654 lines
19 KiB
Python
654 lines
19 KiB
Python
"""
|
|
modified from https://github.com/speechbrain/speechbrain/blob/develop/speechbrain/lobes/models/dual_path.py
|
|
Author: Shengkui Zhao
|
|
"""
|
|
|
|
import math
|
|
import torch
|
|
import torch.nn as nn
|
|
import torch.nn.functional as F
|
|
import copy
|
|
from models.mossformer2_se.mossformer2_block import ScaledSinuEmbedding, MossformerBlock_GFSMN, MossformerBlock
|
|
|
|
|
|
EPS = 1e-8
|
|
|
|
|
|
class GlobalLayerNorm(nn.Module):
|
|
"""Calculate Global Layer Normalization.
|
|
|
|
Arguments
|
|
---------
|
|
dim : (int or list or torch.Size)
|
|
Input shape from an expected input of size.
|
|
eps : float
|
|
A value added to the denominator for numerical stability.
|
|
elementwise_affine : bool
|
|
A boolean value that when set to True,
|
|
this module has learnable per-element affine parameters
|
|
initialized to ones (for weights) and zeros (for biases).
|
|
|
|
Example
|
|
-------
|
|
>>> x = torch.randn(5, 10, 20)
|
|
>>> GLN = GlobalLayerNorm(10, 3)
|
|
>>> x_norm = GLN(x)
|
|
"""
|
|
|
|
def __init__(self, dim, shape, eps=1e-8, elementwise_affine=True):
|
|
super(GlobalLayerNorm, self).__init__()
|
|
self.dim = dim
|
|
self.eps = eps
|
|
self.elementwise_affine = elementwise_affine
|
|
|
|
if self.elementwise_affine:
|
|
if shape == 3:
|
|
self.weight = nn.Parameter(torch.ones(self.dim, 1))
|
|
self.bias = nn.Parameter(torch.zeros(self.dim, 1))
|
|
if shape == 4:
|
|
self.weight = nn.Parameter(torch.ones(self.dim, 1, 1))
|
|
self.bias = nn.Parameter(torch.zeros(self.dim, 1, 1))
|
|
else:
|
|
self.register_parameter("weight", None)
|
|
self.register_parameter("bias", None)
|
|
|
|
def forward(self, x):
|
|
"""Returns the normalized tensor.
|
|
|
|
Arguments
|
|
---------
|
|
x : torch.Tensor
|
|
Tensor of size [N, C, K, S] or [N, C, L].
|
|
"""
|
|
# x = N x C x K x S or N x C x L
|
|
# N x 1 x 1
|
|
# cln: mean,var N x 1 x K x S
|
|
# gln: mean,var N x 1 x 1
|
|
if x.dim() == 3:
|
|
mean = torch.mean(x, (1, 2), keepdim=True)
|
|
var = torch.mean((x - mean) ** 2, (1, 2), keepdim=True)
|
|
if self.elementwise_affine:
|
|
x = (
|
|
self.weight * (x - mean) / torch.sqrt(var + self.eps)
|
|
+ self.bias
|
|
)
|
|
else:
|
|
x = (x - mean) / torch.sqrt(var + self.eps)
|
|
|
|
if x.dim() == 4:
|
|
mean = torch.mean(x, (1, 2, 3), keepdim=True)
|
|
var = torch.mean((x - mean) ** 2, (1, 2, 3), keepdim=True)
|
|
if self.elementwise_affine:
|
|
x = (
|
|
self.weight * (x - mean) / torch.sqrt(var + self.eps)
|
|
+ self.bias
|
|
)
|
|
else:
|
|
x = (x - mean) / torch.sqrt(var + self.eps)
|
|
return x
|
|
|
|
|
|
class CumulativeLayerNorm(nn.LayerNorm):
|
|
"""Calculate Cumulative Layer Normalization.
|
|
|
|
Arguments
|
|
---------
|
|
dim : int
|
|
Dimension that you want to normalize.
