def __init__(self, q, batch_iter): super(BatchProducer, self).__init__() threading.Thread.__init__(self) self.q = q self.batch_iter = batch_iter self.log = logger.get() self._stop = threading.Event() self.daemon = True
def __init__(self,
batch_iter,
max_queue_size=10,
num_threads=5,
log_queue=20,
name=None):
"""
Data provider wrapper that supports concurrent data fetching.
"""
super(ConcurrentBatchIterator, self).__init__()
self.max_queue_size = max_queue_size
self.num_threads = num_threads
self.q = queue.Queue(maxsize=max_queue_size)
self.log = logger.get()
self.batch_iter = batch_iter
self.fetchers = []
self.init_fetchers()
self.counter = 0
self.relaunch = True
self._stopped = False
self.log_queue = log_queue
self.name = name
pass
"""Unit tests for multi-pass model."""
from __future__ import (absolute_import, division, print_function,
unicode_literals)
import os
import numpy as np
import tensorflow as tf
from resnet.configs import get_config
from resnet.configs import test_configs
from resnet.models.model_factory import get_model, get_multi_gpu_model
from resnet.models.multi_pass_model import MultiPassModel
from resnet.models.resnet_model import ResNetModel
from resnet.utils import logger
log = logger.get()
FOLDER = "tmp"
CKPT_FNAME = os.path.join(FOLDER, "test_multi_pass.ckpt")
class MultiPassModelTests(tf.test.TestCase):
def _test_single_pass(self, method):
config = get_config("resnet-test")
config.momentum = 0.0
config.base_learn_rate = 1e-1
np.random.seed(0)
BSIZE = config.batch_size
xval = np.random.uniform(
-1.0, 1.0, [BSIZE, config.height, config.width,
config.num_channel]).astype(np.float32)
yval = np.floor(np.random.uniform(0, 9.9, [BSIZE])).astype(np.int32)
