ASRT_SpeechRecognition/train_speech_model.py

54 lines
1.9 KiB
Python
Raw Permalink Normal View History

2022-09-18 20:56:14 +08:00
# !/usr/bin/env python3
# -*- coding: utf-8 -*-
#
# Copyright 2016-2099 Ailemon.net
#
# This file is part of ASRT Speech Recognition Tool.
#
# ASRT is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
# ASRT is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with ASRT. If not, see <https://www.gnu.org/licenses/>.
# ============================================================================
"""
@author: nl8590687
用于训练语音识别系统语音模型的程序
"""
2022-09-18 20:56:14 +08:00
import os
from tensorflow.keras.optimizers import Adam
from speech_model import ModelSpeech
2022-05-24 00:02:28 +08:00
from model_zoo.speech_model.keras_backend import SpeechModel251BN
from data_loader import DataLoader
from speech_features import SpecAugment
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
2021-11-21 20:20:26 +08:00
AUDIO_LENGTH = 1600
AUDIO_FEATURE_LENGTH = 200
CHANNELS = 1
# 默认输出的拼音的表示大小是1428即1427个拼音+1个空白块
2021-11-21 20:20:26 +08:00
OUTPUT_SIZE = 1428
sm251bn = SpeechModel251BN(
2021-11-21 20:20:26 +08:00
input_shape=(AUDIO_LENGTH, AUDIO_FEATURE_LENGTH, CHANNELS),
output_size=OUTPUT_SIZE
2022-09-18 20:56:14 +08:00
)
feat = SpecAugment()
train_data = DataLoader('train')
2022-09-18 20:56:14 +08:00
opt = Adam(learning_rate=0.0001, beta_1=0.9, beta_2=0.999, decay=0.0, epsilon=10e-8)
ms = ModelSpeech(sm251bn, feat, max_label_length=64)
2022-09-18 20:56:14 +08:00
# ms.load_model('save_models/' + sm251bn.get_model_name() + '.model.h5')
2021-11-21 20:20:26 +08:00
ms.train_model(optimizer=opt, data_loader=train_data,
2022-09-18 20:56:14 +08:00
epochs=50, save_step=1, batch_size=16, last_epoch=0)
ms.save_model('save_models/' + sm251bn.get_model_name())