mirror of
https://github.com/hpcaitech/ColossalAI.git
synced 2025-09-08 04:24:47 +00:00
add ColoDiffusion codes: /ldm/module/, /ldm/data/, /scripts/test/
This commit is contained in:
Maruyama_Aya
314
examples/images/diffusion/ldm/modules/attention.py
Normal file
314
examples/images/diffusion/ldm/modules/attention.py
Normal file
@@ -0,0 +1,314 @@
|
||||
from inspect import isfunction
|
||||
import math
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
from torch import nn, einsum
|
||||
from einops import rearrange, repeat
|
||||
|
||||
from torch.utils import checkpoint
|
||||
|
||||
try:
|
||||
from ldm.modules.flash_attention import flash_attention_qkv, flash_attention_q_kv
|
||||
FlASH_AVAILABLE = True
|
||||
except:
|
||||
FlASH_AVAILABLE = False
|
||||
|
||||
USE_FLASH = False
|
||||
|
||||
|
||||
def enable_flash_attention():
|
||||
global USE_FLASH
|
||||
USE_FLASH = True
|
||||
if FlASH_AVAILABLE is False:
|
||||
print("Please install flash attention to activate new attention kernel.\n" +
|
||||
"Use \'pip install git+https://github.com/HazyResearch/flash-attention.git@c422fee3776eb3ea24e011ef641fd5fbeb212623#egg=flash_attn\'")
|
||||
|
||||
|
||||
def exists(val):
|
||||
return val is not None
|
||||
|
||||
|
||||
def uniq(arr):
|
||||
return{el: True for el in arr}.keys()
|
||||
|
||||
|
||||
def default(val, d):
|
||||
if exists(val):
|
||||
return val
|
||||
return d() if isfunction(d) else d
|
||||
|
||||
|
||||
def max_neg_value(t):
|
||||
return -torch.finfo(t.dtype).max
|
||||
|
||||
|
||||
def init_(tensor):
|
||||
dim = tensor.shape[-1]
|
||||
std = 1 / math.sqrt(dim)
|
||||
tensor.uniform_(-std, std)
|
||||
return tensor
|
||||
|
||||
|
||||
# feedforward
|
||||
class GEGLU(nn.Module):
|
||||
def __init__(self, dim_in, dim_out):
|
||||
super().__init__()
|
||||
self.proj = nn.Linear(dim_in, dim_out * 2)
|
||||
|
||||
def forward(self, x):
|
||||
x, gate = self.proj(x).chunk(2, dim=-1)
|
||||
return x * F.gelu(gate)
|
||||
|
||||
|
||||
class FeedForward(nn.Module):
|
||||
def __init__(self, dim, dim_out=None, mult=4, glu=False, dropout=0.):
|
||||
super().__init__()
|
||||
inner_dim = int(dim * mult)
|
||||
dim_out = default(dim_out, dim)
|
||||
project_in = nn.Sequential(
|
||||
nn.Linear(dim, inner_dim),
|
||||
nn.GELU()
|
||||
) if not glu else GEGLU(dim, inner_dim)
|
||||
|
||||
self.net = nn.Sequential(
|
||||
project_in,
|
||||
nn.Dropout(dropout),
|
||||
nn.Linear(inner_dim, dim_out)
|
||||
)
|
||||
|
||||
def forward(self, x):
|
||||
return self.net(x)
|
||||
|
||||
|
||||
def zero_module(module):
|
||||
"""
|
||||
Zero out the parameters of a module and return it.
