diff --git a/README.md b/README.md
index ba1e844b3..2636ce140 100644
--- a/README.md
+++ b/README.md
@@ -41,29 +41,32 @@ Workflow examples can be found on the [Examples page](https://comfyanonymous.git
## Shortcuts
-| Keybind | Explanation |
-|---------------------------|--------------------------------------------------------------------------------------------------------------------|
-| Ctrl + Enter | Queue up current graph for generation |
-| Ctrl + Shift + Enter | Queue up current graph as first for generation |
-| Ctrl + Z/Ctrl + Y | Undo/Redo |
-| Ctrl + S | Save workflow |
-| Ctrl + O | Load workflow |
-| Ctrl + A | Select all nodes |
-| Alt + C | Collapse/uncollapse selected nodes |
-| Ctrl + M | Mute/unmute selected nodes |
-| Ctrl + B | Bypass selected nodes (acts like the node was removed from the graph and the wires reconnected through) |
-| Delete/Backspace | Delete selected nodes |
-| Ctrl + Delete/Backspace | Delete the current graph |
-| Space | Move the canvas around when held and moving the cursor |
-| Ctrl/Shift + Click | Add clicked node to selection |
-| Ctrl + C/Ctrl + V | Copy and paste selected nodes (without maintaining connections to outputs of unselected nodes) |
-| Ctrl + C/Ctrl + Shift + V | Copy and paste selected nodes (maintaining connections from outputs of unselected nodes to inputs of pasted nodes) |
-| Shift + Drag | Move multiple selected nodes at the same time |
-| Ctrl + D | Load default graph |
-| Q | Toggle visibility of the queue |
-| H | Toggle visibility of history |
-| R | Refresh graph |
-| Double-Click LMB | Open node quick search palette |
+| Keybind | Explanation |
+|------------------------------------|--------------------------------------------------------------------------------------------------------------------|
+| Ctrl + Enter | Queue up current graph for generation |
+| Ctrl + Shift + Enter | Queue up current graph as first for generation |
+| Ctrl + Z/Ctrl + Y | Undo/Redo |
+| Ctrl + S | Save workflow |
+| Ctrl + O | Load workflow |
+| Ctrl + A | Select all nodes |
+| Alt + C | Collapse/uncollapse selected nodes |
+| Ctrl + M | Mute/unmute selected nodes |
+| Ctrl + B | Bypass selected nodes (acts like the node was removed from the graph and the wires reconnected through) |
+| Delete/Backspace | Delete selected nodes |
+| Ctrl + Delete/Backspace | Delete the current graph |
+| Space | Move the canvas around when held and moving the cursor |
+| Ctrl/Shift + Click | Add clicked node to selection |
+| Ctrl + C/Ctrl + V | Copy and paste selected nodes (without maintaining connections to outputs of unselected nodes) |
+| Ctrl + C/Ctrl + Shift + V | Copy and paste selected nodes (maintaining connections from outputs of unselected nodes to inputs of pasted nodes) |
+| Shift + Drag | Move multiple selected nodes at the same time |
+| Ctrl + D | Load default graph |
+| Alt + `+` | Canvas Zoom in |
+| Alt + `-` | Canvas Zoom out |
+| Ctrl + Shift + LMB + Vertical drag | Canvas Zoom in/out |
+| Q | Toggle visibility of the queue |
+| H | Toggle visibility of history |
+| R | Refresh graph |
+| Double-Click LMB | Open node quick search palette |
Ctrl can also be replaced with Cmd instead for macOS users
@@ -99,7 +102,7 @@ Put your VAE in: models/vae
### AMD GPUs (Linux only)
AMD users can install rocm and pytorch with pip if you don't have it already installed, this is the command to install the stable version:
-```pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.7```
+```pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.0```
This is the command to install the nightly with ROCm 6.0 which might have some performance improvements:
@@ -197,9 +200,7 @@ To use a textual inversion concepts/embeddings in a text prompt put them in the
## How to increase generation speed?
-Make sure you use the regular loaders/Load Checkpoint node to load checkpoints. It will auto pick the right settings depending on your GPU.
-
-You can set this command line setting to disable the upcasting to fp32 in some cross attention operations which will increase your speed. Note that this will very likely give you black images on SD2.x models. If you use xformers or pytorch attention this option does not do anything.
+On non Nvidia hardware you can set this command line setting to disable the upcasting to fp32 in some cross attention operations which will increase your speed. Note that this will very likely give you black images on SD2.x models. If you use xformers or pytorch attention this option does not do anything.
```--dont-upcast-attention```
@@ -209,6 +210,14 @@ Use ```--preview-method auto``` to enable previews.
The default installation includes a fast latent preview method that's low-resolution. To enable higher-quality previews with [TAESD](https://github.com/madebyollin/taesd), download the [taesd_decoder.pth](https://github.com/madebyollin/taesd/raw/main/taesd_decoder.pth) (for SD1.x and SD2.x) and [taesdxl_decoder.pth](https://github.com/madebyollin/taesd/raw/main/taesdxl_decoder.pth) (for SDXL) models and place them in the `models/vae_approx` folder. Once they're installed, restart ComfyUI to enable high-quality previews.
+## How to use TLS/SSL?
+Generate a self-signed certificate (not appropriate for shared/production use) and key by running the command: `openssl req -x509 -newkey rsa:4096 -keyout key.pem -out cert.pem -sha256 -days 3650 -nodes -subj "/C=XX/ST=StateName/L=CityName/O=CompanyName/OU=CompanySectionName/CN=CommonNameOrHostname"`
+
+Use `--tls-keyfile key.pem --tls-certfile cert.pem` to enable TLS/SSL, the app will now be accessible with `https://...` instead of `http://...`.
