def _get(self, url):
logging.debug('--> GET: %s' % url)
# fire it off
res = requests.get(url)
if not res.ok:
logging.error("xx> %s when getting %s: %s" %
(res.status_code, url, res.error))
return []
# decode JSON
data = {}
try:
data = json.loads(res.content)
except Exception as e:
logging.error("-----\n%s\n-----\n%s\n-----" % (e, res.content))
return []
self.previousPageURI = data.get('previousPageURI')
self.nextPageURI = data.get('nextPageURI')
if self.nextPageURI:
self.nextPageURI = self.nextPageURI.replace(
' ', '+') # some queries come back with a space!
self.totalCount = int(data.get('totalCount', 1))
# instantiate Trial objects
trials = []
for tr in data.get('results', []):
trial = Trial()
trial.update_from_lilly(tr)
trials.append(trial)
return trials
def make_trial_set(self):
self.trialSet = []
all_pairs = list(
itertools.product(self.sPlusOrientations, self.sMinusOrientations))
for numSMinus in self.sMinusPresentations:
for p in all_pairs:
(sPlusOrientation, sMinusOrientation) = p
if abs(sPlusOrientation - sMinusOrientation) < 0.001:
#make sure SPLUS and SMINUS are different
continue
t = Trial()
t.numSMinus = numSMinus
t.sPlusOrientation = sPlusOrientation
t.sMinusOrientation = sMinusOrientation
self.trialSet.append(t)
print str(len(self.trialSet)) + " different trial conditions."
self.sequencer = Sequencer(self.trialSet, self.sequenceType)
if self.sequenceType == Sequences.INTERVAL or self.sequenceType == Sequences.INTERVAL_RETRY:
self.sequencer.sequencer.easyOris = self.easyOris
self.sequencer.sequencer.hardOris = self.hardOris
self.sequencer.sequencer.numEasy = self.numEasy
self.sequencer.sequencer.numHard = self.numHard
def _get(self, url):
logging.debug('--> GET: %s' % url)
# fire it off
res = requests.get(url)
if not res.ok:
logging.error("xx> %s when getting %s: %s" % (res.status_code, url, res.error))
return []
# decode JSON
data = {}
try:
data = json.loads(res.content)
except Exception as e:
logging.error("-----\n%s\n-----\n%s\n-----" % (e, res.content))
return []
self.previousPageURI = data.get('previousPageURI')
self.nextPageURI = data.get('nextPageURI')
if self.nextPageURI:
self.nextPageURI = self.nextPageURI.replace(' ', '+') # some queries come back with a space!
self.totalCount = int(data.get('totalCount', 1))
# instantiate Trial objects
trials = []
for tr in data.get('results', []):
trial = Trial()
trial.update_from_lilly(tr)
trials.append(trial)
return trials
def move_to_next_trial(self):
if self.current_trial_index is None:
self.current_trial_index = 0
rospy.logwarn(self.current_trial_index)
else:
del self.current_trial
self.current_trial_index += 1
rospy.logwarn(self.current_trial_index)
trial_param = self.get_trial_params(self.current_trial_index)
self.current_trial = Trial(self.elapsed_time, self.mean_angle,
trial_param, self.devices,
self.current_trial_index)
def extract_trials(self):
trials = []
all_poke_idx, poke_times, all_withdraw_idx, withdraw_times, all_odor_idx, odor_times = self.get_trial_times(
)
# set up parameters
self.total_bins_per_trial = int(
(self.config.test.end_sec + self.config.train.window_size -
self.config.test.start_sec) / self.config.original_bin_width)
self.num_bins_per_spike_data = int(self.config.bin_width /
self.config.original_bin_width)
test_data_start_idx = range(
self.total_bins_per_trial)[0::self.num_bins_per_spike_data]
# for each trial, create a Trial instance
for i in range(len(odor_times)):
# get behaviorial data for current trial
odor, position, in_seq, performance = self.get_behavior(
all_odor_idx[i])
# create instance of class Trial
cur_trial = Trial(i + 1, poke_times[i], withdraw_times[i],
odor_times[i], odor, position, in_seq,
performance)
# get start index and end index for a trial
if self.config.onset_time:
start_idx, end_idx = self.get_start_end_idx(
all_odor_idx[i], all_withdraw_idx[i],
self.config.original_bin_width, self.config.test.start_sec,
self.config.test.end_sec + self.config.train.window_size)
else:
start_idx, end_idx = self.get_start_end_idx(
all_poke_idx[i], all_withdraw_idx[i],
self.config.original_bin_width, self.config.test.start_sec,
self.config.test.end_sec + self.config.train.window_size)
# get neural activity data (spike train and lfp) for current trial
all_spikes_data, lfp_data = self.get_spikes_lfp_data(
start_idx, end_idx)
# save neural spike train data as a nested list
cur_trial.spikes['all'] = self.get_test_data(
test_data_start_idx, all_spikes_data,
self.config.compress_data)
trials.append(cur_trial)
# save the trials data in a pickle file
with open(self.config.test.trials_file, 'wb') as f:
pickle.dump(trials, f)
return trials
def run_test(upper_len, incr, num_trials, func, is_c, safety_time=2):
'''Returns multidimensional array of [[array_size, time],...]
upper_len - @type - int
- @param - the longest array size that should be tested on
incr - @type - int
- @param - increment the length of array for every set of trials
num_trials - @type - int
- @param - how many trials for every array length?
func - @type - function
- @param - the function to run the trials on
'''
trials = []
cur_len = 1
int_bound = int(math.log(sys.maxint, 2)) #is C's longint, actually
do_more = True
while (cur_len <= upper_len and do_more):
for i in range(num_trials):
t = Trial()
t.set_test_data(True, [], int, cur_len, int_bound)
t.run_test(func, is_c)
trials.append([cur_len, t.get_elapsed_time()])
e_t = t.get_elapsed_time()
# if it took more than safety_time, finish this set and stop
if (e_t > safety_time):
do_more = False
cur_len += incr
return trials
def next_task(self):
if self.trial.is_end():
if self.sentence_count == 6:
self.sentence_count = 3
chunk = self.stimuli.get_chunk(self.sentence_count)
if chunk is None:
self.current_handler = ExperimentEndHandler(self.update, self.end_callback)
return
else:
self.current_handler = BlockEndHandler(self.update, self.start_sentence_task)
self.trial = Trial(chunk, self.stimuli.get_shuffled_groups())
return
else:
self.sentence_count += 1
self.trial = Trial(self.stimuli.get_chunk(self.sentence_count), self.stimuli.get_shuffled_groups())
self.start_sentence_task()
def rank_image_cb(self, msg):
logger.info(
'Received message with {} compressed_imgs, {} imgs, {} strs'.