|
|
elementwise_affine : True
|
|
Learnable per-element affine parameters.
|
|
|
|
Example
|
|
-------
|
|
>>> x = torch.randn(5, 10, 20)
|
|
>>> CLN = CumulativeLayerNorm(10)
|
|
>>> x_norm = CLN(x)
|
|
"""
|
|
|
|
def __init__(self, dim, elementwise_affine=True):
|
|
super(CumulativeLayerNorm, self).__init__(
|
|
dim, elementwise_affine=elementwise_affine, eps=1e-8
|
|
)
|
|
|
|
def forward(self, x):
|
|
"""Returns the normalized tensor.
|
|
|
|
Arguments
|
|
---------
|
|
x : torch.Tensor
|
|
Tensor size [N, C, K, S] or [N, C, L]
|
|
"""
|
|
# x: N x C x K x S or N x C x L
|
|
# N x K x S x C
|
|
if x.dim() == 4:
|
|
x = x.permute(0, 2, 3, 1).contiguous()
|
|
# N x K x S x C == only channel norm
|
|
x = super().forward(x)
|
|
# N x C x K x S
|
|
x = x.permute(0, 3, 1, 2).contiguous()
|
|
if x.dim() == 3:
|
|
x = torch.transpose(x, 1, 2)
|
|
# N x L x C == only channel norm
|
|
x = super().forward(x)
|
|
# N x C x L
|
|
x = torch.transpose(x, 1, 2)
|
|
return x
|
|
|
|
|
|
def select_norm(norm, dim, shape):
|
|
"""Just a wrapper to select the normalization type.
|
|
"""
|
|
|
|
if norm == "gln":
|
|
return GlobalLayerNorm(dim, shape, elementwise_affine=True)
|
|
if norm == "cln":
|
|
return CumulativeLayerNorm(dim, elementwise_affine=True)
|
|
if norm == "ln":
|
|
return nn.GroupNorm(1, dim, eps=1e-8)
|
|
else:
|
|
return nn.BatchNorm1d(dim)
|
|
|
|
|
|
class Encoder(nn.Module):
|
|
"""Convolutional Encoder Layer.
|
|
|
|
Arguments
|
|
---------
|
|
kernel_size : int
|
|
Length of filters.
|
|
in_channels : int
|
|
Number of input channels.
|
|
out_channels : int
|
|
Number of output channels.
|
|
|
|
Example
|
|
-------
|
|
>>> x = torch.randn(2, 1000)
|
|
>>> encoder = Encoder(kernel_size=4, out_channels=64)
|
|
>>> h = encoder(x)
|
|
>>> h.shape
|
|
torch.Size([2, 64, 499])
|
|
"""
|
|
|
|
def __init__(self, kernel_size=2, out_channels=64, in_channels=1):
|
|
super(Encoder, self).__init__()
|
|
self.conv1d = nn.Conv1d(
|
|
in_channels=in_channels,
|
|
out_channels=out_channels,
|
|
kernel_size=kernel_size,
|
|
stride=kernel_size // 2,
|
|
groups=1,
|
|
bias=False,
|
|
)
|
|
self.in_channels = in_channels
|
|
|
|
def forward(self, x):
|
|
"""Return the encoded output.
|
|
|
|
Arguments
|
|
---------
|
|
x : torch.Tensor
|
|
Input tensor with dimensionality [B, L].
|
|
Return
|
|
------
|
|
x : torch.Tensor
|
|
Encoded tensor with dimensionality [B, N, T_out].
|
|
|
|
where B = Batchsize
|
|
L = Number of timepoints
|
|
N = Number of filters
|
|
T_out = Number of timepoints at the output of the encoder
|
|
"""
|
|
# B x L -> B x 1 x L
|
|
if self.in_channels == 1:
|
|
x = torch.unsqueeze(x, dim=1)
|
|
# B x 1 x L -> B x N x T_out
|
|
x = self.conv1d(x)
|
|
x = F.relu(x)
|
|
|
|
return x
|
|
|
|
|
|
class Decoder(nn.ConvTranspose1d):
|
|
"""A decoder layer that consists of ConvTranspose1d.
