def reset(self):
logger.debug('Resetting actors')
self.clear()
logger.warning('Halting threads')
self.data.put(DataKey.THREAD_HALT, True)
self.vehicle.apply_control(icarla.vehicle_control(throttle=0, steer=0))
icarla.set_velocity(self.vehicle, icarla.vector3d())
logger.debug('Environment reset successful')
def loop(self):
if self.spectator is None or self.vehicle is None or self.path is None:
time.sleep(1.0)
else:
try:
self.update()
except RuntimeError as r:
logger.error(f'Error: {r}')
logger.warning(f'Setting vehicle to None')
self.vehicle = None
def predict(self, state):
if state is None:
return None, None
if np.sign(state.side) < 0:
return 1, choices[1]
elif np.sign(state.side) > 0:
return 0, choices[0]
else:
logger.warning('OmniscientAgent is confused!')
return 0, choices[0]
def __init__(self):
self.actors = []
self.data = Data()
self.vehicle = None
self.path = None
logger.warning('Halting threads')
self.data.put(DataKey.THREAD_HALT, True)
self.c = ControllerThread(self.data)
self.p = PollerThread(self.data)
self.s = SpectatorFollowThread(self.data)
self.connection = Connection()
def work(self):
# Attention! Contains infinite loop
while True:
self.clear_screen()
if self.state is not None:
self.update_screen()
for event in pygame.event.get():
if event.type == pygame.QUIT:
pygame.quit()
logger.warning('PyWindow closing')
return
self.i += 1
pygame.time.wait(50)
pygame.display.update()
def loop(self):
if self.vehicle is None:
time.sleep(1.0)
else:
try:
self.poll_acceleration()
self.poll_position_and_direction()
self.poll_angular_acceleration()
self.poll_velocity()
time.sleep(POLL_TIME)
except RuntimeError as r:
logger.error(f'Error: {r}')
logger.warning(f'Setting vehicle to None')
self.vehicle = None
def control(self, throttle, steering):
if self.vehicle is not None:
if throttle >= 0.0:
try:
self.vehicle.apply_control(icarla.vehicle_control(throttle=throttle, steer=steering))
except RuntimeError as r:
logger.error(f'Error: {r}')
logger.warning(f'Setting vehicle to None')
self.vehicle = None
else:
try:
throttle *= -1.0
self.vehicle.apply_control(icarla.vehicle_control(reverse=True, throttle=throttle, steer=steering))
except RuntimeError as r:
logger.error(f'Error: {r}')
logger.warning(f'Setting vehicle to None')
self.vehicle = None
else:
raise RuntimeError('Vehicle not set in ControllerThread')
def nameof(var, outside_locals):
if var is None:
logger.debug(f'Not searching for name of None var in nameof()')
return 'None'
if issubclass(type(var), Iterable) and all(v is None for v in var):
logger.debug(f'Not searching for name of Nones var in nameof()')
return 'Nones'
for string, val in outside_locals.items():
if type(var) == type(val) and var == val:
return string
elif isinstance(val, object):
try:
string = nameof(var, val.__dict__)
if string is not None:
return string
except AttributeError:
pass
logger.warning(f'Did not find name of var {var} in nameof()')
return None
def move(actor, loc, safe=True):
"""
Moves the actor to a given location
| If safe mode is enabled, move becomes a blocking function which retries if it failed"""
actor.set_location(loc)
if safe:
time.sleep(0.5) # for some reason we must wait for the simulator to process this
d = actor.get_location().distance(loc)
if d > 1.0:
logger.warning(f'Failed moving {actor.type_id} to {loc}, retry...')
