train.py

#!/usr/bin/env python3

import os
# os.add_dll_directory(os.path.join(os.environ['CUDA_PATH'], 'bin'))
from multiprocessing import cpu_count
import tensorflow as tf
# gpus = tf.config.experimental.list_physical_devices('GPU')
# tf.config.experimental.set_memory_growth(gpus[0], True)

gpus = tf.config.list_physical_devices("GPU")
for gpu in gpus:
    tf.config.experimental.set_memory_growth(gpu, True)
logical_gpus = tf.config.list_logical_devices("GPU")
print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs")

tf.debugging.set_log_device_placement(True)
strategy = tf.distribute.MirroredStrategy(logical_gpus)

from keras.models import Model,Sequential
from keras.layers import *
from keras import callbacks
from keras.optimizers import Adam
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
import numpy as np
# import create_problem


os.environ['TF_ENABLE_GPU_GARBAGE_COLLECTION']='false'

max_voice_len = 95000
data_x = np.empty(0)
data_y = np.empty(0)
nb_training = 10000
problem_len = 10000
nb_voice = 88
voice_num = 0
file_pathx = "./data/data_x.csv"
file_pathy = "./data/data_y.csv"

data_x = np.loadtxt(file_pathx, delimiter=',') 
data_y = np.loadtxt(file_pathy, delimiter=',') 

print(data_x,data_y)

data_x = np.reshape(data_x,[nb_training,problem_len,1])
data_y = np.reshape(data_y,[nb_training,2])

x_train, x_valid, y_train, y_valid = train_test_split(data_x, data_y, test_size=0.2, shuffle= True)

Input_shape = x_train.shape[1:]


def get_model():
    with strategy.scope():
        model = Sequential()
        model.add(Conv1D(128, 32, activation='relu',input_shape=Input_shape))
        model.add(MaxPool1D(pool_size=2, padding='same'))
        # model.add(Dense(64,activation='relu'))
        # model.add(Dense(16,activation='relu'))
        model.add(LSTM(64, return_sequences=True))
        model.add(Flatten())
        model.add(Dense(2,activation='softmax'))
        model.compile(loss="categorical_crossentropy", optimizer=Adam(lr=1e-5),metrics=['accuracy'])
        return model

model = get_model()

model.summary()

history = model.fit(x_train, y_train, batch_size=128, epochs=100,verbose=1,validation_data=(x_valid, y_valid))

plt.plot(history.epoch, history.history["accuracy"], label="Train accracy")
plt.plot(history.epoch, history.history["val_accuracy"], label="Validation accracy")
plt.xlabel("epoch")
plt.legend()

plt.savefig(f"train/voice_correct_in{voice_num}.png")
model.save(f"train/voice_correct_in{voice_num}.h5")