+
+> Note: Windows users can use [alexisrolland/docker-openssl](https://github.com/alexisrolland/docker-openssl) or one of the [3rd party binary distributions](https://wiki.openssl.org/index.php/Binaries) to run the command example above.
+
If you use a container, note that the volume mount `-v` can be a relative path so `... -v ".\:/openssl-certs" ...` would create the key & cert files in the current directory of your command prompt or powershell terminal.
+
## Support and dev channel
[Matrix space: #comfyui_space:matrix.org](https://app.element.io/#/room/%23comfyui_space%3Amatrix.org) (it's like discord but open source).
diff --git a/comfy/cli_args.py b/comfy/cli_args.py
index 353bb51e7..569c79380 100644
--- a/comfy/cli_args.py
+++ b/comfy/cli_args.py
@@ -35,6 +35,8 @@ parser = argparse.ArgumentParser()
parser.add_argument("--listen", type=str, default="127.0.0.1", metavar="IP", nargs="?", const="0.0.0.0", help="Specify the IP address to listen on (default: 127.0.0.1). If --listen is provided without an argument, it defaults to 0.0.0.0. (listens on all)")
parser.add_argument("--port", type=int, default=8188, help="Set the listen port.")
+parser.add_argument("--tls-keyfile", type=str, help="Path to TLS (SSL) key file. Enables TLS, makes app accessible at https://... requires --tls-certfile to function")
+parser.add_argument("--tls-certfile", type=str, help="Path to TLS (SSL) certificate file. Enables TLS, makes app accessible at https://... requires --tls-keyfile to function")
parser.add_argument("--enable-cors-header", type=str, default=None, metavar="ORIGIN", nargs="?", const="*", help="Enable CORS (Cross-Origin Resource Sharing) with optional origin or allow all with default '*'.")
parser.add_argument("--max-upload-size", type=float, default=100, help="Set the maximum upload size in MB.")
diff --git a/comfy/controlnet.py b/comfy/controlnet.py
index b6941d8c4..8cf4a61a6 100644
--- a/comfy/controlnet.py
+++ b/comfy/controlnet.py
@@ -138,11 +138,13 @@ class ControlBase:
return out
class ControlNet(ControlBase):
- def __init__(self, control_model, global_average_pooling=False, device=None, load_device=None, manual_cast_dtype=None):
+ def __init__(self, control_model=None, global_average_pooling=False, device=None, load_device=None, manual_cast_dtype=None):
super().__init__(device)
self.control_model = control_model
self.load_device = load_device
- self.control_model_wrapped = comfy.model_patcher.ModelPatcher(self.control_model, load_device=load_device, offload_device=comfy.model_management.unet_offload_device())
+ if control_model is not None:
+ self.control_model_wrapped = comfy.model_patcher.ModelPatcher(self.control_model, load_device=load_device, offload_device=comfy.model_management.unet_offload_device())
+
self.global_average_pooling = global_average_pooling
self.model_sampling_current = None
self.manual_cast_dtype = manual_cast_dtype
@@ -183,7 +185,9 @@ class ControlNet(ControlBase):
return self.control_merge(None, control, control_prev, output_dtype)
def copy(self):
- c = ControlNet(self.control_model, global_average_pooling=self.global_average_pooling, load_device=self.load_device, manual_cast_dtype=self.manual_cast_dtype)
+ c = ControlNet(None, global_average_pooling=self.global_average_pooling, load_device=self.load_device, manual_cast_dtype=self.manual_cast_dtype)
+ c.control_model = self.control_model
+ c.control_model_wrapped = self.control_model_wrapped
self.copy_to(c)
return c
diff --git a/comfy/diffusers_convert.py b/comfy/diffusers_convert.py
index 08018c54d..ed2a45fea 100644
--- a/comfy/diffusers_convert.py
+++ b/comfy/diffusers_convert.py
@@ -206,6 +206,21 @@ textenc_pattern = re.compile("|".join(protected.keys()))
# Ordering is from https://github.com/pytorch/pytorch/blob/master/test/cpp/api/modules.cpp
code2idx = {"q": 0, "k": 1, "v": 2}
+# This function exists because at the time of writing torch.cat can't do fp8 with cuda
+def cat_tensors(tensors):
+ x = 0
+ for t in tensors:
+ x += t.shape[0]
+
+ shape = [x] + list(tensors[0].shape)[1:]
+ out = torch.empty(shape, device=tensors[0].device, dtype=tensors[0].dtype)
+
+ x = 0
+ for t in tensors:
+ out[x:x + t.shape[0]] = t
+ x += t.shape[0]
+
+ return out
def convert_text_enc_state_dict_v20(text_enc_dict, prefix=""):
new_state_dict = {}
@@ -249,13 +264,13 @@ def convert_text_enc_state_dict_v20(text_enc_dict, prefix=""):
if None in tensors:
raise Exception("CORRUPTED MODEL: one of the q-k-v values for the text encoder was missing")
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k_pre)
- new_state_dict[relabelled_key + ".in_proj_weight"] = torch.cat(tensors)
+ new_state_dict[relabelled_key + ".in_proj_weight"] = cat_tensors(tensors)
for k_pre, tensors in capture_qkv_bias.items():
if None in tensors:
raise Exception("CORRUPTED MODEL: one of the q-k-v values for the text encoder was missing")
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k_pre)
- new_state_dict[relabelled_key + ".in_proj_bias"] = torch.cat(tensors)
+ new_state_dict[relabelled_key + ".in_proj_bias"] = cat_tensors(tensors)
return new_state_dict
diff --git a/comfy/k_diffusion/sampling.py b/comfy/k_diffusion/sampling.py
index 7af016829..f9b281894 100644
--- a/comfy/k_diffusion/sampling.py
+++ b/comfy/k_diffusion/sampling.py
@@ -527,6 +527,9 @@ def sample_dpmpp_2s_ancestral(model, x, sigmas, extra_args=None, callback=None,
@torch.no_grad()
def sample_dpmpp_sde(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None, r=1 / 2):
"""DPM-Solver++ (stochastic)."""
