call nstep_q_return from dqn_replay.py, still need test

This commit is contained in:
Dong Yan 2018-03-06 20:48:07 +08:00
parent 2a2274aeea
commit 24d75fd1aa
4 changed files with 19 additions and 110 deletions

View File

@ -1,95 +0,0 @@
#!/usr/bin/env python
from __future__ import absolute_import
import tensorflow as tf
import gym
import numpy as np
import time
# our lib imports here! It's ok to append path in examples
import sys
sys.path.append('..')
from tianshou.core import losses
from tianshou.data.batch import Batch
import tianshou.data.advantage_estimation as advantage_estimation
import tianshou.core.policy.dqn as policy # TODO: fix imports as zhusuan so that only need to import to policy
import tianshou.core.value_function.action_value as value_function
import tianshou.data.replay_buffer.proportional as proportional
import tianshou.data.replay_buffer.rank_based as rank_based
import tianshou.data.replay_buffer.naive as naive
import tianshou.data.replay_buffer.Replay as Replay
# TODO: why this solves cartpole even without training?
if __name__ == '__main__':
env = gym.make('CartPole-v0')
observation_dim = env.observation_space.shape
action_dim = env.action_space.n
clip_param = 0.2
num_batches = 10
batch_size = 512
seed = 0
np.random.seed(seed)
tf.set_random_seed(seed)
### 1. build network with pure tf
observation_ph = tf.placeholder(tf.float32, shape=(None,) + observation_dim)
def my_network():
net = tf.layers.dense(observation_ph, 32, activation=tf.nn.tanh)
net = tf.layers.dense(net, 32, activation=tf.nn.tanh)
action_values = tf.layers.dense(net, action_dim, activation=None)
return None, action_values # no policy head
### 2. build policy, loss, optimizer
dqn = value_function.DQN(my_network, observation_placeholder=observation_ph, weight_update=100)
pi = policy.DQN(dqn)
dqn_loss = losses.qlearning(dqn)
total_loss = dqn_loss
global_step = tf.Variable(0, name='global_step', trainable=False)
optimizer = tf.train.AdamOptimizer(1e-4)
train_op = optimizer.minimize(total_loss, var_list=dqn.trainable_variables, global_step=tf.train.get_global_step())
# replay_memory = naive.NaiveExperience({'size': 1000})
replay_memory = rank_based.RankBasedExperience({'size': 30})
# replay_memory = proportional.PropotionalExperience({'size': 100, 'batch_size': 10})
data_collector = Replay.Replay(replay_memory, env, pi, [advantage_estimation.ReplayMemoryQReturn(1, dqn)], [dqn])
### 3. define data collection
# data_collector = Batch(env, pi, [advantage_estimation.nstep_q_return(1, dqn)], [dqn])
### 4. start training
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
# assign actor to pi_old
pi.sync_weights() # TODO: automate this for policies with target network
start_time = time.time()
#TODO : repeat_num shoulde be defined in some configuration files
repeat_num = 100
for i in range(repeat_num):
# collect data
# data_collector.collect(nums=50)
data_collector.collect(num_episodes=50, epsilon_greedy= (repeat_num - i + 0.0) / repeat_num)
# print current return
print('Epoch {}:'.format(i))
data_collector.statistics()
# update network
for _ in range(num_batches):
feed_dict = data_collector.next_batch(batch_size, tf.train.global_step(sess, global_step))
sess.run(train_op, feed_dict=feed_dict)
print('Elapsed time: {:.1f} min'.format((time.time() - start_time) / 60))