|
|
|
|
Arguments
|
|
---------
|
|
kernel_size : int
|
|
Length of filters.
|
|
in_channels : int
|
|
Number of input channels.
|
|
out_channels : int
|
|
Number of output channels.
|
|
|
|
|
|
Example
|
|
---------
|
|
>>> x = torch.randn(2, 100, 1000)
|
|
>>> decoder = Decoder(kernel_size=4, in_channels=100, out_channels=1)
|
|
>>> h = decoder(x)
|
|
>>> h.shape
|
|
torch.Size([2, 1003])
|
|
"""
|
|
|
|
def __init__(self, *args, **kwargs):
|
|
super(Decoder, self).__init__(*args, **kwargs)
|
|
|
|
def forward(self, x):
|
|
"""Return the decoded output.
|
|
|
|
Arguments
|
|
---------
|
|
x : torch.Tensor
|
|
Input tensor with dimensionality [B, N, L].
|
|
where, B = Batchsize,
|
|
N = number of filters
|
|
L = time points
|
|
"""
|
|
|
|
if x.dim() not in [2, 3]:
|
|
raise RuntimeError(
|
|
"{} accept 3/4D tensor as input".format(self.__name__)
|
|
)
|
|
x = super().forward(x if x.dim() == 3 else torch.unsqueeze(x, 1))
|
|
|
|
if torch.squeeze(x).dim() == 1:
|
|
x = torch.squeeze(x, dim=1)
|
|
else:
|
|
x = torch.squeeze(x)
|
|
return x
|
|
|
|
|
|
class IdentityBlock:
|
|
"""This block is used when we want to have identity transformation within the Dual_path block.
|
|
|
|
Example
|
|
-------
|
|
>>> x = torch.randn(10, 100)
|
|
>>> IB = IdentityBlock()
|
|
>>> xhat = IB(x)
|
|
"""
|
|
|
|
def _init__(self, **kwargs):
|
|
pass
|
|
|
|
def __call__(self, x):
|
|
return x
|
|
|
|
|
|
class MossFormerM(nn.Module):
|
|
"""This class implements the transformer encoder based on MossFormer2 layers.
|
|
|
|
Arguments
|
|
---------
|
|
num_blocks : int
|
|
Number of mossformer2 blocks to include.
|
|
d_model : int
|
|
The dimension of the input embedding.
|
|
attn_dropout : float
|
|
Dropout for the self-attention (Optional).
|
|
group_size: int
|
|
the chunk size for segmenting sequence
|
|
query_key_dim: int
|
|
the attention vector dimension
|
|
expansion_factor: int
|
|
the expansion factor for the linear projection in conv module
|
|
causal: bool
|
|
true for causal / false for non causal
|
|
|
|
Example
|
|
-------
|
|
>>> import torch
|
|
>>> x = torch.rand((8, 60, 512))
|
|
>>> net = MossFormerM(num_blocks=8, d_model=512)
|
|
>>> output, _ = net(x)
|
|
>>> output.shape
|
|
torch.Size([8, 60, 512])
|
|
"""
|
|
def __init__(
|
|
self,
|
|
num_blocks,
|
|
d_model=None,
|
|
causal=False,
|
|
group_size = 256,
|
|
query_key_dim = 128,
|
|
expansion_factor = 4.,
|
|
attn_dropout = 0.1
|
|
):
|
|
super().__init__()
|
|
|
|
self.mossformerM = MossformerBlock_GFSMN(
|
|
dim=d_model,
|
|
depth=num_blocks,
|
|
group_size=group_size,
|
|
query_key_dim=query_key_dim,
|
|
expansion_factor=expansion_factor,
|
|
causal=causal,
|
|
attn_dropout=attn_dropout
|
|
)
|
|
self.norm = nn.LayerNorm(d_model, eps=1e-6)
|
|
def forward(
|
|
self,
|
|
src,
|
|
):
|
|
"""
|
|
Arguments
|
|
----------
|
|
src : torch.Tensor
|
|
Tensor shape [B, L, N],
|
|
where, B = Batchsize,
|
|
L = time points
|
|
N = number of filters
|
|
The sequence to the encoder layer (required).