|
||||
"""
|
||||
for p in module.parameters():
|
||||
p.detach().zero_()
|
||||
return module
|
||||
|
||||
|
||||
def Normalize(in_channels):
|
||||
return torch.nn.GroupNorm(num_groups=32, num_channels=in_channels, eps=1e-6, affine=True)
|
||||
|
||||
|
||||
class LinearAttention(nn.Module):
|
||||
def __init__(self, dim, heads=4, dim_head=32):
|
||||
super().__init__()
|
||||
self.heads = heads
|
||||
hidden_dim = dim_head * heads
|
||||
self.to_qkv = nn.Conv2d(dim, hidden_dim * 3, 1, bias = False)
|
||||
self.to_out = nn.Conv2d(hidden_dim, dim, 1)
|
||||
|
||||
def forward(self, x):
|
||||
b, c, h, w = x.shape
|
||||
qkv = self.to_qkv(x)
|
||||
q, k, v = rearrange(qkv, 'b (qkv heads c) h w -> qkv b heads c (h w)', heads = self.heads, qkv=3)
|
||||
k = k.softmax(dim=-1)
|
||||
context = torch.einsum('bhdn,bhen->bhde', k, v)
|
||||
out = torch.einsum('bhde,bhdn->bhen', context, q)
|
||||
out = rearrange(out, 'b heads c (h w) -> b (heads c) h w', heads=self.heads, h=h, w=w)
|
||||
return self.to_out(out)
|
||||
|
||||
|
||||
class SpatialSelfAttention(nn.Module):
|
||||
def __init__(self, in_channels):
|
||||
super().__init__()
|
||||
self.in_channels = in_channels
|
||||
|
||||
self.norm = Normalize(in_channels)
|
||||
self.q = torch.nn.Conv2d(in_channels,
|
||||
in_channels,
|
||||
kernel_size=1,
|
||||
stride=1,
|
||||
padding=0)
|
||||
self.k = torch.nn.Conv2d(in_channels,
|
||||
in_channels,
|
||||
kernel_size=1,
|
||||
stride=1,
|
||||
padding=0)
|
||||
self.v = torch.nn.Conv2d(in_channels,
|
||||
in_channels,
|
||||
kernel_size=1,
|
||||
stride=1,
|
||||
padding=0)
|
||||
self.proj_out = torch.nn.Conv2d(in_channels,
|
||||
in_channels,
|
||||
kernel_size=1,
|
||||
stride=1,
|
||||
padding=0)
|
||||
|
||||
def forward(self, x):
|
||||
h_ = x
|
||||
h_ = self.norm(h_)
|
||||
q = self.q(h_)
|
||||
k = self.k(h_)
|
||||
v = self.v(h_)
|
||||
|
||||
# compute attention
|
||||
b,c,h,w = q.shape
|
||||
q = rearrange(q, 'b c h w -> b (h w) c')
|
||||
k = rearrange(k, 'b c h w -> b c (h w)')
|
||||
w_ = torch.einsum('bij,bjk->bik', q, k)
|
||||
|
||||
w_ = w_ * (int(c)**(-0.5))
|
||||
w_ = torch.nn.functional.softmax(w_, dim=2)
|
||||
|
||||
# attend to values
|
||||
v = rearrange(v, 'b c h w -> b c (h w)')
|
||||
w_ = rearrange(w_, 'b i j -> b j i')
|
||||
h_ = torch.einsum('bij,bjk->bik', v, w_)
|
||||
h_ = rearrange(h_, 'b c (h w) -> b c h w', h=h)
|
||||
h_ = self.proj_out(h_)
|
||||
|
||||
return x+h_
|
||||
|
||||
|
||||
class CrossAttention(nn.Module):
|
||||
def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.):
|
||||
super().__init__()
|
||||
inner_dim = dim_head * heads
|
||||
context_dim = default(context_dim, query_dim)
|
||||
|
||||
self.scale = dim_head ** -0.5
|
||||
self.heads = heads
|
||||
|
||||
self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
|
||||
self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
|
||||
self.to_v = nn.Linear(context_dim, inner_dim, bias=False)
|
||||
|
||||
self.to_out = nn.Sequential(
|
||||
nn.Linear(inner_dim, query_dim),
|
||||
nn.Dropout(dropout)
|
||||
)
|
||||
|
||||
def forward(self, x, context=None, mask=None):
|
||||
q = self.to_q(x)
|
||||
context = default(context, x)
|
||||
k = self.to_k(context)
|
||||
v = self.to_v(context)
|
||||
dim_head = q.shape[-1] / self.heads
|
||||
|
||||
if USE_FLASH and FlASH_AVAILABLE and q.dtype in (torch.float16, torch.bfloat16) and \
|
||||
dim_head <= 128 and (dim_head % 8) == 0:
|
||||
# print("in flash")
|
||||
if q.shape[1] == k.shape[1]:
|
||||
out = self._flash_attention_qkv(q, k, v)
|
||||
else:
|
||||
out = self._flash_attention_q_kv(q, k, v)
|
||||
else:
|
||||
out = self._native_attention(q, k, v, self.heads, mask)
|
||||
|
||||
return self.to_out(out)
|
||||
|
||||
def _native_attention(self, q, k, v, h, mask):
|
||||
q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))
|
||||
sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
|
||||
if exists(mask):
|
||||
mask = rearrange(mask, 'b ... -> b (...)')
