cosyvoice2/utils/common.py

# Copyright (c) 2020 Mobvoi Inc (Binbin Zhang)
#               2024 Alibaba Inc (authors: Xiang Lyu)
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Modified from ESPnet(https://github.com/espnet/espnet)
"""Unility functions for Transformer."""

import random
from typing import List

import numpy as np
import torch

IGNORE_ID = -1


def pad_list(xs: List[torch.Tensor], pad_value: int):
    """Perform padding for the list of tensors.

    Args:
        xs (List): List of Tensors [(T_1, `*`), (T_2, `*`), ..., (T_B, `*`)].
        pad_value (float): Value for padding.

    Returns:
        Tensor: Padded tensor (B, Tmax, `*`).

    Examples:
        >>> x = [torch.ones(4), torch.ones(2), torch.ones(1)]
        >>> x
        [tensor([1., 1., 1., 1.]), tensor([1., 1.]), tensor([1.])]
        >>> pad_list(x, 0)
        tensor([[1., 1., 1., 1.],
                [1., 1., 0., 0.],
                [1., 0., 0., 0.]])

    """
    max_len = max([len(item) for item in xs])
    batchs = len(xs)
    ndim = xs[0].ndim
    if ndim == 1:
        pad_res = torch.zeros(batchs,
                              max_len,
                              dtype=xs[0].dtype,
                              device=xs[0].device)
    elif ndim == 2:
        pad_res = torch.zeros(batchs,
                              max_len,
                              xs[0].shape[1],
                              dtype=xs[0].dtype,
                              device=xs[0].device)
    elif ndim == 3:
        pad_res = torch.zeros(batchs,
                              max_len,
                              xs[0].shape[1],
                              xs[0].shape[2],
                              dtype=xs[0].dtype,
                              device=xs[0].device)
    else:
        raise ValueError(f"Unsupported ndim: {ndim}")
    pad_res.fill_(pad_value)
    for i in range(batchs):
        pad_res[i, :len(xs[i])] = xs[i]
    return pad_res


def get_padding(kernel_size, dilation=1):
    return int((kernel_size * dilation - dilation) / 2)


def init_weights(m, mean=0.0, std=0.01):
    classname = m.__class__.__name__
    if classname.find("Conv") != -1:
        m.weight.data.normal_(mean, std)


def fade_in_out(fade_in_mel, fade_out_mel, window):
    device = fade_in_mel.device
    fade_in_mel, fade_out_mel = fade_in_mel.cpu(), fade_out_mel.cpu()
    mel_overlap_len = int(window.shape[0] / 2)
    if fade_in_mel.device == torch.device('cpu'):
        fade_in_mel = fade_in_mel.clone()
    fade_in_mel[..., :mel_overlap_len] = fade_in_mel[..., :mel_overlap_len] * window[:mel_overlap_len] + \
        fade_out_mel[..., -mel_overlap_len:] * window[mel_overlap_len:]
    return fade_in_mel.to(device)


def set_all_random_seed(seed):
    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)