2023-12-13 20:52:11 +00:00
|
|
|
import torch
|
|
|
|
from torch import einsum
|
2023-12-13 21:10:03 +00:00
|
|
|
import torch.nn.functional as F
|
|
|
|
import math
|
|
|
|
|
2023-12-13 20:52:11 +00:00
|
|
|
from einops import rearrange, repeat
|
2024-05-14 22:02:27 +00:00
|
|
|
from comfy.ldm.modules.attention import optimized_attention
|
2023-12-13 21:10:03 +00:00
|
|
|
import comfy.samplers
|
2023-12-13 20:52:11 +00:00
|
|
|
|
|
|
|
# from comfy/ldm/modules/attention.py
|
|
|
|
# but modified to return attention scores as well as output
|
2024-05-14 22:02:27 +00:00
|
|
|
def attention_basic_with_sim(q, k, v, heads, mask=None, attn_precision=None):
|
2023-12-13 20:52:11 +00:00
|
|
|
b, _, dim_head = q.shape
|
|
|
|
dim_head //= heads
|
|
|
|
scale = dim_head ** -0.5
|
|
|
|
|
|
|
|
h = heads
|
|
|
|
q, k, v = map(
|
|
|
|
lambda t: t.unsqueeze(3)
|
|
|
|
.reshape(b, -1, heads, dim_head)
|
|
|
|
.permute(0, 2, 1, 3)
|
|
|
|
.reshape(b * heads, -1, dim_head)
|
|
|
|
.contiguous(),
|
|
|
|
(q, k, v),
|
|
|
|
)
|
|
|
|
|
|
|
|
# force cast to fp32 to avoid overflowing
|
2024-05-14 22:02:27 +00:00
|
|
|
if attn_precision == torch.float32:
|
2023-12-15 06:28:16 +00:00
|
|
|
sim = einsum('b i d, b j d -> b i j', q.float(), k.float()) * scale
|
2023-12-13 20:52:11 +00:00
|
|
|
else:
|
|
|
|
sim = einsum('b i d, b j d -> b i j', q, k) * scale
|
|
|
|
|
|
|
|
del q, k
|
|
|
|
|
|
|
|
if mask is not None:
|
|
|
|
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
|
|
|
|
sim = sim.softmax(dim=-1)
|
|
|
|
|
|
|
|
out = einsum('b i j, b j d -> b i d', sim.to(v.dtype), v)
|
|
|
|
out = (
|
|
|
|
out.unsqueeze(0)
|
|
|
|
.reshape(b, heads, -1, dim_head)
|
|
|
|
.permute(0, 2, 1, 3)
|
|
|
|
.reshape(b, -1, heads * dim_head)
|
|
|
|
)
|
|
|
|
return (out, sim)
|
|
|
|
|
2023-12-13 21:10:03 +00:00
|
|
|
def create_blur_map(x0, attn, sigma=3.0, threshold=1.0):
|
|
|
|
# reshape and GAP the attention map
|
|
|
|
_, hw1, hw2 = attn.shape
|
|
|
|
b, _, lh, lw = x0.shape
|
|
|
|
attn = attn.reshape(b, -1, hw1, hw2)
|
|
|
|
# Global Average Pool
|
|
|
|
mask = attn.mean(1, keepdim=False).sum(1, keepdim=False) > threshold
|
2024-11-08 23:16:29 +00:00
|
|
|
|
2024-11-10 05:10:45 +00:00
|
|
|
total = mask.shape[-1]
|
|
|
|
x = round(math.sqrt((lh / lw) * total))
|
|
|
|
xx = None
|
|
|
|
for i in range(0, math.floor(math.sqrt(total) / 2)):
|
|
|
|
for j in [(x + i), max(1, x - i)]:
|
|
|
|
if total % j == 0:
|
|
|
|
xx = j
|
|
|
|
break
|
|
|
|
if xx is not None:
|
|
|
|
break
|
|
|
|
|
|
|
|
x = xx
|
|
|
|
y = total // x
|
2023-12-13 21:10:03 +00:00
|
|
|
|
|
|
|
# Reshape
|
|
|
|
mask = (
|
2024-11-10 05:10:45 +00:00
|
|
|
mask.reshape(b, x, y)
|
2023-12-13 21:10:03 +00:00
|
|
|
.unsqueeze(1)
|
|
|
|
.type(attn.dtype)
|
|
|
|
)
|
|
|
|
# Upsample
|
|
|
|
mask = F.interpolate(mask, (lh, lw))
|
|
|
|
|
|
|
|
blurred = gaussian_blur_2d(x0, kernel_size=9, sigma=sigma)
|
|
|
|
blurred = blurred * mask + x0 * (1 - mask)
|
|
|
|
return blurred
|
|
|
|
|
|
|
|
def gaussian_blur_2d(img, kernel_size, sigma):
|
|
|
|
ksize_half = (kernel_size - 1) * 0.5
|
|
|
|
|
|
|
|
x = torch.linspace(-ksize_half, ksize_half, steps=kernel_size)
|
|
|
|
|
|
|
|
pdf = torch.exp(-0.5 * (x / sigma).pow(2))
|
|
|
|
|
|
|
|
x_kernel = pdf / pdf.sum()
|
|
|
|
x_kernel = x_kernel.to(device=img.device, dtype=img.dtype)
|
|
|
|
|
|
|
|
kernel2d = torch.mm(x_kernel[:, None], x_kernel[None, :])
|
|
|
|
kernel2d = kernel2d.expand(img.shape[-3], 1, kernel2d.shape[0], kernel2d.shape[1])
|
|
|
|
|
|
|
|
padding = [kernel_size // 2, kernel_size // 2, kernel_size // 2, kernel_size // 2]
|
|
|
|
|
|
|
|
img = F.pad(img, padding, mode="reflect")
|
|
|
|
img = F.conv2d(img, kernel2d, groups=img.shape[-3])
