Reformat code and reshape non-square noises so that the model can run
This commit is contained in:
asagi4
+85
-79
@@ -7,26 +7,29 @@ import einops
|
||||
from diffusers.models.normalization import AdaGroupNorm
|
||||
|
||||
from timm.layers import use_fused_attn
|
||||
from timm.models.layers import PatchEmbed, Mlp, DropPath, trunc_normal_, lecun_normal_, get_act_layer
|
||||
|
||||
|
||||
from comfy.utils import common_upscale
|
||||
|
||||
|
||||
class Attention(nn.Module):
|
||||
fused_attn = True
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
dim: int,
|
||||
num_heads: int = 8,
|
||||
qkv_bias: bool = False,
|
||||
qk_norm: bool = False,
|
||||
attn_drop: float = 0.,
|
||||
proj_drop: float = 0.,
|
||||
norm_layer: nn.Module = nn.LayerNorm,
|
||||
self,
|
||||
dim: int,
|
||||
num_heads: int = 8,
|
||||
qkv_bias: bool = False,
|
||||
qk_norm: bool = False,
|
||||
attn_drop: float = 0.0,
|
||||
proj_drop: float = 0.0,
|
||||
norm_layer: nn.Module = nn.LayerNorm,
|
||||
) -> None:
|
||||
super().__init__()
|
||||
assert dim % num_heads == 0, 'dim should be divisible by num_heads'
|
||||
assert dim % num_heads == 0, "dim should be divisible by num_heads"
|
||||
self.num_heads = num_heads
|
||||
self.head_dim = dim // num_heads
|
||||
self.scale = self.head_dim ** -0.5
|
||||
self.scale = self.head_dim**-0.5
|
||||
self.fused_attn = use_fused_attn()
|
||||
|
||||
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
|
||||
@@ -44,8 +47,10 @@ class Attention(nn.Module):
|
||||
|
||||
if self.fused_attn:
|
||||
x = F.scaled_dot_product_attention(
|
||||
q, k, v,
|
||||
dropout_p=self.attn_drop.p if self.training else 0.,
|
||||
q,
|
||||
k,
|
||||
v,
|
||||
dropout_p=self.attn_drop.p if self.training else 0.0,
|
||||
)
|
||||
else:
|
||||
q = q * self.scale
|
||||
@@ -61,7 +66,7 @@ class Attention(nn.Module):
|
||||
|
||||
|
||||
class SVDNoiseUnet(nn.Module):
|
||||
def __init__(self, in_channels=4, out_channels=4, resolution=128): # resolution = size // 8
|
||||
def __init__(self, in_channels=4, out_channels=4, resolution=128): # resolution = size // 8
|
||||
super(SVDNoiseUnet, self).__init__()
|
||||
|
||||
_in = int(resolution * in_channels // 2)
|
||||
@@ -85,7 +90,7 @@ class SVDNoiseUnet(nn.Module):
|
||||
|
||||
self.bn = nn.BatchNorm2d(_out)
|
||||
|
||||
self.mlp4 = nn.Sequential(
|
||||
self.mlp4 = nn.Sequential(
|
||||
nn.Linear(_out, 1024),
|
||||
nn.ReLU(inplace=True),
|
||||
nn.Linear(1024, _out),
|
||||
@@ -93,15 +98,15 @@ class SVDNoiseUnet(nn.Module):
|
||||
|
||||
def forward(self, x, residual=False):
|
||||
b, c, h, w = x.shape
|
||||
x = einops.rearrange(x, "b (a c)h w ->b (a h)(c w)", a=2,c=2) # x -> [1, 256, 256]
|
||||
U, s, V = torch.linalg.svd(x) # U->[b 256 256], s-> [b 256], V->[b 256 256]
|
||||
x = einops.rearrange(x, "b (a c)h w ->b (a h)(c w)", a=2, c=2) # x -> [1, 256, 256]
|
||||
U, s, V = torch.linalg.svd(x) # U->[b 256 256], s-> [b 256], V->[b 256 256]
|
||||
U_T = U.permute(0, 2, 1)
|
||||
out = self.mlp1(U_T) + self.mlp2(V) + self.mlp3(s).unsqueeze(1) # s -> [b, 1, 256] => [b, 256, 256]
|
||||
out = self.mlp1(U_T) + self.mlp2(V) + self.mlp3(s).unsqueeze(1) # s -> [b, 1, 256] => [b, 256, 256]
|
||||
out = self.attention(out).mean(1)
|
||||
out = self.mlp4(out) + s
|
||||
pred = U @ torch.diag_embed(out) @ V
|
||||
return einops.rearrange(pred, "b (a h)(c w) -> b (a c) h w", a=2,c=2)
|
||||
|
||||
return einops.rearrange(pred, "b (a h)(c w) -> b (a c) h w", a=2, c=2)
|
||||
|
||||
|
||||
class SVDNoiseUnet_Concise(nn.Module):
|
||||
def __init__(self, in_channels=4, out_channels=4, resolution=128):
|
||||