|
|
src_mask : tensor
|
|
The mask for the src sequence (optional).
|
|
src_key_padding_mask : tensor
|
|
The mask for the src keys per batch (optional).
|
|
"""
|
|
output = self.mossformerM(src)
|
|
output = self.norm(output)
|
|
|
|
return output
|
|
|
|
class MossFormerM2(nn.Module):
|
|
"""This class implements the transformer encoder.
|
|
|
|
Arguments
|
|
---------
|
|
num_blocks : int
|
|
Number of mossformer blocks to include.
|
|
d_model : int
|
|
The dimension of the input embedding.
|
|
attn_dropout : float
|
|
Dropout for the self-attention (Optional).
|
|
group_size: int
|
|
the chunk size
|
|
query_key_dim: int
|
|
the attention vector dimension
|
|
expansion_factor: int
|
|
the expansion factor for the linear projection in conv module
|
|
causal: bool
|
|
true for causal / false for non causal
|
|
|
|
Example
|
|
-------
|
|
>>> import torch
|
|
>>> x = torch.rand((8, 60, 512))
|
|
>>> net = MossFormerM2(num_blocks=8, d_model=512)
|
|
>>> output, _ = net(x)
|
|
>>> output.shape
|
|
torch.Size([8, 60, 512])
|
|
"""
|
|
def __init__(
|
|
self,
|
|
num_blocks,
|
|
d_model=None,
|
|
causal=False,
|
|
group_size = 256,
|
|
query_key_dim = 128,
|
|
expansion_factor = 4.,
|
|
attn_dropout = 0.1
|
|
):
|
|
super().__init__()
|
|
|
|
self.mossformerM = MossformerBlock(
|
|
dim=d_model,
|
|
depth=num_blocks,
|
|
group_size=group_size,
|
|
query_key_dim=query_key_dim,
|
|
expansion_factor=expansion_factor,
|
|
causal=causal,
|
|
attn_dropout=attn_dropout
|
|
)
|
|
self.norm = nn.LayerNorm(d_model, eps=1e-6)
|
|
|
|
def forward(
|
|
self,
|
|
src,
|
|
):
|
|
"""
|
|
Arguments
|
|
----------
|
|
src : torch.Tensor
|
|
Tensor shape [B, L, N],
|
|
where, B = Batchsize,
|
|
L = time points
|
|
N = number of filters
|
|
The sequence to the encoder layer (required).
|
|
src_mask : tensor
|
|
The mask for the src sequence (optional).
|
|
src_key_padding_mask : tensor
|
|
The mask for the src keys per batch (optional).
|
|
"""
|
|
output = self.mossformerM(src)
|
|
output = self.norm(output)
|
|
|
|
return output
|
|
|
|
class Computation_Block(nn.Module):
|
|
"""Computation block for dual-path processing.
|
|
|
|
Arguments
|
|
---------
|
|
out_channels : int
|
|
Dimensionality of model output.
|
|
norm : str
|
|
Normalization type.
|
|
skip_around_intra : bool
|
|
Skip connection around the intra layer.
