inference.py

# SPDX-FileCopyrightText: Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: MIT
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.

import json
import torch

from hifigan_models import Generator
from hifigan_env import AttrDict
from hifigan_denoiser import Denoiser


def lines_to_list(filename):
    """
    Takes a text file of filenames and makes a list of filenames
    """
    with open(filename, encoding="utf-8") as f:
        files = f.readlines()

    files = [f.rstrip() for f in files]
    return files


def load_vocoder(vocoder_path, config_path, to_cuda=False):
    with open(config_path) as f:
        data_vocoder = f.read()

    config_vocoder = json.loads(data_vocoder)
    h = AttrDict(config_vocoder)

    if "blur" in vocoder_path:
        config_vocoder["gaussian_blur"]["p_blurring"] = 0.5
    else:
        if "gaussian_blur" in config_vocoder:
            config_vocoder["gaussian_blur"]["p_blurring"] = 0.0
        else:
            config_vocoder["gaussian_blur"] = {"p_blurring": 0.0}
            h["gaussian_blur"] = {"p_blurring": 0.0}

    state_dict_g = torch.load(vocoder_path, map_location="cpu")["generator"]

    # load hifigan
    vocoder = Generator(h)
    vocoder.load_state_dict(state_dict_g)
    denoiser = Denoiser(vocoder)

    if to_cuda:
        vocoder.cuda()
        denoiser.cuda()

    vocoder.eval()
    denoiser.eval()

    return vocoder, denoiser