+ if len(sigmas) <= 1:
+ return x
+
sigma_min, sigma_max = sigmas[sigmas > 0].min(), sigmas.max()
seed = extra_args.get("seed", None)
noise_sampler = BrownianTreeNoiseSampler(x, sigma_min, sigma_max, seed=seed, cpu=True) if noise_sampler is None else noise_sampler
@@ -595,6 +598,8 @@ def sample_dpmpp_2m(model, x, sigmas, extra_args=None, callback=None, disable=No
@torch.no_grad()
def sample_dpmpp_2m_sde(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None, solver_type='midpoint'):
"""DPM-Solver++(2M) SDE."""
+ if len(sigmas) <= 1:
+ return x
if solver_type not in {'heun', 'midpoint'}:
raise ValueError('solver_type must be \'heun\' or \'midpoint\'')
@@ -642,6 +647,9 @@ def sample_dpmpp_2m_sde(model, x, sigmas, extra_args=None, callback=None, disabl
def sample_dpmpp_3m_sde(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None):
"""DPM-Solver++(3M) SDE."""
+ if len(sigmas) <= 1:
+ return x
+
seed = extra_args.get("seed", None)
sigma_min, sigma_max = sigmas[sigmas > 0].min(), sigmas.max()
noise_sampler = BrownianTreeNoiseSampler(x, sigma_min, sigma_max, seed=seed, cpu=True) if noise_sampler is None else noise_sampler
@@ -690,18 +698,27 @@ def sample_dpmpp_3m_sde(model, x, sigmas, extra_args=None, callback=None, disabl
@torch.no_grad()
def sample_dpmpp_3m_sde_gpu(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None):
+ if len(sigmas) <= 1:
+ return x
+
sigma_min, sigma_max = sigmas[sigmas > 0].min(), sigmas.max()
noise_sampler = BrownianTreeNoiseSampler(x, sigma_min, sigma_max, seed=extra_args.get("seed", None), cpu=False) if noise_sampler is None else noise_sampler
return sample_dpmpp_3m_sde(model, x, sigmas, extra_args=extra_args, callback=callback, disable=disable, eta=eta, s_noise=s_noise, noise_sampler=noise_sampler)
@torch.no_grad()
def sample_dpmpp_2m_sde_gpu(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None, solver_type='midpoint'):
+ if len(sigmas) <= 1:
+ return x
+
sigma_min, sigma_max = sigmas[sigmas > 0].min(), sigmas.max()
noise_sampler = BrownianTreeNoiseSampler(x, sigma_min, sigma_max, seed=extra_args.get("seed", None), cpu=False) if noise_sampler is None else noise_sampler
return sample_dpmpp_2m_sde(model, x, sigmas, extra_args=extra_args, callback=callback, disable=disable, eta=eta, s_noise=s_noise, noise_sampler=noise_sampler, solver_type=solver_type)
@torch.no_grad()
def sample_dpmpp_sde_gpu(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None, r=1 / 2):
+ if len(sigmas) <= 1:
+ return x
+
sigma_min, sigma_max = sigmas[sigmas > 0].min(), sigmas.max()
noise_sampler = BrownianTreeNoiseSampler(x, sigma_min, sigma_max, seed=extra_args.get("seed", None), cpu=False) if noise_sampler is None else noise_sampler
return sample_dpmpp_sde(model, x, sigmas, extra_args=extra_args, callback=callback, disable=disable, eta=eta, s_noise=s_noise, noise_sampler=noise_sampler, r=r)
diff --git a/comfy/ldm/modules/attention.py b/comfy/ldm/modules/attention.py
index f116efee3..d51a2fae1 100644
--- a/comfy/ldm/modules/attention.py
+++ b/comfy/ldm/modules/attention.py
@@ -297,8 +297,8 @@ def attention_split(q, k, v, heads, mask=None):
BROKEN_XFORMERS = False
try:
x_vers = xformers.__version__
- #I think 0.0.23 is also broken (q with bs bigger than 65535 gives CUDA error)
- BROKEN_XFORMERS = x_vers.startswith("0.0.21") or x_vers.startswith("0.0.22") or x_vers.startswith("0.0.23")
+ # XFormers bug confirmed on all versions from 0.0.21 to 0.0.26 (q with bs bigger than 65535 gives CUDA error)
+ BROKEN_XFORMERS = x_vers.startswith("0.0.2") and not x_vers.startswith("0.0.20")
except:
pass
diff --git a/comfy/model_base.py b/comfy/model_base.py
index 898adb665..841598b73 100644
--- a/comfy/model_base.py
+++ b/comfy/model_base.py
@@ -162,7 +162,7 @@ class BaseModel(torch.nn.Module):
c_concat = kwargs.get("noise_concat", None)
if c_concat is not None:
- out['c_concat'] = comfy.conds.CONDNoiseShape(data)
+ out['c_concat'] = comfy.conds.CONDNoiseShape(c_concat)
return out
@@ -503,8 +503,10 @@ class SD15_instructpix2pix(IP2P, BaseModel):
class SDXL_instructpix2pix(IP2P, SDXL):
def __init__(self, model_config, model_type=ModelType.EPS, device=None):
super().__init__(model_config, model_type, device=device)
- # self.process_ip2p_image_in = lambda image: comfy.latent_formats.SDXL().process_in(image)
- self.process_ip2p_image_in = lambda image: image
+ if model_type == ModelType.V_PREDICTION_EDM:
+ self.process_ip2p_image_in = lambda image: comfy.latent_formats.SDXL().process_in(image) #cosxl ip2p
+ else:
+ self.process_ip2p_image_in = lambda image: image #diffusers ip2p
class StableCascade_C(BaseModel):
diff --git a/comfy/model_detection.py b/comfy/model_detection.py
index 795af828a..23358a2c0 100644
--- a/comfy/model_detection.py
+++ b/comfy/model_detection.py
@@ -357,7 +357,14 @@ def unet_config_from_diffusers_unet(state_dict, dtype=None):
'context_dim': 1024, 'num_head_channels': 64, 'transformer_depth_output': [1, 1, 1, 1, 1, 1],