@@ -111,13 +116,12 @@ class SVDNoiseUnet_Concise(nn.Module):
|
||||
class NoiseTransformer(nn.Module):
|
||||
def __init__(self, resolution=128):
|
||||
super().__init__()
|
||||
self.upsample = lambda x: F.interpolate(x, [224,224])
|
||||
self.downsample = lambda x: F.interpolate(x, [resolution,resolution])
|
||||
self.upconv = nn.Conv2d(7,4,(1,1),(1,1),(0,0))
|
||||
self.downconv = nn.Conv2d(4,3,(1,1),(1,1),(0,0))
|
||||
self.upsample = lambda x: F.interpolate(x, [224, 224])
|
||||
self.downsample = lambda x: F.interpolate(x, [resolution, resolution])
|
||||
self.upconv = nn.Conv2d(7, 4, (1, 1), (1, 1), (0, 0))
|
||||
self.downconv = nn.Conv2d(4, 3, (1, 1), (1, 1), (0, 0))
|
||||
# self.upconv = nn.Conv2d(7,4,(1,1),(1,1),(0,0))
|
||||
self.swin = create_model("swin_tiny_patch4_window7_224",pretrained=True)
|
||||
|
||||
self.swin = create_model("swin_tiny_patch4_window7_224", pretrained=True)
|
||||
|
||||
def forward(self, x, residual=False):
|
||||
if residual:
|
||||
@@ -127,78 +131,76 @@ class NoiseTransformer(nn.Module):
|
||||
|
||||
return x
|
||||
|
||||
|
||||
class NPNet(nn.Module):
|
||||
def __init__(self, model_id, pretrained_path=True, device='cuda') -> None:
|
||||
super(NPNet, self).__init__()
|
||||
def __init__(self, model_id, pretrained_path=True, device="cuda") -> None:
|
||||
super(NPNet, self).__init__()
|
||||
|
||||
assert model_id in ['SDXL', 'DreamShaper', 'DiT']
|
||||
assert model_id in ["SDXL", "DreamShaper", "DiT"]
|
||||
|
||||
self.model_id = model_id
|
||||
self.device = device
|
||||
self.pretrained_path = pretrained_path
|
||||
self.model_id = model_id
|
||||
self.device = device
|
||||
self.pretrained_path = pretrained_path
|
||||
|
||||
(
|
||||
self.unet_svd,
|
||||
self.unet_embedding,
|
||||
self.text_embedding,
|
||||
self._alpha,
|
||||
self._beta
|
||||
) = self.get_model()
|
||||
(self.unet_svd, self.unet_embedding, self.text_embedding, self._alpha, self._beta) = self.get_model()
|
||||
|
||||
def get_model(self):
|
||||
def get_model(self):
|
||||
|
||||
unet_embedding = NoiseTransformer(resolution=128).to(self.device).to(torch.float32)
|
||||
unet_svd = SVDNoiseUnet(resolution=128).to(self.device).to(torch.float32)
|
||||
unet_embedding = NoiseTransformer(resolution=128).to(self.device).to(torch.float32)
|
||||
unet_svd = SVDNoiseUnet(resolution=128).to(self.device).to(torch.float32)
|
||||
|
||||
if self.model_id == 'DiT':
|
||||
text_embedding = AdaGroupNorm(1024 * 77, 4, 1, eps=1e-6).to(self.device).to(torch.float32)
|
||||
else:
|
||||
text_embedding = AdaGroupNorm(2048 * 77, 4, 1, eps=1e-6).to(self.device).to(torch.float32)
|
||||
if self.model_id == "DiT":
|
||||
text_embedding = AdaGroupNorm(1024 * 77, 4, 1, eps=1e-6).to(self.device).to(torch.float32)
|
||||
else:
|
||||
text_embedding = AdaGroupNorm(2048 * 77, 4, 1, eps=1e-6).to(self.device).to(torch.float32)
|
||||
|
||||
|
||||
if '.pth' in self.pretrained_path:
|
||||
gloden_unet = torch.load(self.pretrained_path)
|
||||
unet_svd.load_state_dict(gloden_unet["unet_svd"])
|
||||
unet_embedding.load_state_dict(gloden_unet["unet_embedding"])
|
||||
text_embedding.load_state_dict(gloden_unet["embeeding"])
|
||||
_alpha = gloden_unet["alpha"]
|
||||
_beta = gloden_unet["beta"]
|
||||
if ".pth" in self.pretrained_path:
|
||||
gloden_unet = torch.load(self.pretrained_path)
|
||||
unet_svd.load_state_dict(gloden_unet["unet_svd"])
|
||||
unet_embedding.load_state_dict(gloden_unet["unet_embedding"])
|
||||
text_embedding.load_state_dict(gloden_unet["embeeding"])
|
||||
_alpha = gloden_unet["alpha"]
|
||||
_beta = gloden_unet["beta"]
|
||||
|
||||
print("Load Successfully!")