format(len(msg.compressed_imgs), len(msg.imgs), len(msg.strs)))
if self.trial and not self.trial.mode in (Trial.State.ABORTED,
Trial.State.COMPLETED):
print('Aborting due to ongoing trial')
self.action_server.set_aborted()
return
self.reset()
if len(msg.compressed_imgs) > 0:
self.screen.fill((255, 255, 255))
self.screen.blit(
self.font.render(
'Received {} images'.format(len(msg.compressed_imgs)), 1,
(0, 0, 0)), (40, self.size[1] / 2))
pygame.display.flip()
scaled_imgs = [
ImageConverter.from_ros(img) for img in msg.compressed_imgs
]
self.start_trial(
Trial(zip(msg.option_ids, scaled_imgs),
size=self.size,
preview_time=4000,
image_time=self.PRESENTATION_DELAY))
while self.ranking:
pygame.time.wait(100)
else:
self.action_server.set_aborted()
def main():
gridSize = 15
bounds = [1, 1, gridSize - 2, gridSize - 2]
minDistanceBetweenGrids = 5
condition = [-5, -3, -1, 0, 1, 3, 5]
counter = [0] * len(condition)
initialWorld = InitialWorld(bounds)
updateWorld = UpdateWorld(bounds, condition, counter)
pg.init()
screenWidth = 680
screenHeight = 680
screen = pg.display.set_mode((screenWidth, screenHeight))
leaveEdgeSpace = 2
lineWidth = 1
backgroundColor = [205, 255, 204]
lineColor = [0, 0, 0]
targetColor = [255, 50, 50]
playerColor = [50, 50, 255]
targetRadius = 10
playerRadius = 10
stopwatchUnit = 100
finishTime = 1000 * 30
block = 1
textColorTuple = (255, 50, 50)
stopwatchEvent = pg.USEREVENT + 1
pg.time.set_timer(stopwatchEvent, stopwatchUnit)
pg.event.set_allowed([pg.KEYDOWN, pg.QUIT, stopwatchEvent])
picturePath = os.path.abspath(os.path.join(os.getcwd(), os.pardir)) + '/Pictures/'
resultsPath = os.path.abspath(os.path.join(os.getcwd(), os.pardir)) + '/Results/'
experimentValues = co.OrderedDict()
experimentValues["name"] = input("Please enter your name:").capitalize()
experimentValues["condition"] = 'None'
writerPath = resultsPath + experimentValues["name"] + '.csv'
writer = WriteDataFrameToCSV(writerPath)
introductionImage = pg.image.load(picturePath + 'introduction.png')
restImage = pg.image.load(picturePath + 'rest.png')
finishImage = pg.image.load(picturePath + 'finish.png')
introductionImage = pg.transform.scale(introductionImage, (screenWidth, screenHeight))
finishImage = pg.transform.scale(finishImage, (int(screenWidth * 2 / 3), int(screenHeight / 4)))
drawBackground = DrawBackground(screen, gridSize, leaveEdgeSpace, backgroundColor, lineColor, lineWidth,
textColorTuple)
drawNewState = DrawNewState(screen, drawBackground, targetColor, playerColor, targetRadius, playerRadius)
drawImage = DrawImage(screen)
humanController = HumanController(gridSize, stopwatchEvent, stopwatchUnit, drawNewState, finishTime)
trial = Trial(humanController, drawNewState, stopwatchEvent, finishTime)
experiment = Experiment(trial, writer, experimentValues, initialWorld, updateWorld, drawImage, resultsPath,
minDistanceBetweenGrids)
giveExperimentFeedback = GiveExperimentFeedback(screen, textColorTuple, screenWidth, screenHeight)
drawImage(introductionImage)
score = [0] * block
for i in range(block):
score[i] = experiment(finishTime)
giveExperimentFeedback(i, score)
if i == block - 1:
drawImage(finishImage)
else:
drawImage(restImage)
participantsScore = np.sum(np.array(score))
print(participantsScore)
def __init__(self, db, update, end_callback, groups=None):
self.db = db
self.stimuli = Stimuli(groups)
self.update = update
self.sentence_count = 3
self.trial = Trial(self.stimuli.get_chunk(self.sentence_count), self.stimuli.get_shuffled_groups())
self.start_sentence_task()
self.end_callback = end_callback
def all_trials_main():
# roughly 1 day of trails
machines = [
# "mlcnetA.cs.wpi.edu",
# "mlcnetB.cs.wpi.edu",
# "mlcnetC.cs.wpi.edu",
"mlcnetD.cs.wpi.edu"
]
protocols = [
# "cubic",
# "bbr",
# "hybla",
"pcc"
]
dir = 'data/' + date.today().strftime("%Y-%m-%d")
if not os.path.exists(dir):
os.makedirs(dir)
# roughly 24 hours
for i in range(80):
for machine, protocol in zip(machines, protocols):
print(machine, protocol)
title = f"{dir}/{machine}_{protocol}_{i}"
trial = Trial(name=title, data="1G", remote=machine)
trial.remote_tc(cc=protocol)
trial.start()
def load_trials(mat, session):
if session is None:
return []
trials = []
data = mat['dv']['exp'][0, 0]
keys = ['dir', 'coh', 'durIdx', 'dur', 'tr', 'response', 'correct']
data = [data[key][0, 0] for key in keys]
data = [d.transpose()[0] if d.shape[0] > 1 else d[0] for d in data]
for dirc, coh, durIdx, dur, tr, resp, crct in zip(*data):
# for tr, coh, dirc, durIdx, dur, resp, crct in mat['D']:
trial = Trial(session, tr, coh, dur, durIdx, dirc, resp,
True if crct == 1 else False)
trials.append(trial)
return trials
def parse_trial(trial_path):
with open(trial_path,'rb') as input_file:
trial_result = json.load(input_file)
generation_list = list()
for gen in trial_result['generationalData']:
individual_list = list()
for indi in gen['individualList']:
individual_list.append(
Individual(indi['UUID'], indi['parentUUID'], indi['switch'], indi['lightFirst'], indi['fitness']))
generation_list.append(Generation(gen['generationNumber'], individual_list))
trial = Trial(trial_result['config'], generation_list)
return trial
class TrialHandler():
def __init__(self, db, update, end_callback, groups=None):
self.db = db
self.stimuli = Stimuli(groups)
self.update = update
self.sentence_count = 3
self.trial = Trial(self.stimuli.get_chunk(self.sentence_count), self.stimuli.get_shuffled_groups())
self.start_sentence_task()
self.end_callback = end_callback
def handle_keyboard(self, keycode):
self.current_handler.handle_keyboard(keycode)
def start_sentence_task(self):
self.sentences, self.words = self.trial.get_one_group_data_from_chunk()
self.current_handler = SentenceHandler(self.sentences, self.update, self.start_word_task)
def start_word_task(self):
self.current_handler = WordHandler(self.words, self.update, self.next_task, self.db)
def next_task(self):
if self.trial.is_end():