'use_temporal_attention': False, 'use_temporal_resblock': False, 'disable_self_attentions': [True, False, False]}
- supported_models = [SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega, KOALA_700M, KOALA_1B, SD09_XS, SDXL_diffusers_ip2p]
+ SD_XS = {'use_checkpoint': False, 'image_size': 32, 'out_channels': 4, 'use_spatial_transformer': True, 'legacy': False,
+ 'adm_in_channels': None, 'dtype': dtype, 'in_channels': 4, 'model_channels': 320, 'num_res_blocks': [1, 1, 1],
+ 'transformer_depth': [0, 1, 1], 'channel_mult': [1, 2, 4], 'transformer_depth_middle': -2, 'use_linear_in_transformer': False,
+ 'context_dim': 768, 'num_head_channels': 64, 'transformer_depth_output': [0, 0, 1, 1, 1, 1],
+ 'use_temporal_attention': False, 'use_temporal_resblock': False}
+
+
+ supported_models = [SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega, KOALA_700M, KOALA_1B, SD09_XS, SD_XS, SDXL_diffusers_ip2p]
for unet_config in supported_models:
matches = True
diff --git a/comfy/model_management.py b/comfy/model_management.py
index 26216432a..913b6844f 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -83,7 +83,7 @@ def get_torch_device():
return torch.device("cpu")
else:
if is_intel_xpu():
- return torch.device("xpu")
+ return torch.device("xpu", torch.xpu.current_device())
else:
return torch.device(torch.cuda.current_device())
@@ -304,7 +304,7 @@ class LoadedModel:
raise e
if is_intel_xpu() and not args.disable_ipex_optimize:
- self.real_model = torch.xpu.optimize(self.real_model.eval(), inplace=True, auto_kernel_selection=True, graph_mode=True)
+ self.real_model = ipex.optimize(self.real_model.eval(), graph_mode=True, concat_linear=True)
self.weights_loaded = True
return self.real_model
@@ -385,6 +385,8 @@ def load_models_gpu(models, memory_required=0):
inference_memory = minimum_inference_memory()
extra_mem = max(inference_memory, memory_required)
+ models = set(models)
+
models_to_load = []
models_already_loaded = []
for x in models:
@@ -550,8 +552,6 @@ def text_encoder_device():
if args.gpu_only:
return get_torch_device()
elif vram_state == VRAMState.HIGH_VRAM or vram_state == VRAMState.NORMAL_VRAM:
- if is_intel_xpu():
- return torch.device("cpu")
if should_use_fp16(prioritize_performance=False):
return get_torch_device()
else:
@@ -615,7 +615,8 @@ def supports_dtype(device, dtype): #TODO
def device_supports_non_blocking(device):
if is_device_mps(device):
return False #pytorch bug? mps doesn't support non blocking
- return True
+ return False
+ # return True #TODO: figure out why this causes issues
def cast_to_device(tensor, device, dtype, copy=False):
device_supports_cast = False
diff --git a/comfy/model_patcher.py b/comfy/model_patcher.py
index 8dda84cfd..cf51c4ad8 100644
--- a/comfy/model_patcher.py
+++ b/comfy/model_patcher.py
@@ -18,6 +18,26 @@ def apply_weight_decompose(dora_scale, weight):
return weight * (dora_scale / weight_norm)
+def set_model_options_patch_replace(model_options, patch, name, block_name, number, transformer_index=None):
+ to = model_options["transformer_options"].copy()
+
+ if "patches_replace" not in to:
+ to["patches_replace"] = {}
+ else:
+ to["patches_replace"] = to["patches_replace"].copy()
+
+ if name not in to["patches_replace"]:
+ to["patches_replace"][name] = {}
+ else:
+ to["patches_replace"][name] = to["patches_replace"][name].copy()
+
+ if transformer_index is not None:
+ block = (block_name, number, transformer_index)
+ else:
+ block = (block_name, number)
+ to["patches_replace"][name][block] = patch
+ model_options["transformer_options"] = to
+ return model_options
class ModelPatcher:
def __init__(self, model, load_device, offload_device, size=0, current_device=None, weight_inplace_update=False):
@@ -109,16 +129,7 @@ class ModelPatcher:
to["patches"][name] = to["patches"].get(name, []) + [patch]
def set_model_patch_replace(self, patch, name, block_name, number, transformer_index=None):
- to = self.model_options["transformer_options"]
- if "patches_replace" not in to:
- to["patches_replace"] = {}
- if name not in to["patches_replace"]:
- to["patches_replace"][name] = {}
- if transformer_index is not None:
- block = (block_name, number, transformer_index)
- else:
- block = (block_name, number)
- to["patches_replace"][name][block] = patch
+ self.model_options = set_model_options_patch_replace(self.model_options, patch, name, block_name, number, transformer_index=transformer_index)
def set_model_attn1_patch(self, patch):
self.set_model_patch(patch, "attn1_patch")
@@ -150,6 +161,15 @@ class ModelPatcher:
def add_object_patch(self, name, obj):
self.object_patches[name] = obj
+ def get_model_object(self, name):
+ if name in self.object_patches:
+ return self.object_patches[name]
+ else:
+ if name in self.object_patches_backup:
+ return self.object_patches_backup[name]
+ else:
+ return comfy.utils.get_attr(self.model, name)
+
def model_patches_to(self, device):
to = self.model_options["transformer_options"]
if "patches" in to:
@@ -278,7 +298,7 @@ class ModelPatcher:
if weight_key in self.patches:
m.weight_function = LowVramPatch(weight_key, self)
if bias_key in self.patches:
- m.bias_function = LowVramPatch(weight_key, self)
+ m.bias_function = LowVramPatch(bias_key, self)
m.prev_comfy_cast_weights = m.comfy_cast_weights
m.comfy_cast_weights = True
@@ -462,4 +482,4 @@ class ModelPatcher:
for k in keys:
comfy.utils.set_attr(self.model, k, self.object_patches_backup[k])
- self.object_patches_backup = {}
+ self.object_patches_backup.clear()
diff --git a/comfy/sample.py b/comfy/sample.py
index 5c8a7d130..e51bd67d6 100644
--- a/comfy/sample.py
+++ b/comfy/sample.py
@@ -1,10 +1,9 @@
import torch
import comfy.model_management
import comfy.samplers
-import comfy.conds
import comfy.utils
-import math
import numpy as np
+import logging
def prepare_noise(latent_image, seed, noise_inds=None):
"""
@@ -25,94 +24,21 @@ def prepare_noise(latent_image, seed, noise_inds=None):
noises = torch.cat(noises, axis=0)
return noises
-def prepare_mask(noise_mask, shape, device):
- """ensures noise mask is of proper dimensions"""
- noise_mask = torch.nn.functional.interpolate(noise_mask.reshape((-1, 1, noise_mask.shape[-2], noise_mask.shape[-1])), size=(shape[2], shape[3]), mode="bilinear")
- noise_mask = torch.cat([noise_mask] * shape[1], dim=1)
- noise_mask = comfy.utils.repeat_to_batch_size(noise_mask, shape[0])
- noise_mask = noise_mask.to(device)
- return noise_mask
-
-def get_models_from_cond(cond, model_type):
- models = []
- for c in cond:
- if model_type in c:
- models += [c[model_type]]
- return models
-
-def convert_cond(cond):
- out = []
- for c in cond:
- temp = c[1].copy()
- model_conds = temp.get("model_conds", {})
- if c[0] is not None:
- model_conds["c_crossattn"] = comfy.conds.CONDCrossAttn(c[0]) #TODO: remove
- temp["cross_attn"] = c[0]
- temp["model_conds"] = model_conds
- out.append(temp)
- return out
-
-def get_additional_models(positive, negative, dtype):
- """loads additional models in positive and negative conditioning"""
- control_nets = set(get_models_from_cond(positive, "control") + get_models_from_cond(negative, "control"))
-
- inference_memory = 0
- control_models = []
- for m in control_nets:
- control_models += m.get_models()
- inference_memory += m.inference_memory_requirements(dtype)
-
- gligen = get_models_from_cond(positive, "gligen") + get_models_from_cond(negative, "gligen")
- gligen = [x[1] for x in gligen]
- models = control_models + gligen
- return models, inference_memory
+def prepare_sampling(model, noise_shape, positive, negative, noise_mask):
+ logging.warning("Warning: comfy.sample.prepare_sampling isn't used anymore and can be removed")
+ return model, positive, negative, noise_mask, []
def cleanup_additional_models(models):
- """cleanup additional models that were loaded"""
- for m in models:
- if hasattr(m, 'cleanup'):
- m.cleanup()
-
-def prepare_sampling(model, noise_shape, positive, negative, noise_mask):
- device = model.load_device
- positive = convert_cond(positive)
- negative = convert_cond(negative)
-
- if noise_mask is not None:
- noise_mask = prepare_mask(noise_mask, noise_shape, device)
-
- real_model = None
- models, inference_memory = get_additional_models(positive, negative, model.model_dtype())
- comfy.model_management.load_models_gpu([model] + models, model.memory_required([noise_shape[0] * 2] + list(noise_shape[1:])) + inference_memory)
- real_model = model.model
-
- return real_model, positive, negative, noise_mask, models
-
+ logging.warning("Warning: comfy.sample.cleanup_additional_models isn't used anymore and can be removed")
def sample(model, noise, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=1.0, disable_noise=False, start_step=None, last_step=None, force_full_denoise=False, noise_mask=None, sigmas=None, callback=None, disable_pbar=False, seed=None):
- real_model, positive_copy, negative_copy, noise_mask, models = prepare_sampling(model, noise.shape, positive, negative, noise_mask)
+ sampler = comfy.samplers.KSampler(model, steps=steps, device=model.load_device, sampler=sampler_name, scheduler=scheduler, denoise=denoise, model_options=model.model_options)
- noise = noise.to(model.load_device)
- latent_image = latent_image.to(model.load_device)
-
- sampler = comfy.samplers.KSampler(real_model, steps=steps, device=model.load_device, sampler=sampler_name, scheduler=scheduler, denoise=denoise, model_options=model.model_options)
-
- samples = sampler.sample(noise, positive_copy, negative_copy, cfg=cfg, latent_image=latent_image, start_step=start_step, last_step=last_step, force_full_denoise=force_full_denoise, denoise_mask=noise_mask, sigmas=sigmas, callback=callback, disable_pbar=disable_pbar, seed=seed)
+ samples = sampler.sample(noise, positive, negative, cfg=cfg, latent_image=latent_image, start_step=start_step, last_step=last_step, force_full_denoise=force_full_denoise, denoise_mask=noise_mask, sigmas=sigmas, callback=callback, disable_pbar=disable_pbar, seed=seed)
