461 lines
18 KiB
Python
461 lines
18 KiB
Python
import random
|
|
|
|
import torch
|
|
import torch.nn as nn
|
|
import torch.nn.functional as F
|
|
from torch import autocast
|
|
from x_transformers import Encoder
|
|
|
|
from models.diffusion.nn import timestep_embedding, normalization, zero_module, conv_nd, linear
|
|
from models.diffusion.unet_diffusion import AttentionBlock, TimestepEmbedSequential, \
|
|
Downsample, Upsample, TimestepBlock
|
|
from models.audio.tts.mini_encoder import AudioMiniEncoder
|
|
from models.audio.tts.unet_diffusion_tts7 import CheckpointedXTransformerEncoder
|
|
from scripts.audio.gen.use_diffuse_tts import ceil_multiple
|
|
from trainer.networks import register_model
|
|
from utils.util import checkpoint
|
|
|
|
def is_sequence(t):
|
|
return t.dtype == torch.long
|
|
|
|
|
|
class ResBlock(TimestepBlock):
|
|
def __init__(
|
|
self,
|
|
channels,
|
|
emb_channels,
|
|
dropout,
|
|
out_channels=None,
|
|
dims=2,
|
|
kernel_size=3,
|
|
efficient_config=True,
|
|
use_scale_shift_norm=False,
|
|
):
|
|
super().__init__()
|
|
self.channels = channels
|
|
self.emb_channels = emb_channels
|
|
self.dropout = dropout
|
|
self.out_channels = out_channels or channels
|
|
self.use_scale_shift_norm = use_scale_shift_norm
|
|
padding = {1: 0, 3: 1, 5: 2}[kernel_size]
|
|
eff_kernel = 1 if efficient_config else 3
|
|
eff_padding = 0 if efficient_config else 1
|
|
|
|
self.in_layers = nn.Sequential(
|
|
normalization(channels),
|
|
nn.SiLU(),
|
|
conv_nd(dims, channels, self.out_channels, eff_kernel, padding=eff_padding),
|
|
)
|
|
|
|
self.emb_layers = nn.Sequential(
|
|
nn.SiLU(),
|
|
linear(
|
|
emb_channels,
|
|
2 * self.out_channels if use_scale_shift_norm else self.out_channels,
|
|
),
|
|
)
|
|
self.out_layers = nn.Sequential(
|
|
normalization(self.out_channels),
|
|
nn.SiLU(),
|
|
nn.Dropout(p=dropout),
|
|
zero_module(
|
|
conv_nd(dims, self.out_channels, self.out_channels, kernel_size, padding=padding)
|
|
),
|
|
)
|
|
|
|
if self.out_channels == channels:
|
|
self.skip_connection = nn.Identity()
|
|
else:
|
|
self.skip_connection = conv_nd(dims, channels, self.out_channels, eff_kernel, padding=eff_padding)
|
|
|
|
def forward(self, x, emb):
|
|
"""
|
|
Apply the block to a Tensor, conditioned on a timestep embedding.
|
|
|
|
:param x: an [N x C x ...] Tensor of features.
|
|
:param emb: an [N x emb_channels] Tensor of timestep embeddings.
|
|
:return: an [N x C x ...] Tensor of outputs.
|
|
"""
|
|
return checkpoint(
|
|
self._forward, x, emb
|
|
)
|
|
|
|
def _forward(self, x, emb):
|
|
h = self.in_layers(x)
|
|
emb_out = self.emb_layers(emb).type(h.dtype)
|
|
while len(emb_out.shape) < len(h.shape):
|
|
emb_out = emb_out[..., None]
|
|
if self.use_scale_shift_norm:
|
|
out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
|
|
scale, shift = torch.chunk(emb_out, 2, dim=1)
|
|
h = out_norm(h) * (1 + scale) + shift
|
|
h = out_rest(h)
|
|
else:
|
|
h = h + emb_out
|
|
h = self.out_layers(h)
|
|
return self.skip_connection(x) + h
|
|
|
|
|
|
class ResBlockSimple(nn.Module):
|
|
def __init__(
|
|
self,
|
|
channels,
|
|
dropout,
|
|
out_channels=None,
|
|
dims=1,
|
|
kernel_size=3,
|
|
efficient_config=True,
|
|
):
|
|
super().__init__()
|
|
self.channels = channels
|
|
self.dropout = dropout
|
|
self.out_channels = out_channels or channels
|
|
padding = {1: 0, 3: 1, 5: 2}[kernel_size]
|
|
eff_kernel = 1 if efficient_config else 3
|
|
eff_padding = 0 if efficient_config else 1
|
|
|
|
self.in_layers = nn.Sequential(
|
|
normalization(channels),
|
|
nn.SiLU(),
|
|
conv_nd(dims, channels, self.out_channels, eff_kernel, padding=eff_padding),
|
|
)
|
|
self.out_layers = nn.Sequential(
|
|
normalization(self.out_channels),
|
|
nn.SiLU(),
|
|
nn.Dropout(p=dropout),
|
|
zero_module(
|
|
conv_nd(dims, self.out_channels, self.out_channels, kernel_size, padding=padding)
|
|
),
|
|
)
|
|
|
|
if self.out_channels == channels:
|
|
self.skip_connection = nn.Identity()
|
|
else:
|
|
self.skip_connection = conv_nd(dims, channels, self.out_channels, eff_kernel, padding=eff_padding)
|
|
|
|
def forward(self, x):
|
|
return checkpoint(
|
|
self._forward, x
|
|
)
|
|
|
|
def _forward(self, x):
|
|
h = self.in_layers(x)
|
|
h = self.out_layers(h)
|
|
return self.skip_connection(x) + h
|
|
|
|
|
|
class AudioVAE(nn.Module):
|
|
def __init__(self, channels, dropout):
|
|
super().__init__()
|
|
# 1, 4, 16, 64, 256
|
|
level_resblocks = [1, 1, 2, 2, 2]
|
|
level_ch_mult = [1, 2, 4, 6, 8]
|
|
levels = []
|
|
for i, (resblks, chdiv) in enumerate(zip(level_resblocks, level_ch_mult)):
|
|
blocks = [ResBlockSimple(channels*chdiv, dropout=dropout, kernel_size=5) for _ in range(resblks)]
|
|
if i != len(level_ch_mult)-1:
|
|
blocks.append(nn.Conv1d(channels*chdiv, channels*level_ch_mult[i+1], kernel_size=5, padding=2, stride=4))
|
|
levels.append(nn.Sequential(*blocks))
|
|
self.down_levels = nn.ModuleList(levels)
|
|
|
|
levels = []
|
|
lastdiv = None
|
|
for resblks, chdiv in reversed(list(zip(level_resblocks, level_ch_mult))):
|
|
if lastdiv is not None:
|
|
blocks = [nn.Conv1d(channels*lastdiv, channels*chdiv, kernel_size=5, padding=2)]
|
|
else:
|
|
blocks = []
|
|
blocks.extend([ResBlockSimple(channels*chdiv, dropout=dropout, kernel_size=5) for _ in range(resblks)])
|
|
levels.append(nn.Sequential(*blocks))
|
|
lastdiv = chdiv
|
|
self.up_levels = nn.ModuleList(levels)
|
|
|
|
def forward(self, x):
|
|
h = x
|
|
for level in self.down_levels:
|
|
h = level(h)
|
|
|
|
for k, level in enumerate(self.up_levels):
|
|
h = level(h)
|
|
if k != len(self.up_levels)-1:
|
|
h = F.interpolate(h, scale_factor=4, mode='linear')
|
|
return h
|
|
|
|
|
|
class Diffusion(nn.Module):
|
|
"""
|
|
The full UNet model with attention and timestep embedding.
|
|
|
|
Customized to be conditioned on an aligned prior derived from a autoregressive
|
|
GPT-style model.
|
|
|
|
:param in_channels: channels in the input Tensor.
|
|
:param in_latent_channels: channels from the input latent.
|
|
:param model_channels: base channel count for the model.
|
|
:param out_channels: channels in the output Tensor.
|
|
:param num_res_blocks: number of residual blocks per downsample.
|
|
:param dropout: the dropout probability.
|
|
:param channel_mult: channel multiplier for each level of the UNet.
|
|
:param conv_resample: if True, use learned convolutions for upsampling and
|
|
downsampling.
|
|
:param use_scale_shift_norm: use a FiLM-like conditioning mechanism.
|
|
:param resblock_updown: use residual blocks for up/downsampling.
|
|
:param use_new_attention_order: use a different attention pattern for potentially
|
|
increased efficiency.
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
model_channels,
|
|
in_channels=1,
|
|
out_channels=2, # mean and variance
|
|
dropout=0,
|
|
# res 1, 2, 4, 8,16,32,64,128,256,512, 1K, 2K
|
|
channel_mult= (1,1.5,2, 3, 4, 6, 8, 12, 16, 24, 32, 48),
|
|
num_res_blocks=(1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2),
|
|
# spec_cond: 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0)
|
|
# attn: 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1
|
|
conv_resample=True,
|
|
dims=1,
|
|
use_fp16=False,
|
|
kernel_size=3,
|
|
scale_factor=2,
|
|
time_embed_dim_multiplier=4,
|
|
efficient_convs=True, # Uses kernels with width of 1 in several places rather than 3.
|
|
use_scale_shift_norm=True,
|
|
freeze_main=False,
|
|
# Parameters for regularization.
|
|
unconditioned_percentage=.1, # This implements a mechanism similar to what is used in classifier-free training.
|
|
):
|
|
super().__init__()
|
|
|
|
self.in_channels = in_channels
|
|
self.model_channels = model_channels
|
|
self.out_channels = out_channels
|
|
self.dropout = dropout
|
|
self.channel_mult = channel_mult
|
|
self.conv_resample = conv_resample
|
|
self.dims = dims
|
|
self.unconditioned_percentage = unconditioned_percentage
|
|
self.enable_fp16 = use_fp16
|
|
self.alignment_size = max(2 ** (len(channel_mult)+1), 256)
|
|
padding = 1 if kernel_size == 3 else 2
|
|
down_kernel = 1 if efficient_convs else 3
|
|
|
|
time_embed_dim = model_channels * time_embed_dim_multiplier
|
|
self.time_embed = nn.Sequential(
|
|
linear(model_channels, time_embed_dim),
|
|
nn.SiLU(),
|
|
linear(time_embed_dim, time_embed_dim),
|
|
)
|
|
|
|
self.structural_cond_input = nn.Conv1d(in_channels, model_channels, kernel_size=5, padding=2)
|
|
self.aligned_latent_padding_embedding = nn.Parameter(torch.zeros(1,in_channels,1))
|
|
self.unconditioned_embedding = nn.Parameter(torch.randn(1,model_channels,1))
|
|
self.structural_processor = AudioVAE(model_channels, dropout)
|
|
self.surrogate_head = nn.Conv1d(model_channels, in_channels, 1)
|
|
|
|
self.input_block = conv_nd(dims, in_channels, model_channels, kernel_size, padding=padding)
|
|
self.input_blocks = nn.ModuleList(
|
|
[
|
|
TimestepEmbedSequential(
|
|
conv_nd(dims, model_channels*2, model_channels, 1)
|
|
)
|
|
]
|
|
)
|
|
self._feature_size = model_channels
|
|
input_block_chans = [model_channels]
|
|
ch = model_channels
|
|
ds = 1
|
|
|
|
for level, (mult, num_blocks) in enumerate(zip(channel_mult, num_res_blocks)):
|
|
for _ in range(num_blocks):
|
|
layers = [
|
|
ResBlock(
|
|
ch,
|
|
time_embed_dim,
|
|
dropout,
|
|
out_channels=int(mult * model_channels),
|
|
dims=dims,
|
|
kernel_size=kernel_size,
|
|
efficient_config=efficient_convs,
|
|
use_scale_shift_norm=use_scale_shift_norm,
|
|
)
|
|
]
|
|
ch = int(mult * model_channels)
|
|
self.input_blocks.append(TimestepEmbedSequential(*layers))
|
|
self._feature_size += ch
|
|
input_block_chans.append(ch)
|
|
if level != len(channel_mult) - 1:
|
|
out_ch = ch
|
|
self.input_blocks.append(
|
|
TimestepEmbedSequential(
|
|
Downsample(
|
|
ch, conv_resample, dims=dims, out_channels=out_ch, factor=scale_factor, ksize=down_kernel, pad=0 if down_kernel == 1 else 1
|
|
)
|
|
)
|
|
)
|
|
ch = out_ch
|
|
input_block_chans.append(ch)
|
|
ds *= 2
|
|
self._feature_size += ch
|
|
|
|
self.middle_block = TimestepEmbedSequential(
|
|
ResBlock(
|
|
ch,
|
|
time_embed_dim,
|
|
dropout,
|
|
dims=dims,
|
|
kernel_size=kernel_size,
|
|
efficient_config=efficient_convs,
|
|
use_scale_shift_norm=use_scale_shift_norm,
|
|
),
|
|
)
|
|
self._feature_size += ch
|
|
|
|
self.output_blocks = nn.ModuleList([])
|
|
for level, (mult, num_blocks) in list(enumerate(zip(channel_mult, num_res_blocks)))[::-1]:
|
|
for i in range(num_blocks + 1):
|
|
ich = input_block_chans.pop()
|
|
layers = [
|
|
ResBlock(
|
|
ch + ich,
|
|
time_embed_dim,
|
|
dropout,
|
|
out_channels=int(model_channels * mult),
|
|
dims=dims,
|
|
kernel_size=kernel_size,
|
|
efficient_config=efficient_convs,
|
|
use_scale_shift_norm=use_scale_shift_norm,
|
|
)
|
|
]
|
|
ch = int(model_channels * mult)
|
|
if level and i == num_blocks:
|
|
out_ch = ch
|
|
layers.append(
|
|
Upsample(ch, conv_resample, dims=dims, out_channels=out_ch, factor=scale_factor)
|
|
)
|
|
ds //= 2
|
|
self.output_blocks.append(TimestepEmbedSequential(*layers))
|
|
self._feature_size += ch
|
|
|
|
self.out = nn.Sequential(
|
|
normalization(ch),
|
|
nn.SiLU(),
|
|
zero_module(conv_nd(dims, model_channels, out_channels, kernel_size, padding=padding)),
|
|
)
|
|
|
|
if freeze_main:
|
|
for p in self.parameters():
|
|
p.DO_NOT_TRAIN = True
|
|
p.requires_grad = False
|
|
for m in [self.structural_processor, self.structural_cond_input, self.surrogate_head]:
|
|
for p in m.parameters():
|
|
del p.DO_NOT_TRAIN
|
|
p.requires_grad = True
|
|
|
|
|
|
def get_grad_norm_parameter_groups(self):
|
|
groups = {
|
|
'input_blocks': list(self.input_blocks.parameters()),
|
|
'output_blocks': list(self.output_blocks.parameters()),
|
|
'middle_transformer': list(self.middle_block.parameters()),
|
|
'structural_processor': list(self.structural_processor.parameters()),
|
|
}
|
|
return groups
|
|
|
|
def fix_alignment(self, x, aligned_conditioning):
|
|
"""
|
|
The UNet requires that the input <x> is a certain multiple of 2, defined by the UNet depth. Enforce this by
|
|
padding both <x> and <aligned_conditioning> before forward propagation and removing the padding before returning.
|
|
"""
|
|
cm = ceil_multiple(x.shape[-1], self.alignment_size)
|
|
if cm != 0:
|
|
pc = (cm-x.shape[-1])/x.shape[-1]
|
|
x = F.pad(x, (0,cm-x.shape[-1]))
|
|
aligned_conditioning = F.pad(aligned_conditioning, (0,int(pc*aligned_conditioning.shape[-1])))
|
|
return x, aligned_conditioning
|
|
|
|
def forward(self, x, timesteps, conditioning, return_surrogate=True, conditioning_free=False):
|
|
"""
|
|
Apply the model to an input batch.
|
|
|
|
:param x: an [N x C x ...] Tensor of inputs.
|
|
:param timesteps: a 1-D batch of timesteps.
|
|
:param conditioning: should just be the truth value. produces a latent through an autoencoder, then uses diffusion to decode that latent.
|
|
at inference, only the latent is passed in.
|
|
:param conditioning_free: When set, all conditioning inputs (including tokens and conditioning_input) will not be considered.
|
|
:return: an [N x C x ...] Tensor of outputs.
|
|
"""
|
|
|
|
# Fix input size to the proper multiple of 2 so we don't get alignment errors going down and back up the U-net.
|
|
orig_x_shape = x.shape[-1]
|
|
x, aligned_conditioning = self.fix_alignment(x, conditioning)
|
|
|
|
with autocast(x.device.type, enabled=self.enable_fp16):
|
|
|
|
# Note: this block does not need to repeated on inference, since it is not timestep-dependent.
|
|
if conditioning_free:
|
|
code_emb = self.unconditioned_embedding.repeat(x.shape[0], 1, 1)
|
|
surrogate = torch.zeros_like(x)
|
|
else:
|
|
code_emb = self.structural_cond_input(aligned_conditioning)
|
|
code_emb = self.structural_processor(code_emb)
|
|
code_emb = F.interpolate(code_emb, size=(x.shape[-1],), mode='linear')
|
|
surrogate = self.surrogate_head(code_emb)
|
|
|
|
x = self.input_block(x)
|
|
x = torch.cat([x, code_emb], dim=1)
|
|
|
|
# Everything after this comment is timestep dependent.
|
|
hs = []
|
|
time_emb = self.time_embed(timestep_embedding(timesteps, self.model_channels))
|
|
time_emb = time_emb.float()
|
|
h = x
|
|
for k, module in enumerate(self.input_blocks):
|
|
with autocast(x.device.type, enabled=self.enable_fp16 and not first):
|
|
# First block has autocast disabled to allow a high precision signal to be properly vectorized.
|
|
h = module(h, time_emb)
|
|
hs.append(h)
|
|
h = self.middle_block(h, time_emb)
|
|
for module in self.output_blocks:
|
|
h = torch.cat([h, hs.pop()], dim=1)
|
|
h = module(h, time_emb)
|
|
|
|
# Last block also has autocast disabled for high-precision outputs.
|
|
h = h.float()
|
|
out = self.out(h)
|
|
|
|
# Involve probabilistic or possibly unused parameters in loss so we don't get DDP errors.
|
|
extraneous_addition = 0
|
|
params = [self.aligned_latent_padding_embedding, self.unconditioned_embedding]
|
|
for p in params:
|
|
extraneous_addition = extraneous_addition + p.mean()
|
|
out = out + extraneous_addition * 0
|
|
|
|
if return_surrogate:
|
|
return out[:, :, :orig_x_shape], surrogate[:, :, :orig_x_shape]
|
|
else:
|
|
return out[:, :, :orig_x_shape]
|
|
|
|
|
|
@register_model
|
|
def register_unet_diffusion_waveform_gen2(opt_net, opt):
|
|
return Diffusion(**opt_net['kwargs'])
|
|
|
|
|
|
if __name__ == '__main__':
|
|
clip = torch.randn(2, 1, 32868)
|
|
aligned_sequence = torch.randn(2,1,32868)
|
|
ts = torch.LongTensor([600, 600])
|
|
model = Diffusion(128,
|
|
channel_mult=[1,1.5,2, 3, 4, 6, 8],
|
|
num_res_blocks=[2, 2, 2, 2, 2, 2, 1],
|
|
kernel_size=3,
|
|
scale_factor=2,
|
|
time_embed_dim_multiplier=4,
|
|
efficient_convs=False)
|
|
# Test with sequence aligned conditioning
|
|
o = model(clip, ts, aligned_sequence)
|
|
|