if self.sentence_count == 6:
self.sentence_count = 3
chunk = self.stimuli.get_chunk(self.sentence_count)
if chunk is None:
self.current_handler = ExperimentEndHandler(self.update, self.end_callback)
return
else:
self.current_handler = BlockEndHandler(self.update, self.start_sentence_task)
self.trial = Trial(chunk, self.stimuli.get_shuffled_groups())
return
else:
self.sentence_count += 1
self.trial = Trial(self.stimuli.get_chunk(self.sentence_count), self.stimuli.get_shuffled_groups())
self.start_sentence_task()
def get_best_fit(self, n_try, test_size):
best_trials = []
for a in self.dh.actions:
trials = []
for i in range(n_try):
a.split_train_test(test_size)
a.scale_dataset()
a.offline_train()
print a.get_regression_score(), a.gmm.score(a.y_test)
trials.append(
Trial(a, a.get_regression_score(), a.gmm.score(a.y_test)))
# print trials
trials.sort(key=lambda t: t.cluster_score, reverse=True)
best_trials.append(trials[0])
return best_trials
def main():
dir = 'pcc'
if not os.path.exists(dir):
os.makedirs(dir)
machine = "mlcnetD.cs.wpi.edu"
protocol = "pcc"
title = f"{dir}/{machine}_{protocol}"
trial = Trial(name=title, data="1G", remote=machine)
trial.remote_tc(cc=protocol)
dir = trial.data_dir()
trial.start()
def window_sizes():
dir = 'data/WINDOW_SIZES/' + date.today().strftime("%Y-%m-%d")
if not os.path.exists(dir):
os.makedirs(dir)
# roughly 24 hours
wmems = [
int(BDP * 0.5),
BDP * 1,
BDP * 2,
BDP * 4,
BDP * 8,
BDP * 16,
65625000,
]
for wmem in wmems:
title = f"{dir}/{wmem}"
trial = Trial(name=title, data="1G", remote='mlc1')
trial.remote_tc(cc='cubic', win=wmem)
trial.local_tc(cc='cubic', win=wmem)
trial.start()
def main(batch_size: int = Option(32, "-b"),
G_lr: float = Option(0.01,
help="Learning rate for generator training"),
D_lr: float = Option(0.01,
help="Learning rate for discriminator training"),
GAN_G_lr: float = Option(
0.00008, help="Learning rate for GAN generator training"),
GAN_D_lr: float = Option(
0.00016, help="Learning rate for GAN discriminator training"),
G_epoch: int = Option(10, help="Iteration of generator training"),
D_epoch: int = Option(3, help="Iteration of discriminator training"),
GAN_epoch: int = Option(1, help="Iteration of GAN training"),
itr: int = Option(1, help="Iteration of whole NOGAN training"),
optim_type: str = Option("ADAB",
help="Options of Optimizers for Generator"),
level: str = Option("O1",
help="FP16 Training Flags, Refer to Apex Docs"),
adv_weight: float = Option(1.0, help="Weight for discriminator loss"),
load_dir: Optional[Path] = Option(None)):
"""
NOGAN training Trial.
"""
torch.backends.cudnn.benchmark = True
assert (itr > 0), "Number must be bigger than 0"
trial = Trial(batch_size=batch_size,
G_lr=G_lr,
D_lr=D_lr,
optim_type=optim_type,
level=level)
decreased_lr = False
if load_dir is not None:
trial.load_trial(load_dir)
for _ in range(itr):
trial.Generator_NOGAN(
epochs=G_epoch,
content_weight=3.0,
recon_weight=10.,
loss=['content_loss', 'recon_loss'],
)
trial.Discriminator_NOGAN(epochs=D_epoch, adv_weight=adv_weight)
trial.GAN_NOGAN(GAN_epoch,
GAN_G_lr=GAN_G_lr,
GAN_D_lr=GAN_D_lr,
adv_weight=adv_weight)
def read_input_file(modo_inputs):
input_list = []
if modo_inputs == "prueba":
input_file = open("pairAndResInputs.txt", "r")
if modo_inputs == "biblia":
input_file = open("pairAndResInputsBIBLIA.txt", "r")
try:
for line in input_file:
# rstrip() para evitar que algun \n moleste
prime, pairString, res = line.rstrip().split(" ")
left, right = pairString.split(",")
trial = Trial(prime, int(left), int(right),
res) #res es una string porque a veces es una latra
input_list.append(trial)
input_file.close()
except UnboundLocalError as error:
print(error)
print("\nRecordar debes: es prueba o biblia. Escribir bien debes.\n")
sys.exit()
return input_list
def __init__(self):
self.current_trial = DummyTrial()
self.current_trial_index = None
rospy.init_node('virtual_desert')
self.get_param()
self.rate = rospy.Rate(self.param['update_rate'])
self.lock = threading.Lock()
self.start_time = rospy.get_time()
self.angle_lowpass_filter = lowpass_filter.LowpassFilter(
self.param['angle_lowpass_fcut'])
self.angle_accumulator = angle_utils.AngleAccumulator()
self.angle_fixer = angle_utils.AngleFixer()
self.devices = {}
self.devices['panels_controller'] = display_ctrl.LedController()
self.devices['alicat_proxy'] = AlicatProxy()
self.devices['autostep_proxy'] = AutostepProxy()
self.devices['autostep_tracking_data_pub'] = rospy.Publisher(
'/autostep/tracking_data', TrackingData, queue_size=10)
self.devices['sunled_proxy'] = BasicLedStripProxy()
self.initialize_panels_controller()
self.initialize_autostep()
self.rolling_circ_mean = RollingCircularMean(
self.param['rolling_mean_size'])
self.angle_data_sub = rospy.Subscriber('/angle_data', MsgAngleData,
self.on_angle_data_callback)
self.data_pub = rospy.Publisher('/virtual_desert_data',
VirtualDesertData,
queue_size=10)
self.param_pub = rospy.Publisher('/virtual_desert_param',
std_msgs.msg.String,
queue_size=10)
def all_trials_main():
# roughly 1 day of trails
machines = [
#"mlcnetA.cs.wpi.edu",
"mlcnetB.cs.wpi.edu"
#"mlcnetC.cs.wpi.edu",
# "mlcnetD.cs.wpi.edu"
]
protocols = [
#"cubic",
"bbr"
#"hybla",
# "pcc"
]
initcwnd = [
# 3,
# 5,
#10,
# 20,
# 40,
100,
250
]
dir = 'data/' + date.today().strftime("%Y-%m-%d")
if not os.path.exists(dir):
os.makedirs(dir)
# roughly 24 hours
for i in range(5):
for j in initcwnd:
for machine, protocol in zip(machines, protocols):
print(machine, protocol)
title = f"{dir}/{j}/{machine}_{protocol}_{i}"
# time.sleep(15)
trial = Trial(name=title, data="128MB", remote=machine)
# trial.mock()
trial.remote_tc(cc=protocol, initcwnd=j)
trial.start()
# The same for positions
positions = []
try:
while True:
p = input('Enter number of shares to buy (Ctrl+D to end) ')
try:
positions.append(int(p))
except ValueError as e:
print(str(e))
except KeyboardInterrupt:
print()
sys.exit(1)
except EOFError:
print()
# Continue execution
if len(positions) == 0:
print('No number of shares specified, exiting')
sys.exit(2)
output_file = open('result.txt', 'w')
for p in positions:
trial = Trial(p, num_trial)
trial.plot('histogram_%04d_pos.pdf' % p)
output_file.write(
'Position %d: Mean %f, STD %f\n' % (p, trial.mean, trial.std)
)
output_file.close()
class VirtualDesert(object):
Default_Param_File = 'virtual_desert_param.yaml'
def __init__(self):
self.current_trial = DummyTrial()
self.current_trial_index = None
rospy.init_node('virtual_desert')
self.get_param()
self.rate = rospy.Rate(self.param['update_rate'])
self.lock = threading.Lock()
self.start_time = rospy.get_time()
self.angle_lowpass_filter = lowpass_filter.LowpassFilter(
self.param['angle_lowpass_fcut'])
self.angle_accumulator = angle_utils.AngleAccumulator()
self.angle_fixer = angle_utils.AngleFixer()
self.devices = {}
self.devices['panels_controller'] = display_ctrl.LedController()
self.devices['alicat_proxy'] = AlicatProxy()
self.devices['autostep_proxy'] = AutostepProxy()
self.devices['autostep_tracking_data_pub'] = rospy.Publisher(
'/autostep/tracking_data', TrackingData, queue_size=10)
self.devices['sunled_proxy'] = BasicLedStripProxy()
self.initialize_panels_controller()
self.initialize_autostep()
self.rolling_circ_mean = RollingCircularMean(
self.param['rolling_mean_size'])
self.angle_data_sub = rospy.Subscriber('/angle_data', MsgAngleData,
self.on_angle_data_callback)
self.data_pub = rospy.Publisher('/virtual_desert_data',
VirtualDesertData,
queue_size=10)
self.param_pub = rospy.Publisher('/virtual_desert_param',
std_msgs.msg.String,
queue_size=10)
def initialize_panels_controller(self):
# Wait until we have a subscriber connected or else message may be thrown away
# It would be better if the panels controller used a service.
while self.devices['panels_controller'].pub.get_num_connections() < 1:
# May want to put a try counter on this and error out after some number of attempts
rospy.sleep(0.1)
self.devices['panels_controller'].set_config_id(
self.param['panels_config_id'])
self.devices['panels_controller'].stop()
def shutdown_panels(self):
self.devices['panels_controller'].stop()
self.devices['panels_controller'].all_off()
def initialize_autostep(self):
# Set up known starting angle=0
self.devices['autostep_proxy'].disable_tracking_mode()
self.devices['autostep_proxy'].set_move_mode('jog')
self.devices['autostep_proxy'].move_to(0.0)
self.devices['autostep_proxy'].busy_wait()
while self.devices['autostep_tracking_data_pub'].get_num_connections(
) < 1:
# May want to put a try counter on this and error out after some number of attempts
rospy.sleep(0.1)
def shutdown_autostep(self):
self.devices['autostep_proxy'].disable_tracking_mode()
self.devices['autostep_proxy'].soft_stop()
def get_param(self):
self.param = rospy.get_param('/virtual_desert', None)
if self.param is None:
param_file_path = os.path.join(
os.path.dirname(os.path.abspath(__file__)),
self.Default_Param_File)
with open(param_file_path, 'r') as f:
self.param = yaml.load(f)
@property
def angle_dt(self):
return 1.0 / self.param['angle_framerate']
@property
def elapsed_time(self):
return rospy.get_time() - self.start_time
@property
def mean_angle(self):
with self.lock:
mean_angle = self.rolling_circ_mean.value()
return mean_angle
@property
def angle(self):
with self.lock:
angle = self.angle_lowpass_filter.value
return angle
def on_angle_data_callback(self, data):
with self.lock:
self.rolling_circ_mean.insert_data(data.angle)
angle_unwrapped = self.angle_accumulator.update(data.angle)
angle_fixed = self.angle_fixer.fix_data(angle_unwrapped)
self.angle_lowpass_filter.update(angle_fixed, self.angle_dt)
def move_to_next_trial(self):
if self.current_trial_index is None:
self.current_trial_index = 0
rospy.logwarn(self.current_trial_index)
else:
del self.current_trial
self.current_trial_index += 1
rospy.logwarn(self.current_trial_index)
trial_param = self.get_trial_params(self.current_trial_index)
self.current_trial = Trial(self.elapsed_time, self.mean_angle,
trial_param, self.devices,
self.current_trial_index)
def get_trial_params(self, index):
"""
Get parameters for trial set flow parameters to default and override
where specified in trial
"""
trial_param = copy.deepcopy(self.param['trials'][index])
param_name_list = ['flow', 'panels', 'sunled', 'autostep']
for param_name in param_name_list:
try:
param = copy.deepcopy(
self.param['{}_default'.format(param_name)])
except KeyError:
param = {}
if param is None:
param = {}
param.update(trial_param[param_name])
trial_param[param_name] = param
return trial_param
def run(self):
# Set to true to make sure parameters are in bag file
if self.param['wait_for_param_sub']:
while self.param_pub.get_num_connections() < 1:
rospy.sleep(0.1)
self.param_pub.publish(json.dumps(self.param))
while not rospy.is_shutdown():
elapsed_time = self.elapsed_time
if elapsed_time > self.param['startup_delay']:
if self.current_trial.is_done(elapsed_time):
try:
self.move_to_next_trial()
except IndexError:
break # Done with trials -> exit loop
trial_msg = self.current_trial.update(elapsed_time, self.angle)
#print('t: {:0.2f}'.format(self.elapsed_time))
else:
print('pretrial delay, t={:0.2f}'.format(elapsed_time))
trial_msg = TrialData()
msg = VirtualDesertData()
msg.header.stamp = rospy.Time.now()
msg.angle = self.angle
msg.elapsed_time = elapsed_time
msg.current_trial_index = self.current_trial_index
msg.trial_data = trial_msg
self.data_pub.publish(msg)
self.rate.sleep()
self.shutdown_panels()
self.shutdown_autostep()
def _get_trials_data(self, session_nr, raw_spike_timestamps):
"""
This method use the original experiment stimuli data set to extract the order and categories of the stimuli
shown to each subjects. The label of the stimulus for the new old recognition task is also obtained using
this method.