|
|
|
|
Example
|
|
---------
|
|
>>> comp_block = Computation_Block(64)
|
|
>>> x = torch.randn(10, 64, 100)
|
|
>>> x = comp_block(x)
|
|
>>> x.shape
|
|
torch.Size([10, 64, 100])
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
num_blocks,
|
|
out_channels,
|
|
norm="ln",
|
|
skip_around_intra=True,
|
|
):
|
|
super(Computation_Block, self).__init__()
|
|
|
|
##Default MossFormer2 model
|
|
self.intra_mdl = MossFormerM(num_blocks=num_blocks, d_model=out_channels)
|
|
##The previous MossFormer model
|
|
#self.intra_mdl = MossFormerM2(num_blocks=num_blocks, d_model=out_channels)
|
|
self.skip_around_intra = skip_around_intra
|
|
|
|
# Norm
|
|
self.norm = norm
|
|
if norm is not None:
|
|
self.intra_norm = select_norm(norm, out_channels, 3)
|
|
|
|
def forward(self, x):
|
|
"""Returns the output tensor.
|
|
|
|
Arguments
|
|
---------
|
|
x : torch.Tensor
|
|
Input tensor of dimension [B, N, S].
|
|
|
|
|
|
Return
|
|
---------
|
|
out: torch.Tensor
|
|
Output tensor of dimension [B, N, S].
|
|
where, B = Batchsize,
|
|
N = number of filters
|
|
S = sequence time index
|
|
"""
|
|
B, N, S = x.shape
|
|
# [B, S, N]
|
|
intra = x.permute(0, 2, 1).contiguous()
|
|
|
|
intra = self.intra_mdl(intra)
|
|
|
|
# [B, N, S]
|
|
intra = intra.permute(0, 2, 1).contiguous()
|
|
if self.norm is not None:
|
|
intra = self.intra_norm(intra)
|
|
|
|
# [B, N, S]
|
|
if self.skip_around_intra:
|
|
intra = intra + x
|
|
|
|
out = intra
|
|
return out
|
|
|
|
class MossFormer_MaskNet(nn.Module):
|
|
"""
|
|
The MossFormer MaskNet for mask prediction.
|
|
|
|
This class is designed for predicting masks used in source separation tasks.
|
|
It processes input tensors through various layers including convolutional layers,
|
|
normalization, and a computation block to produce the final output.
|
|
|
|
Arguments
|
|
---------
|
|
in_channels : int
|
|
Number of channels at the output of the encoder.
|
|
out_channels : int
|
|
Number of channels that would be inputted to the MossFormer2 blocks.
|
|
out_channels_final : int
|
|
Number of channels that are finally outputted.
|
|
num_blocks : int
|
|
Number of layers in the Dual Computation Block.
|
|
norm : str
|
|
Normalization type ('ln' for LayerNorm, 'bn' for BatchNorm, etc.).
|
|
num_spks : int
|
|
Number of sources (speakers).
|
|
skip_around_intra : bool
|
|
If True, applies skip connections around intra-block connections.
|
|
use_global_pos_enc : bool
|
|
If True, uses global positional encodings.
|
|
max_length : int
|
|
Maximum sequence length for input tensors.
