
|
||||
actor_loss = -torch.mean(self.discounts * self.returns)
|
||||
return actor_loss
|
||||
|
||||
|
||||
def value_model_losses(self):
|
||||
# value loss
|
||||
with torch.no_grad():
|
||||
value_feat = self.imagined_rollout["sample"][:-1].detach()
|
||||
value_targ = self.returns.detach()
|
||||
value_dist = self.value_model(value_feat)
|
||||
value_loss = -torch.mean(self.discounts * value_dist.log_prob(value_targ).unsqueeze(-1))
|
||||
return value_loss
|
||||
|
||||
|
||||
def select_one_batch(self):
|
||||
# collect sequences
|
||||
non_zero_indices = np.nonzero(self.data_buffer.episode_count)[0]
|
||||
current_obs = self.data_buffer.observations[non_zero_indices]
|
||||
next_obs = self.data_buffer.next_observations[non_zero_indices]
|
||||
actions_raw = self.data_buffer.actions[non_zero_indices]
|
||||
rewards = self.data_buffer.rewards[non_zero_indices]
|
||||
self.terms = np.where(self.data_buffer.terminals[non_zero_indices]!=False)[0]
|
||||
|
||||
# group by episodes
|
||||
current_obs = self.grouped_arrays(current_obs)
|
||||
next_obs = self.grouped_arrays(next_obs)
|
||||
actions_raw = self.grouped_arrays(actions_raw)
|
||||
rewards_ = self.grouped_arrays(rewards)
|
||||
|
||||
# select random chunks of episodes
|
||||
if current_obs.shape[0] < self.args.batch_size:
|
||||
random_episode_number = np.random.randint(0, current_obs.shape[0], self.args.batch_size)
|
||||
else:
|
||||
random_episode_number = random.sample(range(current_obs.shape[0]), self.args.batch_size)
|
||||
|
||||
# select random starting points
|
||||
if current_obs[0].shape[0]-self.args.episode_length < self.args.batch_size:
|
||||
init_index = np.random.randint(0, current_obs[0].shape[0]-self.args.episode_length-2, self.args.batch_size)
|
||||
else:
|
||||
init_index = np.asarray(random.sample(range(current_obs[0].shape[0]-self.args.episode_length), self.args.batch_size))
|
||||
|
||||
# shuffle
|
||||
random.shuffle(random_episode_number)
|
||||
random.shuffle(init_index)
|
||||
|
||||
# select first k elements
|
||||
last_observations = self.select_first_k(current_obs, init_index, random_episode_number)[:-1]
|
||||
current_observations = self.select_first_k(current_obs, init_index, random_episode_number)[1:]
|
||||
next_observations = self.select_first_k(next_obs, init_index, random_episode_number)[:-1]
|
||||
actions = self.select_first_k(actions_raw, init_index, random_episode_number)[:-1].to(device)
|
||||
next_actions = self.select_first_k(actions_raw, init_index, random_episode_number)[1:].to(device)
|
||||
rewards = self.select_first_k(rewards_, init_index, random_episode_number)[:-1].to(device)
|
||||
|
||||
# preprocessing
|
||||
last_observations = preprocess_obs(last_observations).to(device)
|
||||
current_observations = preprocess_obs(current_observations).to(device)
|
||||
next_observations = preprocess_obs(next_observations).to(device)
|
||||
|
||||
return last_observations, current_observations, next_observations, actions, next_actions, rewards
|
||||
|
||||
|
||||
print(step)
|
||||
if step % 2850 == 0 and self.data_buffer.steps!=0: #self.args.evaluation_interval == 0:
|
||||
print("Saving model")
|
||||
path = os.path.dirname(os.path.realpath(__file__)) + "/saved_models/models.pth"
|
||||
self.save_models(path)
|
||||
self.evaluate()
|
||||
|
||||
|
||||
def evaluate(self, eval_episodes=10):
|
||||
@ -511,7 +551,7 @@ class DPI:
|
||||
with torch.no_grad():
|
||||
obs = torch.tensor(obs.copy(), dtype=torch.float32).unsqueeze(0)
|
||||
obs_processed = preprocess_obs(obs).to(device)
|
||||
state = self.obs_encoder(obs_processed)["distribution"].sample()
|
||||
state = self.get_features(obs_processed)["distribution"].rsample()
|
||||
action = self.actor_model(state).cpu().detach().numpy().squeeze()
|
||||
|
||||
next_obs, rew, done, _ = self.env.step(action)
|
||||
@ -534,11 +574,9 @@ class DPI:
|
||||
|
||||
def grouped_arrays(self,array):
|
||||
indices = [0] + self.terms.tolist()
|
||||
|
||||
def subarrays():
|
||||
for start, end in zip(indices[:-1], indices[1:]):
|
||||
yield array[start:end]
|
||||
|
||||
try:
|
||||
subarrays = np.stack(list(subarrays()), axis=0)
|
||||
except ValueError:
|
||||
@ -548,13 +586,13 @@ class DPI:
|
||||
|
||||
def select_first_k(self, array, init_index, episode_number):
|
||||
term_index = init_index + self.args.episode_length
|
||||
|
||||
array = array[episode_number]
|
||||
|
||||
array_list = []
|
||||
for i in range(array.shape[0]):
|
||||
array_list.append(array[i][init_index[i]:term_index[i]])
|
||||
array = np.asarray(array_list)
|
||||
|
||||
if array.ndim == 5:
|
||||
transposed_array = np.transpose(array, (1, 0, 2, 3, 4))
|
||||
elif array.ndim == 4:
|
||||
@ -565,20 +603,16 @@ class DPI:
|
||||
transposed_array = np.transpose(array, (1, 0))
|
||||
else:
|
||||
transposed_array = np.expand_dims(array, axis=0)
|
||||
|
||||
#return torch.tensor(array).float()
|
||||
return torch.tensor(transposed_array).float()
|
||||
|
||||
def _upper_bound_minimization(self, last_states, current_states, negative_current_states, predicted_current_states):
|
||||
club_loss = self.club_sample(current_states, predicted_current_states, negative_current_states)
|
||||
def _upper_bound_minimization(self, current_states, predicted_current_states):
|
||||
club_loss = self.club_sample(current_states, predicted_current_states, current_states)
|
||||
likelihood_loss = 0
|
||||
return likelihood_loss, club_loss
|
||||
|
||||
def _past_encoder_loss(self, curr_states_dict, predicted_curr_states_dict):
|
||||
# current state distribution
|
||||
curr_states_dist = curr_states_dict["distribution"]
|
||||
|
||||
# predicted current state distribution
|
||||
predicted_curr_states_dist = predicted_curr_states_dict["distribution"]
|
||||
|
||||
def _encoder_loss(self, curr_states_dist, predicted_curr_states_dist):
|
||||
# KL divergence loss
|
||||
loss = torch.mean(torch.distributions.kl.kl_divergence(curr_states_dist,predicted_curr_states_dist))
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user