logger.info(f'Actors transform is {actor.get_transform()}')
actor.set_transform(carla.Transform(add_locations(loc, location(0, 0, 1)), rotation([0, 0, 0])))
time.sleep(0.5)
d = actor.get_location().distance(loc)
if d > 3.0:
logger.error(f'Failed moving {actor.type_id} to {loc}')
logger.info(f'Actors transform is {actor.get_transform()}')
return
apply_frame_time(frame_start)
status = env.check()
logger.debug(f'{status}')
return status.traveled
def apply_frame_time(frame_start):
frame_end = time.time_ns()
diff = None
if frame_end is not None and frame_start is not None:
diff = (frame_end - frame_start) // 1_000_000
if diff is not None and diff // 1_000 < TARGET_FRAME_TIME:
time.sleep(TARGET_FRAME_TIME - diff // 1_000)
else:
logger.warning('Frame time issue')
def train(agent, memory):
# memory is a list, containing (prev_state, prev_action, state) triplets
if TRAIN and not TRAIN_PER_DECISION:
if TRAIN_MEMORY_SIZE is None or len(memory) >= TRAIN_MEMORY_SIZE:
logger.info(
f'{type(agent).__name__} is starting training with memory of {len(memory)}'
)
agent.train_on_memory(memory)
memory.clear()
else:
logger.info(f'Memory not full, {len(memory)}/{TRAIN_MEMORY_SIZE}')
def save(self):
if self.network is not None:
self.network.save_weights(file_name)
else:
logger.warning('Cannot save network as it is None')
def train(self, train, test, epochs=None):
# awaiting as {ndarray: (X,)}
# inside: x {ndarray: (2,)}
# inside: {ndarray: (width, height)}* and float? *or other way (height, width)
logger.debug(
f'Num GPUs Available: {len(tf.config.experimental.list_physical_devices("GPU"))}'
)
model_checkpoint_callback = keras.callbacks.ModelCheckpoint(
filepath=checkpoint_file_names,
save_weights_only=True,
monitor=METRIC,
mode='min',
save_best_only=False) # revert to True if needed
initial_epoch = 0
found = False
path = ''
for i in list(range(1000))[::-1]:
path = checkpoint_file_names.format(epoch=i)
if not found and os.path.isfile(path):
logger.debug(f'File exists, loading checkpoint {path}')
self.network.load_weights(path)
if epochs is None:
initial_epoch = i
found = True
if not found:
logger.warning(f'No checkpoint found at {path}')
logger.debug('Reshaping data...')
train_x = []
train_y = []
for i in train:
train_x.append(i[0])
train_y.append(i[1])
train_x = np.array(train_x).reshape(
[-1, INPUT_SHAPE[0], INPUT_SHAPE[1], INPUT_SHAPE[2]])
train_y = np.array(train_y)
logger.debug('Training data finished')
test_x = np.array([i[0] for i in test]).reshape(
[-1, INPUT_SHAPE[0], INPUT_SHAPE[1], INPUT_SHAPE[2]])
test_y = np.array([i[1] for i in test])
logger.debug(
'Starting training on {} data with {} data as test set'.format(
len(train_x), len(test_x)))
if epochs is None:
e = 10
else:
e = epochs
# (4.) Training the CNN
logger.debug('Not printing epoch training data in KerasModel')
self.network.fit(train_x,
train_y,
validation_data=(test_x, test_y),
epochs=e,
initial_epoch=initial_epoch,
callbacks=[model_checkpoint_callback],
verbose=0)
# Saving the weights
self.network.save_weights(file_name)
logger.debug('Finished training, cleaning up...')
# Long training session can crash
tf.keras.backend.clear_session() # TODO (4) check whether it helps
gc.collect() # TODO (2) remove if doesn't help
logger.debug('Done')
# based on kerasmodel.py from Tamás Visy's temalab_gta repo
import gc
from settings import choices_count, AGENT_IM_WIDTH, AGENT_IM_HEIGHT
from support.logger import logger
# TODO (2) move and update warnings - best into readme or somewhere where they are actually read
logger.debug('Use a NumPy version with allow_pickle enabled')
logger.warning('Tensorflow warnings and info suppressed')
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
logger.debug('Tensorflow 2 requires 64 bit Python')
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers
from tensorflow.keras.optimizers import SGD
from tensorflow.lite.python.lite import TFLiteConverterV2
import numpy as np
import os
logger.debug('Keras imports finished')
INPUT_SHAPE = [AGENT_IM_WIDTH, AGENT_IM_HEIGHT, 1] # [WIDTH, HEIGHT, 1]
LEARNING_RATE = 0.1
PATH_PREFIX = ''
METRIC = 'val_loss'