:param session_nr: session number
:param raw_spike_timestamps: raw_spike_timestamps from cell
:return: a list of all trial objects
"""
# Get data from variant stimuli data set
block_id_recog = self.sessions[session_nr]['block_id_recog']
block_id_learn = self.sessions[session_nr]['block_id_learn']
variant = self.sessions[session_nr]['variant']
filename = ''
if variant == 1:
filename = 'NewOldDelay_v3.mat'
filename2 = 'NewOldDelayStimuli.mat'
elif variant == 2:
filename = 'NewOldDelay2_v3.mat'
filename2 = 'NewOldDelayStimuli2.mat'
elif variant == 3:
filename = 'NewOldDelay3_v3.mat'
filename2 = 'NewOldDelayStimuli3.mat'
path_to_labels = os.path.join(self.path, 'stimFiles', filename)
experiment_stimuli = loadmat(path_to_labels)['experimentStimuli']
stimuli_recog_list = experiment_stimuli[0, block_id_recog - 1][3]
new_old_recog_list = experiment_stimuli[0, block_id_recog - 1][4]
stimuli_learn_list = experiment_stimuli[0, block_id_learn - 1][2]
path_to_categories = os.path.join(self.path, 'stimFiles', filename2)
category_mat = loadmat(path_to_categories)
category_names = category_mat['categories']
category_mapping = pd.DataFrame(category_mat['categoryMapping']).set_index(0)
file_mapping = category_mat['fileMapping']
# Get events and raw cell data
events_recog = self._get_event_data(session_nr, experiment_type='recog')
recog_events_time_points = self._extract_event_periods(events_recog)
events_learn = self._get_event_data(session_nr, experiment_type='learn')
learn_events_time_points = self._extract_event_periods(events_learn)
# Get recog response from log file
experiment_id_recog = self.sessions[session_nr]['experiment_id_recog']
log_file_path = os.path.join(self._construct_data_path(session_nr, 'events'), 'newold'+str(experiment_id_recog)+'.txt')
log_file = pd.read_csv(log_file_path, sep=';', header=None)
log_events = np.asarray(log_file[1])
recog_responses = log_events[(log_events >= 31)*(log_events <= 36)] - 30
# Get learn response from log file
experiment_id_learn = self.sessions[session_nr]['experiment_id_learn']
log_file_path = os.path.join(self._construct_data_path(session_nr, 'events'), 'newold'+str(experiment_id_learn)+'.txt')
log_file = pd.read_csv(log_file_path, sep=';', header=None)
log_events = np.asarray(log_file[1])
learn_responses = log_events[(log_events >= 20) * (log_events <= 21)] - 20
# Create the trial list
trials = []
for i in range(0, 100):
# Processing the recog phase
baseline_offset = 1000000
trial_start = recog_events_time_points['stimulus_on'][i] - baseline_offset
trial_end = recog_events_time_points['trial_end'][i]
trial_duration = trial_end - trial_start
trial_timestamps_recog = raw_spike_timestamps[(raw_spike_timestamps > trial_start) *
(raw_spike_timestamps <= trial_end)] - trial_start
stimuli_recog_id = stimuli_recog_list[0][i]
category_recog = category_mapping.loc[stimuli_recog_id, 1]
category_name_recog = category_names[0, category_recog - 1][0]
file_path = file_mapping[0][i][0].replace('C:\code\\', '')
new_old_recog = new_old_recog_list[0][i]
response_recog = recog_responses[i]
# Processing the learning phase
trial_start = learn_events_time_points['stimulus_on'][i] - baseline_offset
trial_end = learn_events_time_points['trial_end'][i]
trial_duration = trial_end - trial_start
trial_timestamps_learn = raw_spike_timestamps[(raw_spike_timestamps > trial_start) *
(raw_spike_timestamps <= trial_end)] - trial_start
stimuli_learn_id = stimuli_learn_list[0][i]
category_learn = category_mapping.loc[stimuli_learn_id, 1]
category_name_learn = category_names[0, category_learn - 1][0]
response_learn = learn_responses[i]
# Creating this trial object
trial = Trial(category_recog, category_name_recog, new_old_recog, response_recog, category_learn,
category_name_learn, response_learn, file_path, stimuli_recog_id, trial_timestamps_recog,
trial_timestamps_learn)
trials.append(trial)
return trials
# The same for positions
positions = []
try:
while True:
p = input('Enter number of shares to buy (Ctrl+D to end) ')
try:
positions.append(int(p))
except ValueError as e:
print(str(e))
except KeyboardInterrupt:
print()