samples = samples.to(comfy.model_management.intermediate_device())
-
- cleanup_additional_models(models)
- cleanup_additional_models(set(get_models_from_cond(positive_copy, "control") + get_models_from_cond(negative_copy, "control")))
return samples
def sample_custom(model, noise, cfg, sampler, sigmas, positive, negative, latent_image, noise_mask=None, callback=None, disable_pbar=False, seed=None):
- real_model, positive_copy, negative_copy, noise_mask, models = prepare_sampling(model, noise.shape, positive, negative, noise_mask)
- noise = noise.to(model.load_device)
- latent_image = latent_image.to(model.load_device)
- sigmas = sigmas.to(model.load_device)
-
- samples = comfy.samplers.sample(real_model, noise, positive_copy, negative_copy, cfg, model.load_device, sampler, sigmas, model_options=model.model_options, latent_image=latent_image, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=seed)
+ samples = comfy.samplers.sample(model, noise, positive, negative, cfg, model.load_device, sampler, sigmas, model_options=model.model_options, latent_image=latent_image, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=seed)
samples = samples.to(comfy.model_management.intermediate_device())
- cleanup_additional_models(models)
- cleanup_additional_models(set(get_models_from_cond(positive_copy, "control") + get_models_from_cond(negative_copy, "control")))
return samples
-
diff --git a/comfy/sampler_helpers.py b/comfy/sampler_helpers.py
new file mode 100644
index 000000000..a18abd9e9
--- /dev/null
+++ b/comfy/sampler_helpers.py
@@ -0,0 +1,76 @@
+import torch
+import comfy.model_management
+import comfy.conds
+
+def prepare_mask(noise_mask, shape, device):
+ """ensures noise mask is of proper dimensions"""
+ noise_mask = torch.nn.functional.interpolate(noise_mask.reshape((-1, 1, noise_mask.shape[-2], noise_mask.shape[-1])), size=(shape[2], shape[3]), mode="bilinear")
+ noise_mask = torch.cat([noise_mask] * shape[1], dim=1)
+ noise_mask = comfy.utils.repeat_to_batch_size(noise_mask, shape[0])
+ noise_mask = noise_mask.to(device)
+ return noise_mask
+
+def get_models_from_cond(cond, model_type):
+ models = []
+ for c in cond:
+ if model_type in c:
+ models += [c[model_type]]
+ return models
+
+def convert_cond(cond):
+ out = []
+ for c in cond:
+ temp = c[1].copy()
+ model_conds = temp.get("model_conds", {})
+ if c[0] is not None:
+ model_conds["c_crossattn"] = comfy.conds.CONDCrossAttn(c[0]) #TODO: remove
+ temp["cross_attn"] = c[0]
+ temp["model_conds"] = model_conds
+ out.append(temp)
+ return out
+
+def get_additional_models(conds, dtype):
+ """loads additional models in conditioning"""
+ cnets = []
+ gligen = []
+
+ for k in conds:
+ cnets += get_models_from_cond(conds[k], "control")
+ gligen += get_models_from_cond(conds[k], "gligen")
+
+ control_nets = set(cnets)
+
+ inference_memory = 0
+ control_models = []
+ for m in control_nets:
+ control_models += m.get_models()
+ inference_memory += m.inference_memory_requirements(dtype)
+
+ gligen = [x[1] for x in gligen]
+ models = control_models + gligen
+ return models, inference_memory
+
+def cleanup_additional_models(models):
+ """cleanup additional models that were loaded"""
+ for m in models:
+ if hasattr(m, 'cleanup'):
+ m.cleanup()
+
+
+def prepare_sampling(model, noise_shape, conds):
+ device = model.load_device
+ real_model = None
+ models, inference_memory = get_additional_models(conds, model.model_dtype())
+ comfy.model_management.load_models_gpu([model] + models, model.memory_required([noise_shape[0] * 2] + list(noise_shape[1:])) + inference_memory)
+ real_model = model.model
+
+ return real_model, conds, models
+
+def cleanup_models(conds, models):
+ cleanup_additional_models(models)
+
+ control_cleanup = []
+ for k in conds:
+ control_cleanup += get_models_from_cond(conds[k], "control")
+
+ cleanup_additional_models(set(control_cleanup))
diff --git a/comfy/samplers.py b/comfy/samplers.py
index 204a98f95..29962a916 100644
--- a/comfy/samplers.py
+++ b/comfy/samplers.py
@@ -5,6 +5,7 @@ import collections
from comfy import model_management
import math
import logging
+import comfy.sampler_helpers
def get_area_and_mult(conds, x_in, timestep_in):
area = (x_in.shape[2], x_in.shape[3], 0, 0)
@@ -33,7 +34,7 @@ def get_area_and_mult(conds, x_in, timestep_in):
mask = conds['mask']
assert(mask.shape[1] == x_in.shape[2])
assert(mask.shape[2] == x_in.shape[3])
- mask = mask[:,area[2]:area[0] + area[2],area[3]:area[1] + area[3]] * mask_strength
+ mask = mask[:input_x.shape[0],area[2]:area[0] + area[2],area[3]:area[1] + area[3]] * mask_strength
mask = mask.unsqueeze(1).repeat(input_x.shape[0] // mask.shape[0], input_x.shape[1], 1, 1)
else:
mask = torch.ones_like(input_x)
@@ -230,63 +231,51 @@ def calc_cond_uncond_batch(model, cond, uncond, x_in, timestep, model_options):
logging.warning("WARNING: The comfy.samplers.calc_cond_uncond_batch function is deprecated please use the calc_cond_batch one instead.")