|
|
|
|
Example
|
|
---------
|
|
>>> mossformer_masknet = MossFormer_MaskNet(64, 64, out_channels_final=8, num_spks=2)
|
|
>>> x = torch.randn(10, 64, 2000) # Example input
|
|
>>> x = mossformer_masknet(x) # Forward pass
|
|
>>> x.shape # Expected output shape
|
|
torch.Size([10, 2, 64, 2000])
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
in_channels,
|
|
out_channels,
|
|
out_channels_final,
|
|
num_blocks=24,
|
|
norm="ln",
|
|
num_spks=2,
|
|
skip_around_intra=True,
|
|
use_global_pos_enc=True,
|
|
max_length=20000,
|
|
):
|
|
super(MossFormer_MaskNet, self).__init__()
|
|
|
|
# Initialize instance variables
|
|
self.num_spks = num_spks # Number of sources
|
|
self.num_blocks = num_blocks # Number of computation blocks
|
|
self.norm = select_norm(norm, in_channels, 3) # Select normalization type
|
|
self.conv1d_encoder = nn.Conv1d(in_channels, out_channels, 1, bias=False) # Encoder convolutional layer
|
|
self.use_global_pos_enc = use_global_pos_enc # Flag for global positional encoding
|
|
|
|
if self.use_global_pos_enc:
|
|
self.pos_enc = ScaledSinuEmbedding(out_channels) # Initialize positional embedding
|
|
|
|
# Define the computation block
|
|
self.mdl = Computation_Block(
|
|
num_blocks,
|
|
out_channels,
|
|
norm,
|
|
skip_around_intra=skip_around_intra,
|
|
)
|
|
|
|
# Output layers
|
|
self.conv1d_out = nn.Conv1d(out_channels, out_channels * num_spks, kernel_size=1) # For multiple speakers
|
|
self.conv1_decoder = nn.Conv1d(out_channels, out_channels_final, 1, bias=False) # Decoder layer
|
|
self.prelu = nn.PReLU() # Activation function
|
|
self.activation = nn.ReLU() # Final activation function
|
|
|
|
# Gated output layers
|
|
self.output = nn.Sequential(
|
|
nn.Conv1d(out_channels, out_channels, 1),
|
|
nn.Tanh() # Non-linear activation
|
|
)
|
|
self.output_gate = nn.Sequential(
|
|
nn.Conv1d(out_channels, out_channels, 1),
|
|
nn.Sigmoid() # Gating mechanism
|
|
)
|
|
|
|
def forward(self, x):
|
|
"""Returns the output tensor.
|
|
|
|
Arguments
|
|
---------
|
|
x : torch.Tensor
|
|
Input tensor of dimension [B, N, S], where B is the batch size,
|
|
N is the number of channels, and S is the sequence length.
|
|
|
|
Returns
|
|
-------
|
|
out : torch.Tensor
|
|
Output tensor of dimension [B, spks, N, S], where spks is the number of sources
|
|
(speakers) and is ordered such that the first index corresponds to the target speech.
|
|
"""
|
|
|
|
# Normalize the input
|
|
# [B, N, L]
|
|
x = self.norm(x)
|
|
|
|
# Apply encoder convolution
|
|
# [B, N, L]
|
|
x = self.conv1d_encoder(x)
|
|
|
|
if self.use_global_pos_enc:
|
|
base = x # Store the base for adding positional embedding
|
|
x = x.transpose(1, -1) # Change shape to [B, L, N] for positional encoding
|
|
emb = self.pos_enc(x) # Get positional embeddings
|
|
emb = emb.transpose(0, -1) # Change back to [B, N, L]
|
|
x = base + emb # Add positional embeddings to the base
|
|
|
|
# Process through the computation block
|
|
# [B, N, S]
|
|
x = self.mdl(x)
|
|
x = self.prelu(x) # Apply activation
|
|
|
|
# Expand to multiple speakers
|
|
# [B, N*spks, S]
|
|
x = self.conv1d_out(x)
|
|
B, _, S = x.shape # Unpack the batch size and sequence length
|
|
|
|
# Reshape to [B*spks, N, S]
|
|
# This prepares the output for gating
|
|
# [B*spks, N, S]
|
|
x = x.view(B * self.num_spks, -1, S)
|
|
|
|
# Apply gated output layers
|
|
# [B*spks, N, S]
|
|
x = self.output(x) * self.output_gate(x) # Element-wise multiplication for gating
|
|
|
|
# Decode to final output
|
|
# [B*spks, N, S]
|
|
x = self.conv1_decoder(x)
|
|
|
|
# Reshape to [B, spks, N, S] for output
|
|
# [B, spks, N, S]
|
|
_, N, L = x.shape
|
|
x = x.view(B, self.num_spks, N, L) # Final reshaping for output
|
|
x = self.activation(x) # Apply final activation
|
|
|
|
# Transpose to [spks, B, N, S] for output
|
|
# return the 1st spk signal as the target speech
|
|
x = x.transpose(0, 1)
|
|
return x[0].transpose(1, 2) # Return only the first speaker's signal
|