def __init__(self): self.Config = main.Config() self.vocab_size = self.Config.vocab_size self.special_chars = self.Config.special_chars self.num_of_questions = self.Config.total_examples self.num_of_paragraphs = self.Config.num_of_paragraphs self.glove_dimensionality = self.Config.glove_dimensionality self.largest_num_of_sentences = 0 self.largest_num_of_words = 0 self.glove_lookup = self.initialise_glove_embeddings() self.glove_lookup_dict = {} for entry in self.glove_lookup: index = entry[0] vector = entry[1] self.glove_lookup_dict[index] = vector self.glove_lookup_dict_reversed = {} self.questions_list, self.paragraphs_list, self.answers_list = self.read_squad( ) self.largest_num_of_sentences, self.largest_num_of_words, self.largest_num_of_words_any_paragraph = self.count_words_paragraphs_in_squad( ) self.largest_num_of_words_in_answer = self.get_largest_num_of_words_in_answer( )
def __init__(self): self.Config = main.Config() self.d = self.Config.d self.num_of_batches = self.Config.num_of_batches self.l_rate = self.Config.l_rate self.total_examples = self.Config.total_examples self.examples_per_batch = self.total_examples / self.num_of_batches self.clip_norm = self.Config.clip_norm self.num_of_epochs = self.Config.num_of_epochs self.util = Data.Util() self.unk_answer = self.util.get_one_hot_encoded_from_glove("<unk>") self.largest_num_of_words_any_paragraph = self.util.largest_num_of_words_any_paragraph self.largest_num_of_words_in_answer = self.util.get_largest_num_of_words_in_answer( ) self.largest_num_of_words_in_question = self.util.get_largest_num_of_words_in_question( ) self.question = tf.placeholder( tf.float32, shape=(self.largest_num_of_words_in_question, self.util.glove_dimensionality), name="question") self.text = tf.placeholder( tf.float32, shape=(self.largest_num_of_words_any_paragraph, self.util.glove_dimensionality), name="text") self.answer = tf.placeholder( tf.float32, shape=(self.largest_num_of_words_in_answer, self.util.vocab_size))
import tensorflow as tf import numpy as np import main Config = main.Config() glove_dimensionality = Config.glove_dimensionality d = Config.d class Memory: def __init__(self, text, util): self.A = tf.Variable(tf.random_normal([ util.largest_num_of_words_any_paragraph, util.glove_dimensionality, d ], stddev=0.1), name="A") self.C = tf.Variable(tf.random_normal([ util.largest_num_of_words_any_paragraph, util.glove_dimensionality, d ], stddev=0.1), name="C") self.m = tf.squeeze(tf.matmul(text, self.A), 1) self.c = tf.squeeze(tf.matmul(text, self.C), 1)
#!/usr/bin/env python3 # -*- coding: utf-8 -*- """ Created on Thu May 10 15:48:14 2018 @author: hellcat """ import main import tensorflow as tf from general_net import net import matplotlib.pyplot as plt import numpy as np from PIL import Image opt = main.Config() slim = tf.contrib.slim config = tf.ConfigProto() config.gpu_options.allow_growth = True text_img = './000000000036.jpg' model_path = "./logs/model" img_raw = tf.gfile.FastGFile(text_img, 'rb').read() img = tf.image.decode_jpeg(img_raw) if img.dtype != tf.float32: img = tf.image.convert_image_dtype(img, dtype=tf.float32) generated = net(tf.expand_dims(img, axis=0), training=False) with tf.Session() as sess: saver = tf.train.Saver()
def test_init(self): config = main.Config('testserver', 'testnick') self.assertEqual(config.nickname, 'testnick') self.assertEqual(config.server, 'testserver')