Merge branch 'comfyanonymous:master' into master

This commit is contained in:
patientx 2024-08-30 12:14:01 +03:00 committed by GitHub
commit a8652a052f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
8 changed files with 25 additions and 9 deletions

View File

@ -14,7 +14,7 @@ run_cpu.bat
IF YOU GET A RED ERROR IN THE UI MAKE SURE YOU HAVE A MODEL/CHECKPOINT IN: ComfyUI\models\checkpoints
You can download the stable diffusion 1.5 one from: https://huggingface.co/runwayml/stable-diffusion-v1-5/blob/main/v1-5-pruned-emaonly.ckpt
You can download the stable diffusion 1.5 one from: https://huggingface.co/Comfy-Org/stable-diffusion-v1-5-archive/blob/main/v1-5-pruned-emaonly-fp16.safetensors
RECOMMENDED WAY TO UPDATE:

View File

@ -444,7 +444,12 @@ def load_controlnet_flux_instantx(sd):
for k in sd:
new_sd[k] = sd[k]
control_model = comfy.ldm.flux.controlnet.ControlNetFlux(latent_input=True, operations=operations, device=offload_device, dtype=unet_dtype, **model_config.unet_config)
num_union_modes = 0
union_cnet = "controlnet_mode_embedder.weight"
if union_cnet in new_sd:
num_union_modes = new_sd[union_cnet].shape[0]
control_model = comfy.ldm.flux.controlnet.ControlNetFlux(latent_input=True, num_union_modes=num_union_modes, operations=operations, device=offload_device, dtype=unet_dtype, **model_config.unet_config)
control_model = controlnet_load_state_dict(control_model, new_sd)
latent_format = comfy.latent_formats.Flux()

View File

@ -14,7 +14,7 @@ import comfy.ldm.common_dit
class ControlNetFlux(Flux):
def __init__(self, latent_input=False, image_model=None, dtype=None, device=None, operations=None, **kwargs):
def __init__(self, latent_input=False, num_union_modes=0, image_model=None, dtype=None, device=None, operations=None, **kwargs):
super().__init__(final_layer=False, dtype=dtype, device=device, operations=operations, **kwargs)
self.main_model_double = 19
@ -29,6 +29,11 @@ class ControlNetFlux(Flux):
for _ in range(self.params.depth_single_blocks):
self.controlnet_single_blocks.append(operations.Linear(self.hidden_size, self.hidden_size, dtype=dtype, device=device))
self.num_union_modes = num_union_modes
self.controlnet_mode_embedder = None
if self.num_union_modes > 0:
self.controlnet_mode_embedder = operations.Embedding(self.num_union_modes, self.hidden_size, dtype=dtype, device=device)
self.gradient_checkpointing = False
self.latent_input = latent_input
self.pos_embed_input = operations.Linear(self.in_channels, self.hidden_size, bias=True, dtype=dtype, device=device)
@ -61,6 +66,7 @@ class ControlNetFlux(Flux):
timesteps: Tensor,
y: Tensor,
guidance: Tensor = None,
control_type: Tensor = None,
) -> Tensor:
if img.ndim != 3 or txt.ndim != 3:
raise ValueError("Input img and txt tensors must have 3 dimensions.")
@ -79,6 +85,11 @@ class ControlNetFlux(Flux):
vec = vec + self.vector_in(y)
txt = self.txt_in(txt)
if self.controlnet_mode_embedder is not None and len(control_type) > 0:
control_cond = self.controlnet_mode_embedder(torch.tensor(control_type, device=img.device), out_dtype=img.dtype).unsqueeze(0).repeat((txt.shape[0], 1, 1))
txt = torch.cat([control_cond, txt], dim=1)
txt_ids = torch.cat([txt_ids[:,:1], txt_ids], dim=1)
ids = torch.cat((txt_ids, img_ids), dim=1)
pe = self.pe_embedder(ids)
@ -137,4 +148,4 @@ class ControlNetFlux(Flux):
img_ids = repeat(img_ids, "h w c -> b (h w) c", b=bs)
txt_ids = torch.zeros((bs, context.shape[1], 3), device=x.device, dtype=x.dtype)
return self.forward_orig(img, img_ids, hint, context, txt_ids, timesteps, y, guidance)
return self.forward_orig(img, img_ids, hint, context, txt_ids, timesteps, y, guidance, control_type=kwargs.get("control_type", []))

View File

@ -540,7 +540,7 @@ def calculate_weight(patches, weight, key, intermediate_dtype=torch.float32):
b2 = comfy.model_management.cast_to_device(v[3].flatten(start_dim=1), weight.device, intermediate_dtype)
try:
lora_diff = (torch.mm(b2, b1) + torch.mm(torch.mm(weight.flatten(start_dim=1), a2), a1)).reshape(weight.shape)
lora_diff = (torch.mm(b2, b1) + torch.mm(torch.mm(weight.flatten(start_dim=1).to(dtype=intermediate_dtype), a2), a1)).reshape(weight.shape)
if dora_scale is not None:
weight = function(weight_decompose(dora_scale, weight, lora_diff, alpha, strength, intermediate_dtype))
else:

View File

@ -79,7 +79,7 @@
"#!wget -c https://huggingface.co/comfyanonymous/clip_vision_g/resolve/main/clip_vision_g.safetensors -P ./models/clip_vision/\n",
"\n",
"# SD1.5\n",
"!wget -c https://huggingface.co/runwayml/stable-diffusion-v1-5/resolve/main/v1-5-pruned-emaonly.ckpt -P ./models/checkpoints/\n",
"!wget -c https://huggingface.co/Comfy-Org/stable-diffusion-v1-5-archive/resolve/main/v1-5-pruned-emaonly-fp16.safetensors -P ./models/checkpoints/\n",
"\n",
"# SD2\n",
"#!wget -c https://huggingface.co/stabilityai/stable-diffusion-2-1-base/resolve/main/v2-1_512-ema-pruned.safetensors -P ./models/checkpoints/\n",

View File

@ -43,7 +43,7 @@ prompt_text = """
"4": {
"class_type": "CheckpointLoaderSimple",
"inputs": {
"ckpt_name": "v1-5-pruned-emaonly.ckpt"
"ckpt_name": "v1-5-pruned-emaonly.safetensors"
}
},
"5": {

View File

@ -84,7 +84,7 @@ prompt_text = """
"4": {
"class_type": "CheckpointLoaderSimple",
"inputs": {
"ckpt_name": "v1-5-pruned-emaonly.ckpt"
"ckpt_name": "v1-5-pruned-emaonly.safetensors"
}
},
"5": {

View File

@ -81,7 +81,7 @@ prompt_text = """
"4": {
"class_type": "CheckpointLoaderSimple",
"inputs": {
"ckpt_name": "v1-5-pruned-emaonly.ckpt"
"ckpt_name": "v1-5-pruned-emaonly.safetensors"
}
},
"5": {