View File

@ -14,6 +14,7 @@ from tianshou.core import losses
import tianshou.data.advantage_estimation as advantage_estimation
import tianshou.core.policy.dqn as policy # TODO: fix imports as zhusuan so that only need to import to policy
import tianshou.core.value_function.action_value as value_function
import sys
from tianshou.data.replay_buffer.vanilla import VanillaReplayBuffer
from tianshou.data.data_collector import DataCollector
@ -79,7 +80,7 @@ if __name__ == '__main__':
start_time = time.time()
epsilon = 0.5
pi.set_epsilon_train(epsilon)
data_collector.collect(num_timesteps=1e3) # warm-up
data_collector.collect(num_timesteps=int(1e3)) # warm-up
for i in range(int(1e8)): # number of training steps
# anneal epsilon step-wise
if (i + 1) % 1e4 == 0 and epsilon > 0.1:
@ -101,4 +102,4 @@ if __name__ == '__main__':
if i % 1000 == 0:
# epsilon 0.05 as in nature paper
pi.set_epsilon_test(0.05)
test(env, pi) # go for act_test of pi, not act
#test(env, pi) # go for act_test of pi, not act

View File

@ -8,20 +8,20 @@ REWARD = 2
DONE = 3
# modified for new interfaces
def full_return(buffer, index=None):
def full_return(buffer, indexes=None):
"""
naively compute full return
:param buffer: buffer with property index and data. index determines the current content in `buffer`.
:param index: (sampled) index to be computed. Defaults to all the data in `buffer`. Not necessarily in order within
:param indexes: (sampled) index to be computed. Defaults to all the data in `buffer`. Not necessarily in order within
each episode.
:return: dict with key 'return' and value the computed returns corresponding to `index`.
"""
index = index or buffer.index
indexes = indexes or buffer.index
raw_data = buffer.data
returns = []
for i_episode in range(len(index)):
index_this = index[i_episode]
for i_episode in range(len(indexes)):
index_this = indexes[i_episode]
if index_this:
episode = raw_data[i_episode]
if not episode[-1][DONE]:
@ -111,7 +111,7 @@ class nstep_q_return:
self.use_target_network = use_target_network
# TODO : we should transfer the tf -> numpy/python -> tf into a monolithic compute graph in tf
def __call__(self, buffer, index=None):
def __call__(self, buffer, indexes=None):
"""
:param buffer: buffer with property index and data. index determines the current content in `buffer`.
:param index: (sampled) index to be computed. Defaults to all the data in `buffer`. Not necessarily in order within
@ -119,7 +119,7 @@ class nstep_q_return:
:return: dict with key 'return' and value the computed returns corresponding to `index`.
"""
qvalue = self.action_value._value_tensor_all_actions
index = index or buffer.index
indexes = indexes or buffer.index
episodes = buffer.data
discount_factor = 0.99
returns = []
@ -128,8 +128,8 @@ class nstep_q_return:
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
for episode_index in range(len(index)):
index = index[episode_index]
for episode_index in range(len(indexes)):
index = indexes[episode_index]
if index:
episode = episodes[episode_index]
episode_q = []
@ -145,9 +145,11 @@ class nstep_q_return:
current_discount_factor *= discount_factor
last_frame_index = lfi
if last_frame_index > i:
target_q += current_discount_factor * \
max(sess.run(qvalue, feed_dict={self.action_value.managed_placeholders['observation']:
episode[last_frame_index][STATE]}))
state = episode[last_frame_index][STATE]
# the shape of qpredict is [batch_size, action_dimension]
qpredict = sess.run(qvalue, feed_dict={self.action_value.managed_placeholders['observation']:
state.reshape(1, state.shape[0])})
target_q += current_discount_factor * max(qpredict[0])
episode_q.append(target_q)
returns.append(episode_q)

View File

@ -1,6 +1,7 @@
import numpy as np
import logging
import itertools
import sys
from .replay_buffer.base import ReplayBufferBase
@ -59,7 +60,7 @@ class DataCollector(object):
sampled_index = self.data_buffer.sample(batch_size)
if self.process_mode == 'sample':
for processor in self.process_functions:
self.data_batch.update(processor(self.data_buffer, index=sampled_index))
self.data_batch.update(processor(self.data_buffer, indexes=sampled_index))
# flatten rank-2 list to numpy array, construct feed_dict
feed_dict = {}