|
||||
print("Load Successfully!")
|
||||
|
||||
return unet_svd, unet_embedding, text_embedding, _alpha, _beta
|
||||
|
||||
else:
|
||||
assert ("No Pretrained Weights Found!")
|
||||
|
||||
return unet_svd, unet_embedding, text_embedding, _alpha, _beta
|
||||
|
||||
def forward(self, initial_noise, prompt_embeds):
|
||||
else:
|
||||
assert "No Pretrained Weights Found!"
|
||||
|
||||
prompt_embeds = prompt_embeds.float().view(prompt_embeds.shape[0], -1)
|
||||
text_emb = self.text_embedding(initial_noise.float(), prompt_embeds)
|
||||
def forward(self, initial_noise, prompt_embeds):
|
||||
|
||||
encoder_hidden_states_svd = initial_noise
|
||||
encoder_hidden_states_embedding = initial_noise + text_emb
|
||||
prompt_embeds = prompt_embeds.float().view(prompt_embeds.shape[0], -1)
|
||||
text_emb = self.text_embedding(initial_noise.float(), prompt_embeds)
|
||||
|
||||
golden_embedding = self.unet_embedding(encoder_hidden_states_embedding.float())
|
||||
encoder_hidden_states_svd = initial_noise
|
||||
encoder_hidden_states_embedding = initial_noise + text_emb
|
||||
|
||||
golden_noise = self.unet_svd(encoder_hidden_states_svd.float()) + (
|
||||
2 * torch.sigmoid(self._alpha) - 1) * text_emb + self._beta * golden_embedding
|
||||
golden_embedding = self.unet_embedding(encoder_hidden_states_embedding.float())
|
||||
|
||||
golden_noise = (
|
||||
self.unet_svd(encoder_hidden_states_svd.float())
|
||||
+ (2 * torch.sigmoid(self._alpha) - 1) * text_emb
|
||||
+ self._beta * golden_embedding
|
||||
)
|
||||
|
||||
return golden_noise
|
||||
|
||||
return golden_noise
|
||||
|
||||
class NPNetGoldenNoise:
|
||||
npnet = None
|
||||
noise = None
|
||||
cond = None
|
||||
seed = None
|
||||
|
||||
@classmethod
|
||||
def INPUT_TYPES(s):
|
||||
return {
|
||||
"required": {
|
||||
"noise": ("NOISE",),
|
||||
"prompt": ("CONDITIONING",),
|
||||
"model_path": ("STRING",{"default": "/path/to/sdxl.pth"}),
|
||||
"model_path": ("STRING", {"default": "/path/to/sdxl.pth"}),
|
||||
"model_type": (["SDXL", "DreamShaper", "DiT"],),
|
||||
}
|
||||
}
|
||||
@@ -210,18 +212,24 @@ class NPNetGoldenNoise:
|
||||
|
||||
def generate_noise(self, input_latent):
|
||||
self.seed = self.noise.seed
|
||||
init_noise = self.noise.generate_noise(input_latent).to('cuda')
|
||||
cond = self.cond[0].clone().to('cuda')
|
||||
self.npnet.to('cuda')
|
||||
orig_shape = input_latent["samples"].shape
|
||||
if orig_shape[-2] != 128 or orig_shape[-1] != 128:
|
||||
input_latent = input_latent.copy()
|
||||
print("Latent must be 128x128 for the NPNet model to work; generating square noise and reshaping...")
|
||||
input_latent["samples"] = common_upscale(input_latent["samples"], 128, 128, "nearest-exact", "disabled")
|
||||
init_noise = self.noise.generate_noise(input_latent).to("cuda")
|
||||
cond = self.cond[0].clone().to("cuda")
|
||||
self.npnet.to("cuda")
|
||||
try:
|
||||
print("Applying NPNet to noise")
|
||||
r = self.npnet(init_noise, cond)
|
||||
r = self.npnet(init_noise, cond).to("cpu")
|
||||
if orig_shape[-2] != 128 or orig_shape[-1] != 128:
|
||||
r = common_upscale(r, orig_shape[-1], orig_shape[-2], "nearest-exact", "disabled")
|
||||
except Exception as e:
|
||||
print("Running NPNet failed with error (non-square latent can cause shape errors):", e)
|
||||
print("Returning unmodified noise")
|
||||
print("Running NPNet failed with error, returning unmodified noise:", e)
|
||||
return init_noise
|
||||
print("NPNet ran ok")
|
||||
return r.to("cpu")
|
||||
return r
|
||||
|
||||
def doit(self, noise, prompt, model_path, model_type):
|
||||
if self.npnet is None:
|
||||
@@ -233,6 +241,4 @@ class NPNetGoldenNoise:
|
||||
return (self,)
|
||||
|
||||
|
||||
|
||||
|
||||
NODE_CLASS_MAPPINGS = {"NPNetGoldenNoise": NPNetGoldenNoise}
|
||||
|
||||
Reference in New Issue
Block a user