def evaluate(evalset, config, checkpoint=None):
assert (evalset in ['validation', 'test'])
logger_config = {
'owner': 'Training.evaluate',
'log_dir': config['log_dir'],
'timezone': 'UTC',
}
logger = utils.Logger(logger_config)
_trainer = get_trainer(config) # 调用本py文件中函数get_trainer↑
if checkpoint is None: # 检查是否有可用checkpoint
checkpoints = sorted(glob.glob(_trainer.checkpoints_dir + '/*.pt'))
assert (len(checkpoints))
checkpoint = checkpoints[-1]
_trainer.load(checkpoint) # 导入checkpoint
#------------分为train，test，validation.根据选用的是validation or test执行以下步骤--------
if evalset == 'validation':
valid_loader_config = {
'batch_size': config['batch_size'],
'images_path': config['images_validation_path'],
'formulas_path': config['formulas_validation_path'],
}
loader = utils.data_loader(_trainer.train_loader.vocab,
valid_loader_config) # 导入validation数据
elif evalset == 'test':
test_loader_config = {
'batch_size': config['batch_size'],
'images_path': config['images_test_path'],
}
loader = utils.data_loader(_trainer.train_loader.vocab,
test_loader_config) # 导入test数据
#------------有label的情况下，计算validation的acc等，没有label，test不进行计算，仅进行预测---------
#-----------实际上，validation的计算过程跟train一样；test只进行bleu和edit_distance的评估---------
if loader.has_label: # 是否存在validation_path
predictions, loss, acc = _trainer.evaluate(loader,
config['generation_method'])
logger('loss={}, acc={}'.format(loss, acc))
else:
predictions = _trainer.evaluate(loader, config['generation_method'])
# 确定是test/validation的formula_path
target_path = config[
'formulas_validation_path'] if evalset == 'validation' else config[
'formulas_test_path']
predicted_path = 'tmp/predicted-{}.txt'.format(evalset) # 预测的fomula_path
bleu_message, edit_message = bleu_and_edit_distance(
predictions, predicted_path, target_path) # 评估↓
logger(bleu_message)
logger(edit_message)
def evaluate(evalset, config, checkpoint=None):
assert (evalset in ['validation', 'test'])
logger_config = {
'owner': 'Training.evaluate',
'log_dir': config['log_dir'],
'timezone': 'UTC',
}
logger = utils.Logger(logger_config)
_trainer = get_trainer(config)
if checkpoint is None:
checkpoints = sorted(glob.glob(_trainer.checkpoints_dir + '/*.pt'))
assert (len(checkpoints))
checkpoint = checkpoints[-1]
_trainer.load(checkpoint)
if evalset == 'validation':
valid_loader_config = {
'batch_size': config['batch_size'],
'images_path': config['images_validation_path'],
'formulas_path': config['formulas_validation_path'],
}
loader = utils.data_loader(_trainer.train_loader.vocab,
valid_loader_config)
elif evalset == 'test':
test_loader_config = {
'batch_size': config['batch_size'],
'images_path': config['images_test_path'],
}
loader = utils.data_loader(_trainer.train_loader.vocab,
test_loader_config)
if loader.has_label:
predictions, attn_weights, loss, acc = _trainer.evaluate(
loader, config['generation_method'])
logger('loss={}, acc={}'.format(loss, acc))
else:
predictions, attn_weights = _trainer.evaluate(
loader, config['generation_method'])
target_path = config[
'formulas_validation_path'] if evalset == 'validation' else config[
'formulas_test_path']
predicted_path = 'tmp/predicted-{}.txt'.format(evalset)
bleu_message, edit_message = bleu_and_edit_distance(
predictions, predicted_path, target_path)
logger(bleu_message)
logger(edit_message)
def __init__(self, instance_id):
UTIL.Logger("TI_Gateway", "ti_gateway.log", logging.DEBUG,
logging.DEBUG)
self.logger = logging.getLogger("TI_Gateway")
self.rest_switches = REST.OpenHabRestInterface(
"192.168.178.24", "8080", "pi", "raspberry",
"%s_device_switch_group" % self.prefix, self.switch_queue)
self.rest_switches.daemon = True # REST classes can be daemonized
self.rest_switches.set_logger("TI_Gateway")
self.rest_values = REST.OpenHabRestInterface("192.168.178.24", "8080",
"pi", "raspberry",
"%s_values_group",
self.value_queue)
self.rest_values.daemon = True # REST classes can be daemonized