return tuple(calc_cond_batch(model, [cond, uncond], x_in, timestep, model_options))
+def cfg_function(model, cond_pred, uncond_pred, cond_scale, x, timestep, model_options={}, cond=None, uncond=None):
+ if "sampler_cfg_function" in model_options:
+ args = {"cond": x - cond_pred, "uncond": x - uncond_pred, "cond_scale": cond_scale, "timestep": timestep, "input": x, "sigma": timestep,
+ "cond_denoised": cond_pred, "uncond_denoised": uncond_pred, "model": model, "model_options": model_options}
+ cfg_result = x - model_options["sampler_cfg_function"](args)
+ else:
+ cfg_result = uncond_pred + (cond_pred - uncond_pred) * cond_scale
+
+ for fn in model_options.get("sampler_post_cfg_function", []):
+ args = {"denoised": cfg_result, "cond": cond, "uncond": uncond, "model": model, "uncond_denoised": uncond_pred, "cond_denoised": cond_pred,
+ "sigma": timestep, "model_options": model_options, "input": x}
+ cfg_result = fn(args)
+
+ return cfg_result
+
#The main sampling function shared by all the samplers
#Returns denoised
def sampling_function(model, x, timestep, uncond, cond, cond_scale, model_options={}, seed=None):
- if math.isclose(cond_scale, 1.0) and model_options.get("disable_cfg1_optimization", False) == False:
- uncond_ = None
- else:
- uncond_ = uncond
+ if math.isclose(cond_scale, 1.0) and model_options.get("disable_cfg1_optimization", False) == False:
+ uncond_ = None
+ else:
+ uncond_ = uncond
+
+ conds = [cond, uncond_]
+ out = calc_cond_batch(model, conds, x, timestep, model_options)
+ return cfg_function(model, out[0], out[1], cond_scale, x, timestep, model_options=model_options, cond=cond, uncond=uncond_)
- conds = [cond, uncond_]
-
- out = calc_cond_batch(model, conds, x, timestep, model_options)
- cond_pred = out[0]
- uncond_pred = out[1]
-
- if "sampler_cfg_function" in model_options:
- args = {"cond": x - cond_pred, "uncond": x - uncond_pred, "cond_scale": cond_scale, "timestep": timestep, "input": x, "sigma": timestep,
- "cond_denoised": cond_pred, "uncond_denoised": uncond_pred, "model": model, "model_options": model_options}
- cfg_result = x - model_options["sampler_cfg_function"](args)
- else:
- cfg_result = uncond_pred + (cond_pred - uncond_pred) * cond_scale
-
- for fn in model_options.get("sampler_post_cfg_function", []):
- args = {"denoised": cfg_result, "cond": cond, "uncond": uncond, "model": model, "uncond_denoised": uncond_pred, "cond_denoised": cond_pred,
- "sigma": timestep, "model_options": model_options, "input": x}
- cfg_result = fn(args)
-
- return cfg_result
-
-class CFGNoisePredictor(torch.nn.Module):
- def __init__(self, model):
- super().__init__()
- self.inner_model = model
- def apply_model(self, x, timestep, cond, uncond, cond_scale, model_options={}, seed=None):
- out = sampling_function(self.inner_model, x, timestep, uncond, cond, cond_scale, model_options=model_options, seed=seed)
- return out
- def forward(self, *args, **kwargs):
- return self.apply_model(*args, **kwargs)
-
-class KSamplerX0Inpaint(torch.nn.Module):
+class KSamplerX0Inpaint:
def __init__(self, model, sigmas):
- super().__init__()
self.inner_model = model
self.sigmas = sigmas
- def forward(self, x, sigma, uncond, cond, cond_scale, denoise_mask, model_options={}, seed=None):
+ def __call__(self, x, sigma, denoise_mask, model_options={}, seed=None):
if denoise_mask is not None:
if "denoise_mask_function" in model_options:
denoise_mask = model_options["denoise_mask_function"](sigma, denoise_mask, extra_options={"model": self.inner_model, "sigmas": self.sigmas})
latent_mask = 1. - denoise_mask
x = x * denoise_mask + self.inner_model.inner_model.model_sampling.noise_scaling(sigma.reshape([sigma.shape[0]] + [1] * (len(self.noise.shape) - 1)), self.noise, self.latent_image) * latent_mask
- out = self.inner_model(x, sigma, cond=cond, uncond=uncond, cond_scale=cond_scale, model_options=model_options, seed=seed)
+ out = self.inner_model(x, sigma, model_options=model_options, seed=seed)
if denoise_mask is not None:
out = out * denoise_mask + self.latent_image * latent_mask
