malar
/
speech-scoring
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

copying constantws to models

master
Malar Kannan 2017-11-20 15:15:27 +05:30
parent a5d4ede35d
commit d2a075422c
2 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
speech_data.py
View File

@ -8,7 +8,7 @@ import numpy as np
from speech_spectrum import generate_aiff_spectrogram from speech_spectrum import generate_aiff_spectrogram
from sklearn.model_selection import train_test_split from sklearn.model_selection import train_test_split
import itertools import itertools
import os import os,shutil
import random import random
import csv import csv
import gc import gc
@ -144,6 +144,9 @@ def read_siamese_tfrecords_generator(audio_group='audio',batch_size=32,test_size
const_file = os.path.join('./outputs',audio_group+'.constants') const_file = os.path.join('./outputs',audio_group+'.constants')
(n_spec,n_features,n_records) = pickle.load(open(const_file,'rb')) (n_spec,n_features,n_records) = pickle.load(open(const_file,'rb'))
def copy_read_consts(dest_dir):
shutil.copy2(const_file,dest_dir)
return (n_spec,n_features,n_records)
# @threadsafe_iter # @threadsafe_iter
def record_generator(): def record_generator():
print('reading tfrecords({}-train)...'.format(audio_group)) print('reading tfrecords({}-train)...'.format(audio_group))
@ -195,7 +198,7 @@ def read_siamese_tfrecords_generator(audio_group='audio',batch_size=32,test_size
output = example.features.feature['output'].int64_list.value output = example.features.feature['output'].int64_list.value
output_data[i] = np.asarray(output) output_data[i] = np.asarray(output)
return record_generator,input_data,output_data,n_spec,n_features,n_records return record_generator,input_data,output_data,copy_read_consts
def audio_samples_word_count(audio_group='audio'): def audio_samples_word_count(audio_group='audio'):
audio_samples = pd.read_csv( './outputs/' + audio_group + '.csv') audio_samples = pd.read_csv( './outputs/' + audio_group + '.csv')
@ -249,7 +252,7 @@ if __name__ == '__main__':
# create_spectrogram_tfrecords('story_all',sample_count=25) # create_spectrogram_tfrecords('story_all',sample_count=25)
# fix_csv('story_words_test') # fix_csv('story_words_test')
#fix_csv('story_phrases') #fix_csv('story_phrases')
create_spectrogram_tfrecords('story_phrases',sample_count=0,train_test_ratio=0.1) create_spectrogram_tfrecords('story_phrases',sample_count=100,train_test_ratio=0.1)
# create_spectrogram_tfrecords('audio',sample_count=50) # create_spectrogram_tfrecords('audio',sample_count=50)
# read_siamese_tfrecords_generator('audio') # read_siamese_tfrecords_generator('audio')
# padd_zeros_siamese_tfrecords('audio') # padd_zeros_siamese_tfrecords('audio')

3
speech_model.py
View File

@ -74,7 +74,8 @@ def train_siamese(audio_group = 'audio'):
create_dir(model_dir) create_dir(model_dir)
log_dir = './logs/'+audio_group log_dir = './logs/'+audio_group
create_dir(log_dir) create_dir(log_dir)
tr_gen_fn,te_pairs,te_y,n_step,n_features,n_records = read_siamese_tfrecords_generator(audio_group,batch_size=batch_size,test_size=batch_size) tr_gen_fn,te_pairs,te_y,copy_read_consts = read_siamese_tfrecords_generator(audio_group,batch_size=batch_size,test_size=batch_size)
n_step,n_features,n_records = copy_read_consts()
tr_gen = tr_gen_fn() tr_gen = tr_gen_fn()
input_dim = (n_step, n_features) input_dim = (n_step, n_features)