LMM / mogen /models /attentions /base_attention.py
mingyuan's picture
initial commit
373af33
raw
history blame
8.3 kB
import torch
import torch.nn as nn
import torch.nn.functional as F
from typing import Optional, Dict, Any
from ..builder import ATTENTIONS
from ..utils.stylization_block import StylizationBlock
@ATTENTIONS.register_module()
class BaseMixedAttention(nn.Module):
"""
Base class for Mixed Attention, combining text and motion attention.
Args:
latent_dim (int): Dimension of the latent space for motion input.
text_latent_dim (int): Dimension of the latent space for text input.
num_heads (int): Number of attention heads.
dropout (float): Dropout probability.
time_embed_dim (int): Dimension of the time embedding.
"""
def __init__(self, latent_dim: int, text_latent_dim: int, num_heads: int, dropout: float, time_embed_dim: int):
super().__init__()
self.num_heads = num_heads
self.norm = nn.LayerNorm(latent_dim)
self.text_norm = nn.LayerNorm(text_latent_dim)
self.query = nn.Linear(latent_dim, latent_dim)
self.key_text = nn.Linear(text_latent_dim, latent_dim)
self.value_text = nn.Linear(text_latent_dim, latent_dim)
self.key_motion = nn.Linear(latent_dim, latent_dim)
self.value_motion = nn.Linear(latent_dim, latent_dim)
self.dropout = nn.Dropout(dropout)
self.proj_out = StylizationBlock(latent_dim, time_embed_dim, dropout)
def forward(self, x: torch.Tensor, xf: torch.Tensor, emb: torch.Tensor, src_mask: torch.Tensor,
cond_type: torch.Tensor, **kwargs: Dict[str, Any]) -> torch.Tensor:
"""
Forward pass of Mixed Attention.
Args:
x (torch.Tensor): Input motion tensor of shape [B, T, D].
xf (torch.Tensor): Input text tensor of shape [B, N, L].
emb (torch.Tensor): Time embedding tensor of shape [B, D].
src_mask (torch.Tensor): Source mask tensor of shape [B, T].
cond_type (torch.Tensor): Conditioning type tensor of shape [B].
Returns:
torch.Tensor: Output of the mixed attention module.
"""
B, T, D = x.shape
N = xf.shape[1] + x.shape[1]
H = self.num_heads
query = self.query(self.norm(x)).view(B, T, H, -1)
# Text conditioning type
text_cond_type = ((cond_type % 10) > 0).float().view(B, 1, 1)
text_cond_type = text_cond_type.repeat(1, xf.shape[1], 1)
key = torch.cat(
(self.key_text(self.text_norm(xf)), self.key_motion(self.norm(x))),
dim=1
).view(B, N, H, -1)
attention = torch.einsum('bnhl,bmhl->bnmh', query, key)
motion_mask = src_mask.view(B, 1, T, 1)
text_mask = text_cond_type.view(B, 1, -1, 1)
mask = torch.cat((text_mask, motion_mask), dim=2)
attention = attention + (1 - mask) * -1000000 # Masking for softmax
attention = F.softmax(attention, dim=2)
value = torch.cat(
(self.value_text(self.text_norm(xf)) * text_cond_type, self.value_motion(self.norm(x)) * src_mask),
dim=1
).view(B, N, H, -1)
y = torch.einsum('bnmh,bmhl->bnhl', attention, value).reshape(B, T, D)
y = x + self.proj_out(y, emb)
return y
@ATTENTIONS.register_module()
class BaseSelfAttention(nn.Module):
"""
Base class for Self-Attention mechanism.
Args:
latent_dim (int): Dimension of the latent space.
num_heads (int): Number of attention heads.
dropout (float): Dropout probability.
time_embed_dim (Optional[int]): Dimension of the time embedding (optional).
"""
def __init__(self, latent_dim: int, num_heads: int, dropout: float, time_embed_dim: Optional[int] = None):
super().__init__()
self.num_heads = num_heads
self.norm = nn.LayerNorm(latent_dim)
self.query = nn.Linear(latent_dim, latent_dim)
self.key = nn.Linear(latent_dim, latent_dim)
self.value = nn.Linear(latent_dim, latent_dim)
self.dropout = nn.Dropout(dropout)
self.time_embed_dim = time_embed_dim
if time_embed_dim is not None:
self.proj_out = StylizationBlock(latent_dim, time_embed_dim, dropout)
def forward(self, x: torch.Tensor, src_mask: Optional[torch.Tensor] = None, emb: Optional[torch.Tensor] = None, **kwargs: Dict[str, Any]) -> torch.Tensor:
"""
Forward pass of Self-Attention.