self.rest_values.set_logger("TI_Gateway")
self.sensortag = TIInterface("24:71:89:BC:1D:01", self.ble_queue, 1.0,
self.update)
self.sensortag.daemon = True
self.sensortag.set_logger("TI_Gateway")
self.sensortag.start()
# Create Switch items
self.rest_switches.add_item("%s_device_switch_group" % self.prefix,
"Group",
"Group of Switches for %s" % self.prefix,
"rest", "")
self.rest_switches.add_item("%s_device_switch" % self.prefix, "Switch",
"Switch for %s" % self.prefix, "rest",
"%s_device_switch_group" % self.prefix)
self.rest_switches.update_item_state("%s_device_switch" % self.prefix,
"ON")
self.rest_switches.start()
# Create Value items
self.rest_values.add_item("%s_values_group" % self.prefix, "Group",
"Group of Values for %s" % self.prefix,
"rest", "")
for service_id in self.sensortag.get_service_ids():
self.rest_values.add_item("%s_%s" % (self.prefix, service_id),
"String", service_id, "rest",
"%s_values" % self.prefix)
def train(config):
logger_config = {
'owner': 'Training.train',
'log_dir': config['log_dir'],
'timezone': 'UTC',
}
logger = utils.Logger(logger_config)
_trainer = get_trainer(config)
checkpoints = sorted(glob.glob(_trainer.checkpoints_dir + '/*.pt'))
print('checkpoints')
if len(checkpoints):
_trainer.load(checkpoints[-1])
last_epoch = _trainer.current_epoch
print('last_epoch')
predicted_path = 'tmp/predicted-train.txt'
# if shuffled or sorted by length -> should update target_path accordingly
target_path = config['formulas_train_path']
dot = target_path.rfind('.')
target_path = target_path[:dot] + '_' + target_path[dot:]
if _trainer.train_loader.shuffle or _trainer.train_loader.sort_by_formulas_len:
f = open(target_path, 'w')
for formula in _trainer.train_loader.formulas:
joinedformula = ' '.join(token for token in formula)
f.write(joinedformula + '\n')
f.close()
for epoch in range(last_epoch, config['epochs']):
predictions, epoch_loss, epoch_acc = _trainer.train_one_epoch()
checkpoint = _trainer.save(epoch + 1, epoch_loss, epoch_acc)
bleu_message, edit_message = bleu_and_edit_distance(
predictions, predicted_path, target_path)
logger(bleu_message)
logger(edit_message)
evaluate('validation', config, checkpoint)
def main(unused_args):
output_dir = FLAGS.output_dir
print('-' * 80)
if not gfile.IsDirectory(output_dir):
print('Path {} does not exist. Creating'.format(output_dir))
gfile.MakeDirs(output_dir)
elif FLAGS.reset_output_dir:
print('Path {} exists. Reseting'.format(output_dir))
gfile.DeleteRecursively(output_dir)
gfile.MakeDirs(output_dir)
print('-' * 80)
log_file = os.path.join(output_dir, 'stdout')
print('Logging to {}'.format(log_file))
sys.stdout = utils.Logger(log_file)
params = tf.contrib.training.HParams(
data_path=FLAGS.data_path,
log_every=FLAGS.log_every,
output_dir=FLAGS.output_dir,
)
train(params)
parser.add_argument('--use_feature', default='bottom', type=str, help='use bottom-up feature or grid feature')
# SAN
parser.add_argument('--num_stacks', default=2, type=int,
help='num of stacks in Stack Attention Networks')
args = parser.parse_args()
return args
if __name__ == '__main__':
args = parse_args()
utils.create_dir(args.output)
logger = utils.Logger(os.path.join(args.output, 'log.txt'))
logger.write(args.__repr__())
# Initializes the distributed backend which will take care of sychronizing nodes/GPUs
device = torch.device("cuda:" + str(args.gpu) if args.gpu >= 0 else "cpu")
args.device = device
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
torch.backends.cudnn.benchmark = True
torch.cuda.set_device(args.gpu)
dictionary = Dictionary.load_from_file('data_v7w/dictionary.pkl')
train_dset = V7WDataset('train', args, dictionary, adaptive=True, max_boxes=args.max_boxes,
question_len=args.question_len)
val_dset = V7WDataset('val', args, dictionary, adaptive=True, max_boxes=args.max_boxes,
question_len=args.question_len)
import sys
sys.path.append('.') # if user is outside tests dir.