sys.exit(1)
except EOFError:
print()
# Continue execution
if len(positions) == 0:
print('No number of shares specified, exiting')
sys.exit(2)
output_file = open('result.txt', 'w')
for p in positions:
trial = Trial(p, num_trial)
trial.plot('histogram_%04d_pos.pdf' % p)
output_file.write('Position %d: Mean %f, STD %f\n' %
(p, trial.mean, trial.std))
output_file.close()
def Instruct(self, config, lang="E"):
inst = Func(get_instructions, config).result
# with If(practice==True):
with Parallel():
if not config.TOUCH:
MouseCursor(blocking=False)
with Serial(blocking=False):
with Parallel():
toplbl = Label(text=inst['top_text'],
markup=True,
halign='center',
center_x=self.exp.screen.center_x,
center_y=self.exp.screen.center_y,
font_size=s(config.INST_FONT_SIZE))
# upper left
Flanker(center_x=self.exp.screen.center_x - s(400),
center_y=toplbl.top + s(125),
sep=s(config.CONFIG_SEP),
df=s(config.CONFIG_DF),
line_width=s(config.LW),
stim="__<__\n_<<<_\n<<<<<\n_<<<_\n__<__\n")
# upper middle
Flanker(center_x=self.exp.screen.center_x,
center_y=toplbl.top + s(125),
sep=s(config.CONFIG_SEP),
df=s(config.CONFIG_DF),
line_width=s(config.LW),
stim="__>__\n_><>_\n><<<>\n_><>_\n__>__\n")
# upper right
Flanker(center_x=self.exp.screen.center_x + s(400),
center_y=toplbl.top + s(125),
sep=s(config.CONFIG_SEP),
df=s(config.CONFIG_DF),
line_width=s(config.LW),
stim="__<__\n_<><_\n<><><\n_<><_\n__<__\n")
# lower left
Flanker(center_x=self.exp.screen.center_x - s(400),
center_y=toplbl.bottom - s(125),
sep=s(config.CONFIG_SEP),
df=s(config.CONFIG_DF),
line_width=s(config.LW),
stim="__<__\n_<><_\n<><><\n_<><_\n__<__\n")
# lower middle
Flanker(center_x=self.exp.screen.center_x,
center_y=toplbl.bottom - s(125),
sep=s(config.CONFIG_SEP),
df=s(config.CONFIG_DF),
line_width=s(config.LW),
stim="__>__\n_><>_\n><><>\n_><>_\n__>__\n")
# lower right
Flanker(center_x=self.exp.screen.center_x + s(400),
center_y=toplbl.bottom - s(125),
sep=s(config.CONFIG_SEP),
df=s(config.CONFIG_DF),
line_width=s(config.LW),
stim="__<__\n_<><_\n<>>><\n_<><_\n__<__\n")
with UntilDone():
Wait(until=toplbl.appear_time)
GetResponse(keys=config.CONT_KEY)
Wait(1.0)
with Loop(
[["__<__\n_<<<_\n<<<<<\n_<<<_\n__<__\n", config.RESP_KEYS[0]],
["__>__\n_><>_\n><<<>\n_><>_\n__>__\n", config.RESP_KEYS[0]],
["__<__\n_<><_\n<>>><\n_<><_\n__<__\n", config.RESP_KEYS[-1]],
["__>__\n_><>_\n><><>\n_><>_\n__>__\n",
config.RESP_KEYS[-1]]]) as prac_ev:
Wait(1.0)
p2 = Trial(config,
stim=prac_ev.current[0],
center_x=self.exp.screen.center_x,
center_y=self.exp.screen.center_y,
correct_resp=prac_ev.current[1],
condition='+')
with If(p2.correct):
# They got it right
Label(text=u"\u2713",
color='green',
duration=config.FEEDBACK_TIME,
font_size=s(config.FEEDBACK_FONT_SIZE),
font_name='DejaVuSans.ttf')
with Else():
# they got it wrong
Label(text=u"\u2717",
color='red',
font_size=s(config.FEEDBACK_FONT_SIZE),
duration=config.FEEDBACK_TIME,
font_name='DejaVuSans.ttf')
with Meanwhile():
with Parallel():
Ellipse(color='red', size=(s(55), s(55)))
Ellipse(color=config.INNER_CIRCLE_COLOR,
size=(s(45), s(45)))
Label(text=inst['inst_1'],
markup=True,
halign='center',
center_x=self.exp.screen.center_x,
center_y=self.exp.screen.center_y + s(300),
font_size=s(config.INST_FONT_SIZE))
Wait(1.0)
Label(text=inst['inst_2'],
markup=True,
halign='center',
center_x=self.exp.screen.center_x,
center_y=self.exp.screen.center_y + s(300),
font_size=s(config.INST_FONT_SIZE))
with UntilDone():
GetResponse(keys=config.CONT_KEY)
with Loop([[
"__<__\n_<><_\n<>>><\n_<><_\n__<__\n", config.RESP_KEYS[1],
cos(radians((360. / config.NUM_LOCS) * 5)),
sin(radians((360. / config.NUM_LOCS) * 5))
],
[
"__<__\n_<><_\n<><><\n_<><_\n__<__\n",
config.RESP_KEYS[0],
cos(radians((360. / config.NUM_LOCS) * 1)),
sin(radians((360. / config.NUM_LOCS) * 1))
],
[
"__>__\n_><>_\n><><>\n_><>_\n__>__\n",
config.RESP_KEYS[1],
cos(radians((360. / config.NUM_LOCS) * 3.)),
sin(radians((360. / config.NUM_LOCS) * 3.))
],
[
"__>__\n_><>_\n><><>\n_><>_\n__>__\n",
config.RESP_KEYS[1],
cos(radians((360. / config.NUM_LOCS) * 4.)),
sin(radians((360. / config.NUM_LOCS) * 4.))
