213 lines
8.7 KiB
Python
213 lines
8.7 KiB
Python
import torch
|
|
import torch.nn as nn
|
|
import torch.nn.functional as F
|
|
|
|
from models.diffusion.nn import timestep_embedding, normalization, zero_module, conv_nd, linear
|
|
from models.diffusion.unet_diffusion import TimestepBlock
|
|
from models.lucidrains.x_transformers import Encoder, Attention, FeedForward, RMSScaleShiftNorm, RotaryEmbedding
|
|
from trainer.networks import register_model
|
|
from utils.util import checkpoint, print_network
|
|
|
|
|
|
def is_latent(t):
|
|
return t.dtype == torch.float
|
|
|
|
def is_sequence(t):
|
|
return t.dtype == torch.long
|
|
|
|
|
|
class MultiGroupEmbedding(nn.Module):
|
|
def __init__(self, tokens, groups, dim):
|
|
super().__init__()
|
|
self.m = nn.ModuleList([nn.Embedding(tokens, dim // groups) for _ in range(groups)])
|
|
|
|
def forward(self, x):
|
|
h = [embedding(x[:, :, i]) for i, embedding in enumerate(self.m)]
|
|
return torch.cat(h, dim=-1)
|
|
|
|
|
|
class TimestepRotaryEmbedSequential(nn.Sequential, TimestepBlock):
|
|
def forward(self, x, emb, rotary_emb):
|
|
for layer in self:
|
|
if isinstance(layer, TimestepBlock):
|
|
x = layer(x, emb, rotary_emb)
|
|
else:
|
|
x = layer(x, rotary_emb)
|
|
return x
|
|
|
|
|
|
class DietAttentionBlock(TimestepBlock):
|
|
def __init__(self, in_dim, dim, heads, dropout):
|
|
super().__init__()
|
|
self.rms_scale_norm = RMSScaleShiftNorm(in_dim)
|
|
self.proj = nn.Linear(in_dim, dim)
|
|
self.attn = Attention(dim, heads=heads, causal=False, dropout=dropout)
|
|
self.ff = FeedForward(dim, in_dim, mult=1, dropout=dropout, zero_init_output=True)
|
|
|
|
def forward(self, x, timestep_emb, rotary_emb):
|
|
h = self.rms_scale_norm(x, norm_scale_shift_inp=timestep_emb)
|
|
h = self.proj(h)
|
|
h, _, _, _ = checkpoint(self.attn, h, None, None, None, None, None, rotary_emb)
|
|
h = checkpoint(self.ff, h)
|
|
return h + x
|
|
|
|
|
|
class TransformerDiffusion(nn.Module):
|
|
"""
|
|
A diffusion model composed entirely of stacks of transformer layers. Why would you do it any other way?
|
|
"""
|
|
def __init__(
|
|
self,
|
|
prenet_channels=256,
|
|
model_channels=512,
|
|
block_channels=256,
|
|
num_layers=8,
|
|
in_channels=256,
|
|
rotary_emb_dim=32,
|
|
input_vec_dim=512,
|
|
out_channels=512, # mean and variance
|
|
dropout=0,
|
|
use_fp16=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.prenet_channels = prenet_channels
|
|
self.out_channels = out_channels
|
|
self.dropout = dropout
|
|
self.unconditioned_percentage = unconditioned_percentage
|
|
self.enable_fp16 = use_fp16
|
|
|
|
self.inp_block = conv_nd(1, in_channels, prenet_channels, 3, 1, 1)
|
|
|
|
self.time_embed = nn.Sequential(
|
|
linear(prenet_channels, prenet_channels),
|
|
nn.SiLU(),
|
|
linear(prenet_channels, prenet_channels),
|
|
)
|
|
prenet_heads = prenet_channels//64
|
|
self.conditioning_embedder = nn.Sequential(nn.Conv1d(in_channels, prenet_channels // 2, 3, padding=1, stride=2),
|
|
nn.Conv1d(prenet_channels//2, prenet_channels,3,padding=1,stride=2))
|
|
self.conditioning_encoder = Encoder(
|
|
dim=prenet_channels,
|
|
depth=4,
|
|
heads=prenet_heads,
|
|
ff_dropout=dropout,
|
|
attn_dropout=dropout,
|
|
use_rmsnorm=True,
|
|
ff_glu=True,
|
|
rotary_pos_emb=True,
|
|
zero_init_branch_output=True,
|
|
ff_mult=1,
|
|
)
|
|
|
|
self.input_converter = nn.Linear(input_vec_dim, prenet_channels)
|
|
self.code_converter = Encoder(
|
|
dim=prenet_channels,
|
|
depth=3,
|
|
heads=prenet_heads,
|
|
ff_dropout=dropout,
|
|
attn_dropout=dropout,
|
|
use_rmsnorm=True,
|
|
ff_glu=True,
|
|
rotary_pos_emb=True,
|
|
zero_init_branch_output=True,
