Voice Recognition Sagemaker Script (Part1)

voice-recognition-sagemaker-script.py

import base64
import glob
import json
import logging
import subprocess
import sys
import tarfile
import traceback
import uuid
import wave
from os import unlink, environ, makedirs
from os.path import basename
from pickle import load
from random import randint
from shutil import copy2, rmtree
from urllib.request import urlretrieve

import mxnet as mx
import numpy as np
from mxnet import autograd, nd, gluon
from mxnet.gluon import Trainer
from mxnet.gluon.loss import SoftmaxCrossEntropyLoss
from mxnet.gluon.nn import Conv2D, MaxPool2D, Dropout, Flatten, Dense, Sequential
from mxnet.initializer import Xavier


def install(package):
    subprocess.call([sys.executable, "-m", "pip", "install", package])


install("opencv-python")
install("pydub")
install("matplotlib")

import cv2
import matplotlib

matplotlib.use("agg")
import matplotlib.pyplot as plt

environ["PATH"] += ":/tmp"

rmtree("ffmpeg-tmp", True)
makedirs("ffmpeg-tmp")
urlretrieve("https://johnvansickle.com/ffmpeg/builds/ffmpeg-git-64bit-static.tar.xz",
            "ffmpeg-tmp/ffmpeg-git-64bit-static.tar.xz")

tar = tarfile.open("ffmpeg-tmp/ffmpeg-git-64bit-static.tar.xz")
tar.extractall("ffmpeg-tmp")
tar.close()

for file in [src for src in glob.glob("ffmpeg-tmp/*/**") if basename(src) in ["ffmpeg", "ffprobe"]]:
    copy2(file, ".")
rmtree("ffmpeg-tmp", True)

from pydub import AudioSegment

logging.basicConfig(level=logging.INFO)

voices = ["Ivy", "Joanna", "Joey", "Justin", "Kendra", "Kimberly", "Matthew", "Salli"]


def train(hyperparameters, channel_input_dirs, num_gpus, hosts):
    batch_size = hyperparameters.get("batch_size", 64)
    epochs = hyperparameters.get("epochs", 3)

    mx.random.seed(42)

    training_dir = channel_input_dirs['training']

    with open("{}/train/data.p".format(training_dir), "rb") as pickle:
        train_nd = load(pickle)
    with open("{}/validation/data.p".format(training_dir), "rb") as pickle:
        validation_nd = load(pickle)

    train_data = gluon.data.DataLoader(train_nd, batch_size, shuffle=True)
    validation_data = gluon.data.DataLoader(validation_nd, batch_size, shuffle=True)

    net = Sequential()
    with net.name_scope():
        net.add(Conv2D(channels=32, kernel_size=(3, 3), padding=0, activation="relu"))
        net.add(Conv2D(channels=32, kernel_size=(3, 3), padding=0, activation="relu"))
        net.add(MaxPool2D(pool_size=(2, 2)))
        net.add(Dropout(.25))
        net.add(Flatten())
        net.add(Dense(8))

    ctx = mx.gpu() if num_gpus > 0 else mx.cpu()

    net.collect_params().initialize(Xavier(magnitude=2.24), ctx=ctx)

    loss = SoftmaxCrossEntropyLoss()

    if len(hosts) == 1:
        kvstore = "device" if num_gpus > 0 else "local"
    else:
        kvstore = "dist_device_sync'" if num_gpus > 0 else "dist_sync"

    trainer = Trainer(net.collect_params(), optimizer="adam", kvstore=kvstore)

    smoothing_constant = .01

    for e in range(epochs):
        moving_loss = 0
        for i, (data, label) in enumerate(train_data):
            data = data.as_in_context(ctx)
            label = label.as_in_context(ctx)
            with autograd.record():
                output = net(data)
                loss_result = loss(output, label)
            loss_result.backward()
            trainer.step(batch_size)

            curr_loss = nd.mean(loss_result).asscalar()
            moving_loss = (curr_loss if ((i == 0) and (e == 0))
                           else (1 - smoothing_constant) * moving_loss + smoothing_constant * curr_loss)

        validation_accuracy = measure_performance(net, ctx, validation_data)
        train_accuracy = measure_performance(net, ctx, train_data)
        print("Epoch %s. Loss: %s, Train_acc %s, Test_acc %s" % (e, moving_loss, train_accuracy, validation_accuracy))

    return net


def measure_performance(model, ctx, data_iter):
    acc = mx.metric.Accuracy()
    for _, (data, labels) in enumerate(data_iter):
        data = data.as_in_context(ctx)
        labels = labels.as_in_context(ctx)
        output = model(data)
        predictions = nd.argmax(output, axis=1)
        acc.update(preds=predictions, labels=labels)
    return acc.get()[1]


def save(net, model_dir):
    y = net(mx.sym.var("data"))
    y.save("{}/model.json".format(model_dir))
    net.collect_params().save("{}/model.params".format(model_dir))