|
||||
max_neg_value = -torch.finfo(sim.dtype).max
|
||||
mask = repeat(mask, 'b j -> (b h) () j', h=h)
|
||||
sim.masked_fill_(~mask, max_neg_value)
|
||||
# attention, what we cannot get enough of
|
||||
out = sim.softmax(dim=-1)
|
||||
out = einsum('b i j, b j d -> b i d', out, v)
|
||||
out = rearrange(out, '(b h) n d -> b n (h d)', h=h)
|
||||
return out
|
||||
|
||||
def _flash_attention_qkv(self, q, k, v):
|
||||
qkv = torch.stack([q, k, v], dim=2)
|
||||
b = qkv.shape[0]
|
||||
n = qkv.shape[1]
|
||||
qkv = rearrange(qkv, 'b n t (h d) -> (b n) t h d', h=self.heads)
|
||||
out = flash_attention_qkv(qkv, self.scale, b, n)
|
||||
out = rearrange(out, '(b n) h d -> b n (h d)', b=b, h=self.heads)
|
||||
return out
|
||||
|
||||
def _flash_attention_q_kv(self, q, k, v):
|
||||
kv = torch.stack([k, v], dim=2)
|
||||
b = q.shape[0]
|
||||
q_seqlen = q.shape[1]
|
||||
kv_seqlen = kv.shape[1]
|
||||
q = rearrange(q, 'b n (h d) -> (b n) h d', h=self.heads)
|
||||
kv = rearrange(kv, 'b n t (h d) -> (b n) t h d', h=self.heads)
|
||||
out = flash_attention_q_kv(q, kv, self.scale, b, q_seqlen, kv_seqlen)
|
||||
out = rearrange(out, '(b n) h d -> b n (h d)', b=b, h=self.heads)
|
||||
return out
|
||||
|
||||
|
||||
class BasicTransformerBlock(nn.Module):
|
||||
def __init__(self, dim, n_heads, d_head, dropout=0., context_dim=None, gated_ff=True, use_checkpoint=False):
|
||||
super().__init__()
|
||||
self.attn1 = CrossAttention(query_dim=dim, heads=n_heads, dim_head=d_head, dropout=dropout) # is a self-attention
|
||||
self.ff = FeedForward(dim, dropout=dropout, glu=gated_ff)
|
||||
self.attn2 = CrossAttention(query_dim=dim, context_dim=context_dim,
|
||||
heads=n_heads, dim_head=d_head, dropout=dropout) # is self-attn if context is none
|
||||
self.norm1 = nn.LayerNorm(dim)
|
||||
self.norm2 = nn.LayerNorm(dim)
|
||||
self.norm3 = nn.LayerNorm(dim)
|
||||
self.use_checkpoint = use_checkpoint
|
||||
|
||||
def forward(self, x, context=None):
|
||||
|
||||
|
||||
if self.use_checkpoint:
|
||||
return checkpoint(self._forward, x, context)
|
||||
else:
|
||||
return self._forward(x, context)
|
||||
|
||||
def _forward(self, x, context=None):
|
||||
x = self.attn1(self.norm1(x)) + x
|
||||
x = self.attn2(self.norm2(x), context=context) + x
|
||||
x = self.ff(self.norm3(x)) + x
|
||||
return x
|
||||
|
||||
|
||||
|
||||
class SpatialTransformer(nn.Module):
|
||||
"""
|
||||
Transformer block for image-like data.
|
||||
First, project the input (aka embedding)
|
||||
and reshape to b, t, d.
|
||||
Then apply standard transformer action.
|
||||
Finally, reshape to image
|
||||
"""
|
||||
def __init__(self, in_channels, n_heads, d_head,
|
||||
depth=1, dropout=0., context_dim=None, use_checkpoint=False):
|
||||
super().__init__()
|
||||
self.in_channels = in_channels
|
||||
inner_dim = n_heads * d_head
|
||||
self.norm = Normalize(in_channels)
|
||||
|
||||
self.proj_in = nn.Conv2d(in_channels,
|
||||
inner_dim,
|
||||
kernel_size=1,
|
||||
stride=1,
|
||||
padding=0)
|
||||
|
||||
self.transformer_blocks = nn.ModuleList(
|
||||
[BasicTransformerBlock(inner_dim, n_heads, d_head, dropout=dropout, context_dim=context_dim, use_checkpoint=use_checkpoint)
|
||||
for d in range(depth)]
|
||||
)
|
||||
|
||||
self.proj_out = zero_module(nn.Conv2d(inner_dim,
|
||||
in_channels,
|
||||
kernel_size=1,
|
||||
stride=1,
|
||||
padding=0))
|
||||
|
||||
|
||||
def forward(self, x, context=None):
|
||||
# note: if no context is given, cross-attention defaults to self-attention
|
||||
b, c, h, w = x.shape
|
||||
x_in = x
|
||||
x = self.norm(x)
|
||||
x = self.proj_in(x)
|
||||
x = rearrange(x, 'b c h w -> b (h w) c')
|
||||
x = x.contiguous()
|
||||
for block in self.transformer_blocks:
|
||||
x = block(x, context=context)
|
||||
x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w)
|
||||
x = x.contiguous()
|
||||
x = self.proj_out(x)
|
||||
return x + x_in
|
||||
Reference in New Issue
Block a user