|
|
ff_mult=1,
|
|
)
|
|
|
|
self.unconditioned_embedding = nn.Parameter(torch.randn(1,1,prenet_channels))
|
|
self.rotary_embeddings = RotaryEmbedding(rotary_emb_dim)
|
|
self.cond_intg = nn.Linear(prenet_channels*2, model_channels)
|
|
self.intg = nn.Linear(prenet_channels*2, model_channels)
|
|
self.layers = TimestepRotaryEmbedSequential(*[DietAttentionBlock(model_channels, block_channels, block_channels // 64, dropout) for _ in range(num_layers)])
|
|
|
|
self.out = nn.Sequential(
|
|
normalization(model_channels),
|
|
nn.SiLU(),
|
|
zero_module(conv_nd(1, model_channels, out_channels, 3, padding=1)),
|
|
)
|
|
|
|
self.debug_codes = {}
|
|
|
|
def get_grad_norm_parameter_groups(self):
|
|
groups = {
|
|
'contextual_embedder': list(self.conditioning_embedder.parameters()),
|
|
'layers': list(self.layers.parameters()) + list(self.inp_block.parameters()),
|
|
'code_converters': list(self.embeddings.parameters()) + list(self.code_converter.parameters()),
|
|
'time_embed': list(self.time_embed.parameters()),
|
|
}
|
|
return groups
|
|
|
|
def timestep_independent(self, codes, conditioning_input, expected_seq_len):
|
|
cond_emb = self.conditioning_embedder(conditioning_input).permute(0,2,1)
|
|
cond_emb = self.conditioning_encoder(cond_emb)[:, 0]
|
|
code_emb = self.input_converter(codes)
|
|
|
|
# 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(codes.shape[0], 1, 1),
|
|
code_emb)
|
|
code_emb = self.code_converter(code_emb)
|
|
|
|
expanded_code_emb = F.interpolate(code_emb.permute(0,2,1), size=expected_seq_len, mode='nearest').permute(0,2,1)
|
|
return expanded_code_emb, cond_emb
|
|
|
|
def forward(self, x, timesteps, codes=None, conditioning_input=None, precomputed_code_embeddings=None,
|
|
precomputed_cond_embeddings=None, conditioning_free=False):
|
|
if precomputed_code_embeddings is not None:
|
|
assert codes is None and conditioning_input is None, "Do not provide precomputed embeddings and the other parameters. It is unclear what you want me to do here."
|
|
|
|
unused_params = []
|
|
if conditioning_free:
|
|
code_emb = self.unconditioned_embedding.repeat(x.shape[0], 1, x.shape[-1])
|
|
unused_params.extend(list(self.code_converter.parameters()) + list(self.code_embedding.parameters()))
|
|
else:
|
|
if precomputed_code_embeddings is not None:
|
|
code_emb = precomputed_code_embeddings
|
|
cond_emb = precomputed_cond_embeddings
|
|
else:
|
|
code_emb, cond_emb = self.timestep_independent(codes, conditioning_input, x.shape[-1])
|
|
unused_params.append(self.unconditioned_embedding)
|
|
|
|
blk_emb = torch.cat([self.time_embed(timestep_embedding(timesteps, self.prenet_channels)), cond_emb], dim=-1)
|
|
blk_emb = self.cond_intg(blk_emb)
|
|
x = self.inp_block(x).permute(0,2,1)
|
|
|
|
rotary_pos_emb = self.rotary_embeddings(x.shape[1], x.device)
|
|
x = self.intg(torch.cat([x, code_emb], dim=-1))
|
|
for layer in self.layers:
|
|
x = checkpoint(layer, x, blk_emb, rotary_pos_emb)
|
|
|
|
x = x.float().permute(0,2,1)
|
|
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
|
|
|
|
return out
|
|
|
|
|
|
@register_model
|
|
def register_transformer_diffusion4(opt_net, opt):
|
|
return TransformerDiffusion(**opt_net['kwargs'])
|
|
|
|
|
|
if __name__ == '__main__':
|
|
clip = torch.randn(2, 256, 400)
|
|
aligned_sequence = torch.randn(2,100,512)
|
|
cond = torch.randn(2, 256, 400)
|
|
ts = torch.LongTensor([600, 600])
|
|
model = TransformerDiffusion(model_channels=3072, block_channels=1536, prenet_channels=1536)
|
|
torch.save(model, 'sample.pth')
|
|
print_network(model)
|
|
o = model(clip, ts, aligned_sequence, cond)
|
|
|