],
[
"__>__\n_>>>_\n>>>>>\n_>>>_\n__>__\n",
config.RESP_KEYS[1],
cos(radians((360. / config.NUM_LOCS) * 0)),
sin(radians((360. / config.NUM_LOCS) * 0))
],
[
"__>__\n_><>_\n><<<>\n_><>_\n__>__\n",
config.RESP_KEYS[0],
cos(radians((360. / config.NUM_LOCS) * 2)),
sin(radians((360. / config.NUM_LOCS) * 2))
]]) as prac_ev:
Wait(1.0)
with Parallel():
Ellipse(color='red',
size=(s(55), s(55)),
center_x=self.exp.screen.center_x +
prac_ev.current[2] * s(config.FROM_CENTER),
center_y=self.exp.screen.center_y +
prac_ev.current[3] * s(config.FROM_CENTER),
blocking=False)
Ellipse(color=config.INNER_CIRCLE_COLOR,
size=(s(45), s(45)),
center_x=self.exp.screen.center_x +
prac_ev.current[2] * s(config.FROM_CENTER),
center_y=self.exp.screen.center_y +
prac_ev.current[3] * s(config.FROM_CENTER),
blocking=False)
p4 = Trial(config,
stim=prac_ev.current[0],
center_x=self.exp.screen.center_x +
prac_ev.current[2] * s(config.FROM_CENTER),
center_y=self.exp.screen.center_y +
prac_ev.current[3] * s(config.FROM_CENTER),
correct_resp=prac_ev.current[1],
condition='+')
with If(p4.correct):
# They got it right
Label(text=u"\u2713",
color='green',
duration=config.FEEDBACK_TIME,
font_size=s(config.FEEDBACK_FONT_SIZE),
font_name='DejaVuSans.ttf')
with Else():
# they got it wrong
Label(text=u"\u2717",
color='red',
font_size=s(config.FEEDBACK_FONT_SIZE),
duration=config.FEEDBACK_TIME,
font_name='DejaVuSans.ttf')
Label(text=inst['to_cont'],
halign='center',
center_x=self.exp.screen.center_x,
center_y=self.exp.screen.center_y,
font_size=s(config.INST_FONT_SIZE))
with UntilDone():
GetResponse(keys=config.CONT_KEY)
Wait(1.0)
Label(text=inst['inst_3'],
markup=True,
halign='center',
center_x=self.exp.screen.center_x,
center_y=self.exp.screen.center_y + s(200),
font_size=s(config.INST_FONT_SIZE))
with UntilDone():
GetResponse(keys=config.CONT_KEY)
with Loop([[
"__<__\n_<><_\n<>>><\n_<><_\n__<__\n", config.RESP_KEYS[1],
cos(radians((360. / config.NUM_LOCS) * 5)),
sin(radians((360. / config.NUM_LOCS) * 5))
],
[
"__<__\n_<><_\n<><><\n_<><_\n__<__\n",
config.RESP_KEYS[0],
cos(radians((360. / config.NUM_LOCS) * 1)),
sin(radians((360. / config.NUM_LOCS) * 1))
],
[
"__>__\n_><>_\n><><>\n_><>_\n__>__",
config.RESP_KEYS[1],
cos(radians((360. / config.NUM_LOCS) * 3)),
sin(radians((360. / config.NUM_LOCS) * 3))
],
[
"__>__\n_>>>_\n>>>>>\n_>>>_\n__>__\n",
config.RESP_KEYS[1],
cos(radians((360. / config.NUM_LOCS) * 6)),
sin(radians((360. / config.NUM_LOCS) * 6))
],
[
"__>__\n_><>_\n><<<>\n_><>_\n__>__\n",
config.RESP_KEYS[0],
cos(radians((360. / config.NUM_LOCS) * 2)),
sin(radians((360. / config.NUM_LOCS) * 2))
]]) as prac_ev:
Wait(1.0)
p5 = Trial(config,
stim=prac_ev.current[0],
center_x=self.exp.screen.center_x +
prac_ev.current[2] * s(config.FROM_CENTER),
center_y=self.exp.screen.center_y +
prac_ev.current[3] * s(config.FROM_CENTER),
correct_resp=prac_ev.current[1],
condition='+')
with If(p5.correct):
# They got it right
Label(text=u"\u2713",
color='green',
duration=config.FEEDBACK_TIME,
font_size=s(config.FEEDBACK_FONT_SIZE),
center_y=self.exp.screen.center_y + s(50),
font_name='DejaVuSans.ttf')
with Else():
# they got it wrong
Label(text=u"\u2717",
color='red',
font_size=s(config.FEEDBACK_FONT_SIZE),
center_y=self.exp.screen.center_y + s(50),
duration=config.FEEDBACK_TIME,
font_name='DejaVuSans.ttf')
with Meanwhile():
Label(text="+", font_size=s(config.ORIENT_FONT_SIZE))
with Serial(blocking=False):
with ButtonPress():
Button(text=inst['skip_text'],
right=self.exp.screen.width,
bottom=0,
width=s(config.SKIP_SIZE[0]),
height=s(config.SKIP_SIZE[1]),
blocking=False,
font_size=s(config.SKIP_FONT_SIZE))
Wait(1.0)
Label(text=inst['done'],
halign='center',
center_x=self.exp.screen.center_x,
center_y=self.exp.screen.center_y,
font_size=s(config.INST_FONT_SIZE))
with UntilDone():
GetResponse(keys=config.CONT_KEY)
def __init__(self, expparadigm, config, phrases, args):
# TODO: add a 'question' argument
debug("__init__")
Trial.__init__(self, config, args['experiment'], args['item'],
args['condition'], args['mode'],
phrases, len(phrases)+1, question=None)
if expparadigm == "MR-SAT":
self.paradigm = "MR"
elif expparadigm == "SAT":
self.paradigm = "SR"
else:
raise("Invalid experiment paradigm.")
mode = args['mode']
if not args.has_key('speededAcceptability'):
args['speededAcceptability'] = False
#if not args.has_key('lastFeedbackOffset'):
# args['lastFeedbackOffset'] = 0
self.args = args
self.displayPhrase = None
self.lastPhraseRead = None
self.lastResponseCorrect = None
# create phrase chunks
masked_phrases = []
text_phrases = []
space = Text(text=" ", color=(0.0,0.0,0.0), position=(0,0),
font_size=self.config.sprStimulusSize, anchor='left',
on=0, font_name=self.config.font)
space_size = space.parameters.size[0]
# set initial word coordinates
phrase_positions = []
if mode == CPresentationMode.MovingWindow:
x_pos = self.config.xPositionStart
y_pos = globalScreen.size[1]/2
anchor = "left"
elif mode == CPresentationMode.Centered:
x_pos = globalScreen.size[0]/2
y_pos = globalScreen.size[1]/2
anchor = "center"
# create phrases
self.sentence = ""
for i in range(0, len(phrases)):
if phrases[i] == "$":
self.phraseSignalReference = i
continue
self.sentence = self.sentence + " " + phrases[i]
text = Text(text = phrases[i], color = (0.0,0.0,0.0),
position = (x_pos, y_pos),
font_size = self.config.sprStimulusSize,
anchor=anchor, on=0, font_name=self.config.font)
if mode == CPresentationMode.MovingWindow:
phrase_positions.append( (x_pos,y_pos) )
x_pos = x_pos + text.parameters.size[0] + space_size
text_phrases.append(text)