Args:
x (torch.Tensor): Input tensor of shape [B, T, D].
emb (torch.Tensor): Time embedding tensor of shape [B, D].
src_mask (torch.Tensor): Source mask tensor of shape [B, T].
Returns:
torch.Tensor: Output of the self-attention module.
"""
B, T, D = x.shape
H = self.num_heads
query = self.query(self.norm(x)).view(B, T, H, -1)
key = self.key(self.norm(x)).view(B, T, H, -1)
attention = torch.einsum('bnhl,bmhl->bnmh', query, key)
if src_mask is not None:
mask = src_mask.view(B, 1, T, 1)
attention = attention + (1 - mask) * -1000000 # Masking for softmax
attention = F.softmax(attention, dim=2)
if src_mask is not None:
value = (self.value(self.norm(x)) * src_mask).view(B, T, H, -1)
else:
value = self.value(self.norm(x)).view(B, T, H, -1)
y = torch.einsum('bnmh,bmhl->bnhl', attention, value).reshape(B, T, D)
if self.time_embed_dim is None:
y = x + y
else:
y = x + self.proj_out(y, emb)
return y
@ATTENTIONS.register_module()
class BaseCrossAttention(nn.Module):
"""
Base class for Cross-Attention mechanism, attending over text and motion inputs.
Args:
latent_dim (int): Dimension of the latent space for motion input.
text_latent_dim (int): Dimension of the latent space for text input.
num_heads (int): Number of attention heads.
dropout (float): Dropout probability.
time_embed_dim (int): Dimension of the time embedding.
"""
def __init__(self, latent_dim: int, text_latent_dim: int, num_heads: int, dropout: float, time_embed_dim: int):
super().__init__()
self.num_heads = num_heads
self.norm = nn.LayerNorm(latent_dim)
self.text_norm = nn.LayerNorm(text_latent_dim)
self.query = nn.Linear(latent_dim, latent_dim)
self.key = nn.Linear(text_latent_dim, latent_dim)
self.value = nn.Linear(text_latent_dim, latent_dim)
self.dropout = nn.Dropout(dropout)
self.proj_out = StylizationBlock(latent_dim, time_embed_dim, dropout)
def forward(self, x: torch.Tensor, xf: torch.Tensor, emb: torch.Tensor, src_mask: torch.Tensor,
cond_type: Optional[torch.Tensor] = None, **kwargs: Dict[str, Any]) -> torch.Tensor:
"""
Forward pass of Cross-Attention.
Args:
x (torch.Tensor): Input motion tensor of shape [B, T, D].
xf (torch.Tensor): Input text tensor of shape [B, N, L].
emb (torch.Tensor): Time embedding tensor of shape [B, D].
src_mask (torch.Tensor): Source mask tensor of shape [B, T].
cond_type (Optional[torch.Tensor]): Conditioning type tensor of shape [B]. Defaults to None.
Returns:
torch.Tensor: Output of the cross-attention module.
"""
B, T, D = x.shape
N = xf.shape[1]
H = self.num_heads
query = self.query(self.norm(x)).view(B, T, H, -1)
if cond_type is None:
text_cond_type = 1
mask = 1
else:
text_cond_type = ((cond_type % 10) > 0).float().view(B, 1, 1)
text_cond_type = text_cond_type.repeat(1, xf.shape[1], 1)
mask = text_cond_type.view(B, 1, -1, 1)
key = self.key(self.text_norm(xf)).view(B, N, H, -1)
attention = torch.einsum('bnhl,bmhl->bnmh', query, key)
attention = attention + (1 - mask) * -1000000 # Masking for softmax
attention = F.softmax(attention, dim=2)
value = (self.value(self.text_norm(xf)) * text_cond_type).view(B, N, H, -1)
y = torch.einsum('bnmh,bmhl->bnhl', attention, value).reshape(B, T, D)
y = x + self.proj_out(y, emb)
return y