fish_speech/models/flow_dit/modules/utils.py

import torch
import torch.nn as nn
import torch.nn.functional as F

from matplotlib import pyplot as plt

import numpy as np

from typing import Any, Dict, Tuple, Union


def make_pad_mask(lengths: torch.Tensor, max_len: int = 0) -> torch.Tensor:
    assert lengths.ndim == 1, lengths.ndim
    max_len = max(max_len, lengths.max())
    n = lengths.size(0)
    seq_range = torch.arange(0, max_len, device=lengths.device)
    expaned_lengths = seq_range.unsqueeze(0).expand(n, max_len)
    return expaned_lengths >= lengths.unsqueeze(-1)


def make_attn_mask(lengths: torch.Tensor, num_heads: int, causal: False) -> torch.Tensor:

    key_padding_mask = make_pad_mask(lengths)

    bsz = key_padding_mask.size(0)
    seq_len = key_padding_mask.size(1)

    key_padding_mask = key_padding_mask.view(bsz, 1, 1, seq_len).expand(-1, num_heads, -1, -1)

    if causal:
        assert seq_len == lengths.max(), "Causal mask requires all lengths to be equal to max_len"
        causal_mask = torch.triu(torch.ones(seq_len, seq_len, dtype=torch.bool, device=key_padding_mask.device), diagonal=1)
        causal_mask = causal_mask.view(1, 1, seq_len, seq_len).expand(bsz, num_heads, -1, -1)
        causal_mask = causal_mask.logical_or(key_padding_mask)
        return causal_mask.float().masked_fill(causal_mask, float("-inf"))
    else:
        key_padding_mask_float = key_padding_mask.float()
        key_padding_mask_float = key_padding_mask_float.masked_fill(key_padding_mask, float("-inf"))
        return key_padding_mask_float

def save_figure_to_numpy(fig: plt.Figure) -> np.ndarray:
    """
    Save a matplotlib figure to a numpy array.

    Args:
        fig (Figure): Matplotlib figure object.

    Returns:
        ndarray: Numpy array representing the figure.
    """
    data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep="")
    data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
    return data

def plot_spectrogram_to_numpy(spec_target: np.ndarray, spec_output: np.ndarray) -> np.ndarray:
    """
    Plot a spectrogram and convert it to a numpy array.

    Args:
        spectrogram (ndarray): Spectrogram data.

    Returns:
        ndarray: Numpy array representing the plotted spectrogram.
    """

    fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(12, 12))
    ax1.set_title("Target")
    im = ax1.imshow(spec_target.astype(np.float32), aspect="auto", origin="lower", interpolation="none")
    plt.colorbar(im, ax=ax1)
    plt.xlabel("Frames")
    plt.ylabel("Channels")

    ax2.set_title("Output")
    im = ax2.imshow(spec_output.astype(np.float32), aspect="auto", origin="lower", interpolation="none")
    plt.colorbar(im, ax=ax2)
    plt.xlabel("Frames")
    plt.ylabel("Channels")

    plt.tight_layout()

    fig.canvas.draw()
    data = save_figure_to_numpy(fig)
    plt.close()
    return data

def instantiate_class(args: Union[Any, Tuple[Any, ...]], init: Dict[str, Any]) -> Any:
    """Instantiates a class with the given args and init.

    Args:
        args: Positional arguments required for instantiation.
        init: Dict of the form {"class_path":...,"init_args":...}.

    Returns:
        The instantiated class object.
    """
    kwargs = init.get("init_args", {})
    if not isinstance(args, tuple):
        args = (args,)
    class_module, class_name = init["class_path"].rsplit(".", 1)
    module = __import__(class_module, fromlist=[class_name])
    args_class = getattr(module, class_name)
    return args_class(*args, **kwargs)


if __name__ == "__main__":
    lengths = torch.tensor([3, 5])

    b = torch.ones(3, 5, 5)
    m = make_attn_mask(lengths, 1, True)
    print(m)
    print(b + m)