inference.py

import numpy as np
import torch
from tqdm import tqdm 
import os

WaveGlow = utils.get_WaveGlow()
WaveGlow = WaveGlow.cuda()

from model.fastspeech import FastSpeech
model = FastSpeech()
model.load_state_dict(torch.load('../model_new/checkpoint_225000.pth.tar', map_location='cuda:0')['model'])
model = model.eval()


def synthesis(model, text, alpha=1.0):
    text = np.array(phn)
    text = np.stack([text])
    src_pos = np.array([i+1 for i in range(text.shape[1])])
    src_pos = np.stack([src_pos])
    sequence = torch.from_numpy(text).long().to(train_config.device)
    src_pos = torch.from_numpy(src_pos).long().to(train_config.device)
    
    with torch.no_grad():
        mel = model.forward(sequence, src_pos, alpha=alpha)
    return mel[0].cpu().transpose(0, 1), mel.contiguous().transpose(1, 2)


def get_data():
    tests = [ 
        "I am very happy to see you again!",
        "Durian model is a very good speech synthesis!",
        "When I was twenty, I fell in love with a girl.",
        "I remove attention module in decoder and use average pooling to implement predicting r frames at once",
        "You can not improve your past, but you can improve your future. Once time is wasted, life is wasted.",
        "Death comes to all, but great achievements raise a monument which shall endure until the sun grows old."
    ]
    data_list = list(text.text_to_sequence(test, train_config.text_cleaners) for test in tests)

    return data_list

data_list = get_data()
for speed in [0.8, 1., 1.3]:
    for i, phn in tqdm(enumerate(data_list)):
        mel, mel_cuda = synthesis(model, phn, speed)
        
        os.makedirs("results", exist_ok=True)
        
        audio.tools.inv_mel_spec(
            mel, f"results/s={speed}_{i}.wav"
        )
        
        waveglow.inference.inference(
            mel_cuda, WaveGlow,
            f"results/s={speed}_{i}_waveglow.wav"
        )