def experiment(exp_name):
onehot_words, word_pos, sentence_lens_nchars, sentence_lens_nwords, vocabulary_size, max_char_len = read(
file_name=
'/Users/kevindsouza/Documents/UBC/Research/ML/VaeLM/data/english/train.h5',
train=True)
onehot_words_val, word_pos_val, sentence_lens_nchars_val, sentence_lens_nwords_val, _, _ = read(
file_name=
'/Users/kevindsouza/Documents/UBC/Research/ML/VaeLM/data/english/test.h5',
train=False)
max_char_len = 371
batch_size = 20
hidden_size = 256
decoder_dim = 256
decoder_units_p3 = 256
vocabulary = ["<SOS>"] + ["a"] + ["b"] + ["c"] + ["d"] + ["e"] + ["f"] + \
["g"] + ["h"] + ["i"] + ["j"] + ["k"] + ["l"] + ["m"] + ["n"] + ["o"] + \
["p"] + ["q"] + ["r"] + ["s"] + ["t"] + ["u"] + ["v"] + ["w"] + \
["x"] + ["y"] + ["z"] + ["<EOW>"] + ["<EOS>"] + [">"] + ["-"] + ["."] + ["'"] + ["0"] + ["1"] + [
"2"] + ["3"] + \
["4"] + ["5"] + ["6"] + ["7"] + ["8"] + ["9"] + ["&"] + ["<"] + ["$"] + ["#"] + ["/"] + [","] + ["|"] + \
["@"] + ["%"] + ["^"] + ["\\"] + ["*"] + ["("] + [")"] + ["{"] + ["}"] + [":"] + [";"]
vocabulary_size = len(vocabulary)
# token2index = {token:index for index,token in enumerate(vocabulary)}
index2token = {index: token for index, token in enumerate(vocabulary)}
train_dict = {
'decoder_units_p3': decoder_units_p3,
'batch_size': batch_size,
'hidden_size': hidden_size,
'decoder_dim': decoder_dim,
'max_char_len': max_char_len,
'onehot_words': onehot_words,
'word_pos': word_pos,
'sentence_lens_nchars': sentence_lens_nchars,
'vocabulary_size': vocabulary_size,
'sentence_lens_nwords': sentence_lens_nwords,
'onehot_words_val': onehot_words_val,
'word_pos_val': word_pos_val,
'sentence_lens_nchars_val': sentence_lens_nchars_val,
'sentence_lens_nwords_val': sentence_lens_nwords_val
}
network_dict = {
'vocabulary_size': vocabulary_size,
'max_char_len': max_char_len,
'batch_size': batch_size,
'hidden_size': hidden_size
}
print("start")
train(log_dir='/Users/kevindsouza/Documents/UBC/Research/ML/VaeLM/log/',
n_epochs=500,
network_dict=network_dict,
index2token=index2token,
mode="new_model",
**train_dict)
print("end")
def experiment(exp_name):
from decoder import Decoder
import encoder
import tensorflow as tf
import numpy as np
from functools import reduce
import tensorflow as tf
import numpy as np
from tensorflow.python.ops.rnn import _transpose_batch_time
import collections
import os
import argparse
import datetime as dt
from collections import Counter
from random import random
from nltk import word_tokenize
from train_vaeLM import train, prep_perm_matrix, permute_encoder_output
from pre import read
onehot_words, word_pos, sentence_lens_nchars, sentence_lens_nwords, vocabulary_size, max_char_len = read(
file_name='/home/zalperst/PycharmProjects/vae_proj/train.h5', train=True)
onehot_words_val, word_pos_val, sentence_lens_nchars_val, sentence_lens_nwords_val, _, _ = read(
file_name='/home/zalperst/PycharmProjects/vae_proj/test.h5', train=False)
max_char_len = 371
batch_size = 40
hidden_size = 256
decoder_dim = 256
decoder_units_p3 = 256
vocabulary = ["<SOS>"] + ["a"] + ["b"] + ["c"] + ["d"] + ["e"] + ["f"] + \
["g"] + ["h"] + ["i"] + ["j"] + ["k"] + ["l"] + ["m"] + ["n"] + ["o"] + \
["p"] + ["q"] + ["r"] + ["s"] + ["t"] + ["u"] + ["v"] + ["w"] + \
["x"] + ["y"] + ["z"] + ["<EOW>"] + ["<EOS>"] + [">"] + ["-"] + ["."] + ["'"] + ["0"] + ["1"] + [
"2"] + ["3"] + \
["4"] + ["5"] + ["6"] + ["7"] + ["8"] + ["9"] + ["&"] + ["<"] + ["$"] + ["#"] + ["/"] + [","] + ["|"] + \
["@"] + ["%"] + ["^"] + ["\\"] + ["*"] + ["("] + [")"] + ["{"] + ["}"] + [":"] + [";"]
vocabulary_size = len(vocabulary)
# token2index = {token:index for index,token in enumerate(vocabulary)}
index2token = {index: token for index, token in enumerate(vocabulary)}
train_dict = {'decoder_units_p3': decoder_units_p3, 'batch_size': batch_size, 'hidden_size': hidden_size,
'decoder_dim': decoder_dim, 'max_char_len': max_char_len, 'onehot_words': onehot_words,
'word_pos': word_pos, 'sentence_lens_nchars': sentence_lens_nchars,
'vocabulary_size': vocabulary_size, 'sentence_lens_nwords': sentence_lens_nwords,
'onehot_words_val': onehot_words_val, 'word_pos_val': word_pos_val,
'sentence_lens_nchars_val': sentence_lens_nchars_val,
'sentence_lens_nwords_val': sentence_lens_nwords_val}
network_dict = {'vocabulary_size': vocabulary_size, 'max_char_len': max_char_len, 'batch_size': batch_size,
'hidden_size': hidden_size}
print("start")
train(log_dir='/home/zalperst/PycharmProjects/vae_proj/', n_epochs=500, network_dict=network_dict,
index2token=index2token, **train_dict)
print("end")
'sentence_lens_nwords': sentence_lens_nwords,
'onehot_words': onehot_words_val,
'word_pos': word_pos_val,
'sentence_lens_nchars': sentence_lens_nchars_val,
'sentence_lens_nwords_val': sentence_lens_nwords_val
}
network_dict = {
'max_char_len': max_char_len,
'batch_size': batch_size,
'hidden_size': hidden_size
}
log("start")
train(n_epochs=1,
network_dict=network_dict,
index2token=index2token,
**train_dict)
log("end")
if __name__ == '__main__':
exp_name = 'vae_model'
pool = WorkerPool(1, worker_types='gpu_gtx_1080_ti')
pool.attach_files([
'train_vaeLM.py', 'decoder.py', 'encoder.py', 'cluster_print.py',
'create_h5.py', h5py
])
pool(experiment, [exp_name])