ADNC/adnc/data/loader.py
2018-07-11 14:35:33 +02:00

137 lines
4.7 KiB
Python

# Copyright 2018 Jörg Franke
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
import threading
from queue import Queue
import numpy as np
from adnc.data.utils.batch_generator import BatchGenerator
from adnc.data.tasks.repeat_copy import CopyTask
from adnc.data.tasks.cnn_rc import ReadingComprehension
from adnc.data.tasks.babi import bAbI
class DataLoader():
"""
The data loader loads and process the datasets and provides iterators for training or inference.
"""
def __init__(self, config, word_dict=None, re_word_dict=None):
"""
Args:
config: dict with the config to pre-process the dataset
word_dict: dict with word-feature pairs, optional
re_word_dict: dict with feature-word pairs, optional
"""
self.config = config
if config['data_set'] == 'copy_task':
self.dataset = CopyTask(self.config)
elif config['data_set'] == 'cnn':
self.dataset = ReadingComprehension(self.config)
elif config['data_set'] == 'babi':
self.dataset = bAbI(self.config, word_dict, re_word_dict)
@property
def vocabulary_size(self):
return self.dataset.vocabulary_size
@property
def x_size(self):
return self.dataset.x_size
@property
def y_size(self):
return self.dataset.y_size
def batch_amount(self, set_name):
"""
Calculates the batch amount given a batch size
Args:
set_name: str, name of dataset (train, test, valid)
Returns: int, number of batches
"""
if 'max_len' in self.config.keys():
return np.floor(
self.dataset.sample_amount(set_name, self.config['max_len']) / self.config['batch_size']).astype(int)
else:
return np.floor(self.dataset.sample_amount(set_name) / self.config['batch_size']).astype(int)
def sample_amount(self, set_name, ):
return self.dataset.sample_amount(set_name)
def get_sample(self, set, number):
return self.dataset.get_sample(set, number)
def decode_output(self, sample, prediction):
return self.dataset.decode_output(sample, prediction)
def get_data_loader(self, set_name, shuffle=True, max_len=False, batch_size=None, get_shuffle_option=False):
"""
Provides a data iterator of the given dataset.
Args:
set_name: str, name of dataset
shuffle: bool, shuffle set or not
max_len: int, max length in time of sample
batch_size: int, batch size
get_shuffle_option: bool, returns shuffle function
Returns: iter, iterator over dataset
"""
if batch_size == None:
batch_size = self.config['batch_size']
stream_loader_pre = BatchGenerator(self.dataset, set_name, batch_size, shuffle=shuffle, max_len=max_len)
stream_loader = self._generate_in_background(stream_loader_pre, num_cached=self.config['num_chached'],
threads=self.config['threads'])
if get_shuffle_option:
return stream_loader, stream_loader_pre.shuffle_order
else:
return stream_loader
@staticmethod
def _generate_in_background(batch_gen, num_cached=10, threads=1):
"""
Starts threads with parallel batch generator for faster iteration
Args:
batch_gen: func, the batch generator
num_cached: int, numb of caches batches
threads: int, numb of parallel threads
Returns: iter, iterator over dataset
"""
queue = Queue(maxsize=num_cached)
sentinel = object()
def producer():
for item in batch_gen:
queue.put(item)
queue.put(sentinel)
threads = [threading.Thread(target=producer) for _ in range(threads)]
for t in threads:
t.daemon = True
t.start()
item = queue.get()
while item is not sentinel:
yield item
item = queue.get()