|
|
|
|
return img
|
|
|
|
|
|
|
|
class SelfAttentionGuidance:
|
2023-12-13 20:52:11 +00:00
|
|
|
@classmethod
|
|
|
|
def INPUT_TYPES(s):
|
|
|
|
return {"required": { "model": ("MODEL",),
|
2024-08-02 14:29:03 +00:00
|
|
|
"scale": ("FLOAT", {"default": 0.5, "min": -2.0, "max": 5.0, "step": 0.01}),
|
2023-12-13 20:52:11 +00:00
|
|
|
"blur_sigma": ("FLOAT", {"default": 2.0, "min": 0.0, "max": 10.0, "step": 0.1}),
|
|
|
|
}}
|
|
|
|
RETURN_TYPES = ("MODEL",)
|
|
|
|
FUNCTION = "patch"
|
|
|
|
|
|
|
|
CATEGORY = "_for_testing"
|
|
|
|
|
|
|
|
def patch(self, model, scale, blur_sigma):
|
|
|
|
m = model.clone()
|
2023-12-13 21:10:03 +00:00
|
|
|
|
2023-12-13 20:52:11 +00:00
|
|
|
attn_scores = None
|
|
|
|
|
|
|
|
# TODO: make this work properly with chunked batches
|
|
|
|
# currently, we can only save the attn from one UNet call
|
|
|
|
def attn_and_record(q, k, v, extra_options):
|
|
|
|
nonlocal attn_scores
|
|
|
|
# if uncond, save the attention scores
|
|
|
|
heads = extra_options["n_heads"]
|
|
|
|
cond_or_uncond = extra_options["cond_or_uncond"]
|
|
|
|
b = q.shape[0] // len(cond_or_uncond)
|
|
|
|
if 1 in cond_or_uncond:
|
|
|
|
uncond_index = cond_or_uncond.index(1)
|
|
|
|
# do the entire attention operation, but save the attention scores to attn_scores
|
2024-05-14 22:02:27 +00:00
|
|
|
(out, sim) = attention_basic_with_sim(q, k, v, heads=heads, attn_precision=extra_options["attn_precision"])
|
2023-12-13 20:52:11 +00:00
|
|
|
# when using a higher batch size, I BELIEVE the result batch dimension is [uc1, ... ucn, c1, ... cn]
|
|
|
|
n_slices = heads * b
|
|
|
|
attn_scores = sim[n_slices * uncond_index:n_slices * (uncond_index+1)]
|
|
|
|
return out
|
|
|
|
else:
|
2024-05-14 22:02:27 +00:00
|
|
|
return optimized_attention(q, k, v, heads=heads, attn_precision=extra_options["attn_precision"])
|
2023-12-13 20:52:11 +00:00
|
|
|
|
2023-12-13 21:10:03 +00:00
|
|
|
def post_cfg_function(args):
|
|
|
|
nonlocal attn_scores
|
|
|
|
uncond_attn = attn_scores
|
|
|
|
|
|
|
|
sag_scale = scale
|
|
|
|
sag_sigma = blur_sigma
|
|
|
|
sag_threshold = 1.0
|
|
|
|
model = args["model"]
|
|
|
|
uncond_pred = args["uncond_denoised"]
|
|
|
|
uncond = args["uncond"]
|
|
|
|
cfg_result = args["denoised"]
|
|
|
|
sigma = args["sigma"]
|
|
|
|
model_options = args["model_options"]
|
|
|
|
x = args["input"]
|
2024-01-10 09:08:43 +00:00
|
|
|
if min(cfg_result.shape[2:]) <= 4: #skip when too small to add padding
|
|
|
|
return cfg_result
|
2023-12-13 21:10:03 +00:00
|
|
|
|
|
|
|
# create the adversarially blurred image
|
|
|
|
degraded = create_blur_map(uncond_pred, uncond_attn, sag_sigma, sag_threshold)
|
|
|
|
degraded_noised = degraded + x - uncond_pred
|
|
|
|
# call into the UNet
|
2024-04-01 21:23:07 +00:00
|
|
|
(sag,) = comfy.samplers.calc_cond_batch(model, [uncond], degraded_noised, sigma, model_options)
|
2023-12-13 21:10:03 +00:00
|
|
|
return cfg_result + (degraded - sag) * sag_scale
|
|
|
|
|
2023-12-18 22:03:32 +00:00
|
|
|
m.set_model_sampler_post_cfg_function(post_cfg_function, disable_cfg1_optimization=True)
|
2023-12-13 21:10:03 +00:00
|
|
|
|
2023-12-13 20:52:11 +00:00
|
|
|
# from diffusers:
|
|
|
|
# unet.mid_block.attentions[0].transformer_blocks[0].attn1.patch
|
2023-12-13 21:10:03 +00:00
|
|
|
m.set_model_attn1_replace(attn_and_record, "middle", 0, 0)
|
|
|
|
|
2023-12-13 20:52:11 +00:00
|
|
|
return (m, )
|
|
|
|
|
|
|
|
NODE_CLASS_MAPPINGS = {
|
2023-12-13 21:10:03 +00:00
|
|
|
"SelfAttentionGuidance": SelfAttentionGuidance,
|
|
|
|
}
|
|
|
|
|
|
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
|
|
"SelfAttentionGuidance": "Self-Attention Guidance",
|
2023-12-13 20:52:11 +00:00
|
|
|
}
|