sys.path.append("..") # if user is inside tests dir
import unittest
from src import utils, other_ciphers
logger = utils.Logger(print)
def _test_cipher(self, encryptor, decryptor, _121=True):
"""
Tests single cipher text algorithm for correctness.
_121: whether algo gives one text op for one text ip.
"""
strings = [utils.get_rand_ascii_str() for i in range(20)]
strings += ['']
for string in strings:
ct = encryptor(string)
pt = decryptor(ct)
if _121:
self.assertEqual(len(ct), len(string))
self.assertEqual(pt, string.upper())
class Tests(unittest.TestCase):
def test_shift_chars(self):
# shift char gets the element after displacing elements with dx.
ip_op = {('a', 2): 'c', ('b', -1): 'a', ('z', 3): 'c', ('c', -4): 'y'}
for ip, op in ip_op.items():
def train(args,
model,
train_loader,
eval_loader,
num_epochs,
output,
opt=None,
s_epoch=0):
device = args.device
lr_default = args.lr
lr_decay_step = 2
lr_decay_rate = .25
lr_decay_epochs = range(
10, 20, lr_decay_step) if eval_loader is not None else range(
10, 20, lr_decay_step)
gradual_warmup_steps = [
0.5 * lr_default, 1.0 * lr_default, 1.5 * lr_default, 2.0 * lr_default
]
saving_epoch = 0
grad_clip = args.clip_norm
utils.create_dir(output)
optim = torch.optim.Adamax(filter(lambda p: p.requires_grad, model.parameters()), lr=lr_default) \
if opt is None else opt
# Initial loss function
criterion = torch.nn.BCEWithLogitsLoss(reduction='sum')
logger = utils.Logger(os.path.join(output, 'log.txt'))
logger.write(args.__repr__())
best_eval_score = 0
utils.print_model(model, logger)
logger.write('optim: adamax lr=%.4f, decay_step=%d, decay_rate=%.2f, grad_clip=%.2f' % \
(lr_default, lr_decay_step, lr_decay_rate, grad_clip))
# Create trainer
trainer = Trainer(args, model, criterion, optim)
update_freq = int(args.update_freq)
wall_time_start = time.time()
for epoch in range(s_epoch, num_epochs):
total_loss = 0
train_score = 0
total_norm = 0
count_norm = 0
num_updates = 0
t = time.time()
N = len(train_loader.dataset)
num_batches = int(N / args.batch_size + 1)
if epoch < len(gradual_warmup_steps):
trainer.optimizer.param_groups[0]['lr'] = gradual_warmup_steps[
epoch]
logger.write('gradual warmup lr: %.8f' %
trainer.optimizer.param_groups[0]['lr'])
elif epoch in lr_decay_epochs:
trainer.optimizer.param_groups[0]['lr'] *= lr_decay_rate
logger.write('decreased lr: %.8f' %
trainer.optimizer.param_groups[0]['lr'])
else:
logger.write('lr: %.8f' % trainer.optimizer.param_groups[0]['lr'])
for i, (v, b, q, a, ans_mc, ans_gt) in enumerate(train_loader):
v = v.to(device)
b = b.to(device)
q = q.to(device)
a = a.to(device)
ans_mc = ans_mc.to(device)
# Clone each sample to 4 samples
v = v.unsqueeze(1).expand(v.size(0), 4, v.size(1),
v.size(2)).contiguous().view(
v.size(0) * 4, v.size(1), v.size(2))
q = q.unsqueeze(1).expand(q.size(0), 4,
q.size(1)).contiguous().view(
q.size(0) * 4, q.size(1))
ans_mc = ans_mc.view(
ans_mc.size(0) * ans_mc.size(1), ans_mc.size(2))
a = a.view(ans_mc.size(0), 1)
labels = torch.cat([a, 1 - a], 1)
labels = labels.to(device)
sample = [v, b, q, labels, ans_mc]
if i < num_batches - 1 and (i + 1) % update_freq > 0:
trainer.train_step(sample, update_params=False)
else:
loss, grad_norm, batch_score = trainer.train_step(
sample, update_params=True)
total_norm += grad_norm
count_norm += 1
total_loss += loss.item()
train_score += batch_score
num_updates += 1
if num_updates % int(args.print_interval / update_freq) == 0:
print(
"Iter: {}, Loss {:.4f}, Norm: {:.4f}, Total norm: {:.4f}, Num updates: {}, Wall time: {:.2f},"
" ETA: {}".format(i + 1,
total_loss / ((num_updates + 1)),
grad_norm, total_norm, num_updates,
time.time() - wall_time_start,
utils.time_since(t,
i / num_batches)))
total_loss /= num_updates
train_score = 100 * train_score / (num_updates * args.batch_size)
if eval_loader is not None:
print("Evaluating...")
trainer.model.train(False)
eval_score, bound = evaluate(model, eval_loader, args)
trainer.model.train(True)
logger.write('epoch %d, time: %.2f' % (epoch, time.time() - t))
logger.write('\ttrain_loss: %.2f, norm: %.4f, score: %.2f' %
(total_loss, total_norm / count_norm, train_score))
if eval_loader is not None:
logger.write('\teval score: %.2f (%.2f)' %
(100 * eval_score, 100 * bound))
# Save per epoch
if epoch >= saving_epoch:
model_path = os.path.join(output, 'model_epoch%d.pth' % epoch)
utils.save_model(model_path, model, epoch, trainer.optimizer)
# Save best epoch
if eval_loader is not None and eval_score > best_eval_score:
model_path = os.path.join(output, 'model_epoch_best.pth')
utils.save_model(model_path, model, epoch, trainer.optimizer)
best_eval_score = eval_score
# Configurations
CURRENT_DIR = os.path.dirname(os.path.abspath(__file__))
STATION_CONFIG_DIR = os.path.join(CURRENT_DIR, '../config/station_config.json')
FULL_STATION_CONFIG_DIR = os.path.join(CURRENT_DIR,
'../config/full_station_config.json')
FORECAST_STATION_CONFIG_DIR = os.path.join(
CURRENT_DIR, '../config/forecast_station_config.json')
DB_CONFIG_DIR = os.path.join(CURRENT_DIR, '../config/db_config.json')
WEATHER_CODE_DIR = os.path.join(CURRENT_DIR, '../config/weather_code.json')
# Constants
ERROR_FETCH = 1
SUCCESS = 0
# logger config in this module
_logger = utils.Logger(__name__)
_db_config = utils.parse_json_file(DB_CONFIG_DIR)
_station_config = utils.parse_json_file(STATION_CONFIG_DIR)
_full_station_config = utils.parse_json_file(FULL_STATION_CONFIG_DIR)
_forecast_config = utils.parse_json_file(FORECAST_STATION_CONFIG_DIR)
def fetch_weather_data_of_site(grid_id):
"""
Deprecated. Fetch weather data of a site by the site's grid_id from hko
:param grid_id: grid id of a station, you can find the grid id in the station_config.json file
:return:
a dict containing weather data, if return a string, it means an error occurs when requesting data
"""
if not isinstance(grid_id, str):
grid_id = str(grid_id)