# TODO: Fix y coordinates too. Necessary once the stimulus
# becomes longer than one line.
assert( hasattr(self, "phraseSignalReference"))
self.sentence.strip()
# create masks if the mode is moving window
if mode == CPresentationMode.MovingWindow:
for i in range(0, len(phrases)):
text_size = text_phrases[i].parameters.size
mask_size = (text_size[0], config.sprMaskHeight )
mask_position = [ phrase_positions[i][0],
phrase_positions[i][1] ]
mask_position[1] = mask_position[1]-(text_size[1]/2-mask_size[1]/2)
phrase_mask = Target2D(color=(0.0,0.0,0.0),
position=mask_position,
anchor='left',
on=1, size=mask_size)
masked_phrases.append(phrase_mask)
self.screen = TrialScreenSAT(config, text_phrases,
masked_phrases, mode,
self.args['signalsCnt'],
self.args['feedbackIndicators'],
self.args['mappingIndicators'])
if not self.args['speededAcceptability']:
self.responses = [False]*self.args['signalsCnt']
else:
self.responses = [False]
self._directionBeforeSignal = None
self._directionAfterSignal = None
def do_experiment(self, targets, nontargets, slug):
t = Trial(list(enumerate(targets + nontargets)),
size=self.size,
preview_time=4000,
image_time=self.PRESENTATION_DELAY)
self.start_trial(t)
accumulated_raw_results = []
while self.running:
self.clock.tick(self.FRAMERATE)
for event in pygame.event.get():
if event.type == pygame.QUIT:
self.reset()
return
elif event.type == pygame.KEYDOWN:
if event.key == pygame.K_ESCAPE:
self.reset()
return
elif event.key == pygame.K_SPACE:
self.reset()
t = Trial(list(enumerate(targets + nontargets)),
size=self.size,
preview_time=4000,
image_time=self.PRESENTATION_DELAY)
self.start_trial(t)
elif event.type == self.EVENT_ID:
if self.trial:
pygame.time.set_timer(
self.EVENT_ID,
self.trial.show_next_image(self.screen, self.bci))
pygame.display.flip()
if self.trial.mode == Trial.State.COMPLETED:
self.bci and self.bci.end_block()
results = None
print('Trial completed')
logger.info('completed trial')
results = self.trial.process_results(
self.screen, self.bci)
if results:
self.action_server.set_succeeded(results)
else:
print('Trial insufficient')
pygame.display.flip()
logger.info('raw results:')
raw_results = sorted(self.trial.option_results,
key=lambda x: x.idx)
accumulated_raw_results.append(raw_results)
with open('data/{}.pickle'.format(slug), 'w') as f:
pickle.dump(accumulated_raw_results, f)
logger.info('\n'.join(map(str, raw_results)))
logger.info('processed results:')
logger.info(results)
self.ranking = False
self.trial = None
print('press space to run again')
elif self.trial.mode == Trial.State.ABORTED:
print('Trial aborted')
self.action_server.set_aborted()
self.ranking = False
self.trial = None
self.bci and self.bci.end_block()
self.reset()
from trial import Trial, MyException
import json
trials = []
with open('../data.json') as json_data:
d = json.load(json_data)
for obj in d["data"]:
try:
t = Trial(obj)
except MyException:
continue
t.clean()
trials.append(t)
done = []
done_obj = []
for t in trials:
if t.participant_id not in done:
done_obj.append(t)
done.append(t.participant_id)
def __init__(self, expparadigm, config, phrases, args):
# TODO: add a 'question' argument
debug("__init__")
Trial.__init__(self,
config,
args['experiment'],
args['item'],
args['condition'],
args['mode'],
phrases,
len(phrases) + 1,
question=None)
if expparadigm == "MR-SAT":
self.paradigm = "MR"
elif expparadigm == "SAT":
self.paradigm = "SR"
else:
raise ("Invalid experiment paradigm.")
mode = args['mode']
if not args.has_key('speededAcceptability'):
args['speededAcceptability'] = False
#if not args.has_key('lastFeedbackOffset'):
# args['lastFeedbackOffset'] = 0
self.args = args
self.displayPhrase = None
self.lastPhraseRead = None
self.lastResponseCorrect = None
# create phrase chunks
masked_phrases = []
text_phrases = []
space = Text(text=" ",
color=(0.0, 0.0, 0.0),
position=(0, 0),
font_size=self.config.sprStimulusSize,
anchor='left',
on=0,
font_name=self.config.font)
space_size = space.parameters.size[0]
# set initial word coordinates
phrase_positions = []
if mode == CPresentationMode.MovingWindow:
x_pos = self.config.xPositionStart
y_pos = globalScreen.size[1] / 2
anchor = "left"
elif mode == CPresentationMode.Centered:
x_pos = globalScreen.size[0] / 2
y_pos = globalScreen.size[1] / 2
anchor = "center"
# create phrases
self.sentence = ""
for i in range(0, len(phrases)):
if phrases[i] == "$":
self.phraseSignalReference = i
continue
self.sentence = self.sentence + " " + phrases[i]
text = Text(text=phrases[i],
color=(0.0, 0.0, 0.0),
position=(x_pos, y_pos),
font_size=self.config.sprStimulusSize,
anchor=anchor,
on=0,
font_name=self.config.font)
if mode == CPresentationMode.MovingWindow:
phrase_positions.append((x_pos, y_pos))
x_pos = x_pos + text.parameters.size[0] + space_size
text_phrases.append(text)
# TODO: Fix y coordinates too. Necessary once the stimulus
# becomes longer than one line.
assert (hasattr(self, "phraseSignalReference"))
self.sentence.strip()
# create masks if the mode is moving window
if mode == CPresentationMode.MovingWindow:
for i in range(0, len(phrases)):
text_size = text_phrases[i].parameters.size
mask_size = (text_size[0], config.sprMaskHeight)
mask_position = [
phrase_positions[i][0], phrase_positions[i][1]
]
mask_position[1] = mask_position[1] - (text_size[1] / 2 -
mask_size[1] / 2)
phrase_mask = Target2D(color=(0.0, 0.0, 0.0),
position=mask_position,
anchor='left',
on=1,
size=mask_size)
masked_phrases.append(phrase_mask)
self.screen = TrialScreenSAT(config, text_phrases, masked_phrases,
mode, self.args['signalsCnt'],
self.args['feedbackIndicators'],
self.args['mappingIndicators'])
if not self.args['speededAcceptability']:
self.responses = [False] * self.args['signalsCnt']
else:
self.responses = [False]
self._directionBeforeSignal = None
self._directionAfterSignal = None
path = args.restore_path
full_path = gen_full_path(subdir, path, unique=False)
init_logger(full_path, args.debug)
vae, agent, environment, params = restore_experiment(
full_path, env_class, args)
trigger = params['env_params'].get('trigger')
if (trigger == 'None') or (trigger == ''):
trigger = None
if trigger is not None:
trigger_class = importlib.import_module(subdir + ".triggers")
trigger = trigger_class.__dict__[trigger](
params['env_params']['trigger_step'])
train = not params['test']
trial = Trial(full_path, vae, agent, environment, trigger)
try:
update_step = params['concurrent_batches'] * params['train_step']
trial.run(params['num_steps'], params['summary_step'], update_step, train,
params['record'])
except Exception as e:
logging.exception("Exception in trial.")
finally:
try:
trial.agent.lookup.table.shutdown()
trial.model.exit_signal.set()
except AttributeError:
pass
try:
trial.vae.sess.close()
except:
