forked from mrq/DL-Art-School
277 lines
11 KiB
Python
277 lines
11 KiB
Python
|
import random
|
||
|
|
|
||
|
|
import torch
|
||
|
|
import torch.nn as nn
|
||
|
|
import torch.nn.functional as F
|
||
|
|
from torch import autocast
|
||
|
|
from torchaudio.transforms import TimeMasking, FrequencyMasking
|
||
|
|
|
||
|
|
from models.diffusion.nn import timestep_embedding, normalization, zero_module, conv_nd, linear
|
||
|
|
from models.diffusion.unet_diffusion import AttentionBlock, TimestepEmbedSequential, TimestepBlock
|
||
|
|
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 DiffusionLayer(TimestepBlock):
|
||
|
|
def __init__(self, model_channels, dropout, num_heads):
|
||
|
|
super().__init__()
|
||
|
|
self.resblk = ResBlock(model_channels, model_channels, dropout, model_channels, dims=1, use_scale_shift_norm=True)
|
||
|
|
self.attn = AttentionBlock(model_channels, num_heads, relative_pos_embeddings=True)
|
||
|
|
|
||
|
|
def forward(self, x, time_emb):
|
||
|
|
y = self.resblk(x, time_emb)
|
||
|
|
return self.attn(y)
|
||
|
|
|
||
|
|
|
||
|
|
class MusicGenerator(nn.Module):
|
||
|
|
def __init__(
|
||
|
|
self,
|
||
|
|
model_channels=512,
|
||
|
|
num_layers=8,
|
||
|
|
in_channels=100,
|
||
|
|
out_channels=200, # mean and variance
|
||
|
|
dropout=0,
|
||
|
|
use_fp16=False,
|
||
|
|
num_heads=16,
|
||
|
|
# Parameters for regularization.
|
||
|
|
layer_drop=.1,
|
||
|
|
unconditioned_percentage=.1, # This implements a mechanism similar to what is used in classifier-free training.
|
||
|
|
# Masking parameters.
|
||
|
|
time_mask_percent_max=.4,
|
||
|
|
frequency_mask_percent_max=.4,
|
||
|
|
):
|
||
|
|
super().__init__()
|
||
|
|
|
||
|
|
self.in_channels = in_channels
|
||
|
|
self.model_channels = model_channels
|
||
|
|
self.out_channels = out_channels
|
||
|
|
self.dropout = dropout
|
||
|
|
self.num_heads = num_heads
|
||
|
|
self.unconditioned_percentage = unconditioned_percentage
|
||
|
|
self.enable_fp16 = use_fp16
|
||
|
|
self.layer_drop = layer_drop
|
||
|
|
self.time_mask_percent_max = time_mask_percent_max
|
||
|
|
self.frequency_mask_percent_mask = frequency_mask_percent_max
|
||
|
|
|
||
|
|
self.inp_block = conv_nd(1, in_channels, model_channels, 3, 1, 1)
|
||
|
|
self.time_embed = nn.Sequential(
|
||
|
|
linear(model_channels, model_channels),
|
||
|
|
nn.SiLU(),
|
||
|
|
linear(model_channels, model_channels),
|
||
|
|
)
|
||
|
|
|
||
|
|
self.conditioner = nn.Sequential(
|
||
|
|
nn.Conv1d(in_channels, model_channels, 3, padding=1),
|
||
|
|
AttentionBlock(model_channels, num_heads, relative_pos_embeddings=True),
|
||
|
|
AttentionBlock(model_channels, num_heads, relative_pos_embeddings=True),
|
||
|
|
)
|
||
|
|
self.unconditioned_embedding = nn.Parameter(torch.randn(1,model_channels,1))
|
||
|
|
self.conditioning_timestep_integrator = TimestepEmbedSequential(
|
||
|
|
DiffusionLayer(model_channels, dropout, num_heads),
|
||
|
|
DiffusionLayer(model_channels, dropout, num_heads),
|
||
|
|
)
|
||
|
|
self.integrating_conv = nn.Conv1d(model_channels*2, model_channels, kernel_size=1)
|
||
|
|
self.mel_head = nn.Conv1d(model_channels, in_channels, kernel_size=3, padding=1)
|
||
|
|
self.layers = nn.ModuleList([DiffusionLayer(model_channels, dropout, num_heads) for _ in range(num_layers)] +
|
||
|
|
[ResBlock(model_channels, model_channels, dropout, dims=1, use_scale_shift_norm=True) for _ in range(3)])
|
||
|
|
|
||
|
|
self.out = nn.Sequential(
|
||
|
|
normalization(model_channels),
|
||
|
|
nn.SiLU(),
|
||
|
|
zero_module(conv_nd(1, model_channels, out_channels, 3, padding=1)),
|
||
|
|
)
|
||
|
|
|
||
|
|
def get_grad_norm_parameter_groups(self):
|
||
|
|
groups = {
|
||
|
|
'layers': list(self.layers.parameters()),
|
||
|
|
'conditioner': list(self.conditioner.parameters()) + list(self.conditioner.parameters()),
|
||
|
|
'timestep_integrator': list(self.conditioning_timestep_integrator.parameters()) + list(self.integrating_conv.parameters()),
|
||
|
|
'time_embed': list(self.time_embed.parameters()),
|
||
|
|
}
|
||
|
|
return groups
|
||
|
|
|
||
|
|
def do_masking(self, truth):
|
||
|
|
b, c, s = truth.shape
|
||
|
|
mask = torch.ones_like(truth)
|
||
|
|
if random.random() < .5:
|
||
|
|
# Frequency mask
|
||
|
|
cs = random.randint(0, c-10)
|
||
|
|
ce = min(c-1, cs+random.randint(1, int(self.frequency_mask_percent_mask*c)))
|
||
|
|
mask[:, cs:ce] = 0
|
||
|
|
else:
|
||
|
|
# Time mask
|
||
|
|
cs = random.randint(0, s-5)
|
||
|
|
ce = min(s-1, cs+random.randint(1, int(self.frequency_mask_percent_mask*s)))
|
||
|
|
mask[:, :, cs:ce] = 0
|
||
|
|
return truth * mask
|
||
|
|
|
||
|
|
|
||
|
|
def timestep_independent(self, aligned_conditioning, expected_seq_len, return_code_pred):
|
||
|
|
# Shuffle aligned_latent to BxCxS format
|
||
|
|
aligned_conditioning = aligned_conditioning.permute(0, 2, 1)
|
||
|
|
code_emb = self.conditioner(aligned_conditioning)
|
||
|
|
unconditioned_batches = torch.zeros((code_emb.shape[0], 1, 1), device=code_emb.device)
|
||
|
|
# Mask out the conditioning branch for whole batch elements, implementing something similar to classifier-free guidance.
|
||
|
|
if self.training and self.unconditioned_percentage > 0:
|
||
|
|
unconditioned_batches = torch.rand((code_emb.shape[0], 1, 1),
|
||
|
|
device=code_emb.device) < self.unconditioned_percentage
|
||
|
|
code_emb = torch.where(unconditioned_batches, self.unconditioned_embedding.repeat(aligned_conditioning.shape[0], 1, 1),
|
||
|
|
code_emb)
|
||
|
|
expanded_code_emb = F.interpolate(code_emb, size=expected_seq_len, mode='nearest')
|
||
|
|
|
||
|
|
if not return_code_pred:
|
||
|
|
return expanded_code_emb
|
||
|
|
else:
|
||
|
|
mel_pred = self.mel_head(expanded_code_emb)
|
||
|
|
# Multiply mel_pred by !unconditioned_branches, which drops the gradient on unconditioned branches. This is because we don't want that gradient being used to train parameters through the codes_embedder as it unbalances contributions to that network from the MSE loss.
|
||
|
|
mel_pred = mel_pred * unconditioned_batches.logical_not()
|
||
|
|
return expanded_code_emb, mel_pred
|
||
|
|
|
||
|
|
|
||
|
|
def forward(self, x, timesteps, truth=None, precomputed_aligned_embeddings=None, conditioning_free=False, return_code_pred=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 truth: Input value is either pre-masked (in inference), or unmasked (during training)
|
||
|
|
:param precomputed_aligned_embeddings: Embeddings returned from self.timestep_independent()
|
||
|
|
: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.
|
||
|
|
"""
|
||
|
|
assert precomputed_aligned_embeddings is not None or truth is not None
|
||
|
|
assert not (return_code_pred and precomputed_aligned_embeddings is not None) # These two are mutually exclusive.
|
||
|
|
|
||
|
|
unused_params = []
|
||
|
|
if conditioning_free:
|
||
|
|
code_emb = self.unconditioned_embedding.repeat(x.shape[0], 1, x.shape[-1])
|
||
|
|
unused_params.extend(list(self.conditioner.parameters()))
|
||
|
|
else:
|
||
|
|
if precomputed_aligned_embeddings is not None:
|
||
|
|
code_emb = precomputed_aligned_embeddings
|
||
|
|
else:
|
||
|
|
truth = self.do_masking(truth)
|
||
|
|
code_emb, mel_pred = self.timestep_independent(truth, x.shape[-1], True)
|
||
|
|
unused_params.append(self.unconditioned_embedding)
|
||
|
|
|
||
|
|
time_emb = self.time_embed(timestep_embedding(timesteps, self.model_channels))
|
||
|
|
code_emb = self.conditioning_timestep_integrator(code_emb, time_emb)
|
||
|
|
x = self.inp_block(x)
|
||
|
|
x = torch.cat([x, code_emb], dim=1)
|
||
|
|
x = self.integrating_conv(x)
|
||
|
|
for i, lyr in enumerate(self.layers):
|
||
|
|
# Do layer drop where applicable. Do not drop first and last layers.
|
||
|
|
if self.training and self.layer_drop > 0 and i != 0 and i != (len(self.layers)-1) and random.random() < self.layer_drop:
|
||
|
|
unused_params.extend(list(lyr.parameters()))
|
||
|
|
else:
|
||
|
|
# First and last blocks will have autocast disabled for improved precision.
|
||
|
|
with autocast(x.device.type, enabled=self.enable_fp16 and i != 0):
|
||
|
|
x = lyr(x, time_emb)
|
||
|
|
|
||
|
|
x = x.float()
|
||
|
|
out = self.out(x)
|
||
|
|
|
||
|
|
# Involve probabilistic or possibly unused parameters in loss so we don't get DDP errors.
|
||
|
|
extraneous_addition = 0
|
||
|
|
for p in unused_params:
|
||
|
|
extraneous_addition = extraneous_addition + p.mean()
|
||
|
|
out = out + extraneous_addition * 0
|
||
|
|
|
||
|
|
if return_code_pred:
|
||
|
|
return out, mel_pred
|
||
|
|
return out
|
||
|
|
|
||
|
|
|
||
|
|
@register_model
|
||
|
|
def register_music_gap_gen(opt_net, opt):
|
||
|
|
return MusicGenerator(**opt_net['kwargs'])
|
||
|
|
|
||
|
|
|
||
|
|
if __name__ == '__main__':
|
||
|
|
clip = torch.randn(2, 100, 400)
|
||
|
|
aligned_latent = torch.randn(2,388,100)
|
||
|
|
ts = torch.LongTensor([600, 600])
|
||
|
|
model = MusicGenerator(512, layer_drop=.3, unconditioned_percentage=.5)
|
||
|
|
o = model(clip, ts, aligned_latent)
|
||
|
|
|