return out
-def simple_scheduler(model, steps):
- s = model.model_sampling
+def simple_scheduler(model_sampling, steps):
+ s = model_sampling
sigs = []
ss = len(s.sigmas) / steps
for x in range(steps):
@@ -294,8 +283,8 @@ def simple_scheduler(model, steps):
sigs += [0.0]
return torch.FloatTensor(sigs)
-def ddim_scheduler(model, steps):
- s = model.model_sampling
+def ddim_scheduler(model_sampling, steps):
+ s = model_sampling
sigs = []
ss = max(len(s.sigmas) // steps, 1)
x = 1
@@ -306,8 +295,8 @@ def ddim_scheduler(model, steps):
sigs += [0.0]
return torch.FloatTensor(sigs)
-def normal_scheduler(model, steps, sgm=False, floor=False):
- s = model.model_sampling
+def normal_scheduler(model_sampling, steps, sgm=False, floor=False):
+ s = model_sampling
start = s.timestep(s.sigma_max)
end = s.timestep(s.sigma_min)
@@ -550,6 +539,9 @@ class KSAMPLER(Sampler):
def ksampler(sampler_name, extra_options={}, inpaint_options={}):
if sampler_name == "dpm_fast":
def dpm_fast_function(model, noise, sigmas, extra_args, callback, disable):
+ if len(sigmas) <= 1:
+ return noise
+
sigma_min = sigmas[-1]
if sigma_min == 0:
sigma_min = sigmas[-2]
@@ -558,6 +550,9 @@ def ksampler(sampler_name, extra_options={}, inpaint_options={}):
sampler_function = dpm_fast_function
elif sampler_name == "dpm_adaptive":
def dpm_adaptive_function(model, noise, sigmas, extra_args, callback, disable, **extra_options):
+ if len(sigmas) <= 1:
+ return noise
+
sigma_min = sigmas[-1]
if sigma_min == 0:
sigma_min = sigmas[-2]
@@ -568,61 +563,122 @@ def ksampler(sampler_name, extra_options={}, inpaint_options={}):
return KSAMPLER(sampler_function, extra_options, inpaint_options)
-def wrap_model(model):
- model_denoise = CFGNoisePredictor(model)
- return model_denoise
-def sample(model, noise, positive, negative, cfg, device, sampler, sigmas, model_options={}, latent_image=None, denoise_mask=None, callback=None, disable_pbar=False, seed=None):
- positive = positive[:]
- negative = negative[:]
+def process_conds(model, noise, conds, device, latent_image=None, denoise_mask=None, seed=None):
+ for k in conds:
+ conds[k] = conds[k][:]
+ resolve_areas_and_cond_masks(conds[k], noise.shape[2], noise.shape[3], device)
- resolve_areas_and_cond_masks(positive, noise.shape[2], noise.shape[3], device)
- resolve_areas_and_cond_masks(negative, noise.shape[2], noise.shape[3], device)
-
- model_wrap = wrap_model(model)
-
- calculate_start_end_timesteps(model, negative)
- calculate_start_end_timesteps(model, positive)
-
- if latent_image is not None and torch.count_nonzero(latent_image) > 0: #Don't shift the empty latent image.
- latent_image = model.process_latent_in(latent_image)
+ for k in conds:
+ calculate_start_end_timesteps(model, conds[k])
if hasattr(model, 'extra_conds'):
- positive = encode_model_conds(model.extra_conds, positive, noise, device, "positive", latent_image=latent_image, denoise_mask=denoise_mask, seed=seed)
- negative = encode_model_conds(model.extra_conds, negative, noise, device, "negative", latent_image=latent_image, denoise_mask=denoise_mask, seed=seed)
+ for k in conds:
+ conds[k] = encode_model_conds(model.extra_conds, conds[k], noise, device, k, latent_image=latent_image, denoise_mask=denoise_mask, seed=seed)
#make sure each cond area has an opposite one with the same area
- for c in positive:
- create_cond_with_same_area_if_none(negative, c)
- for c in negative:
- create_cond_with_same_area_if_none(positive, c)
+ for k in conds:
+ for c in conds[k]:
+ for kk in conds:
+ if k != kk:
+ create_cond_with_same_area_if_none(conds[kk], c)
- pre_run_control(model, negative + positive)
+ for k in conds:
+ pre_run_control(model, conds[k])
- apply_empty_x_to_equal_area(list(filter(lambda c: c.get('control_apply_to_uncond', False) == True, positive)), negative, 'control', lambda cond_cnets, x: cond_cnets[x])
- apply_empty_x_to_equal_area(positive, negative, 'gligen', lambda cond_cnets, x: cond_cnets[x])
+ if "positive" in conds:
+ positive = conds["positive"]
+ for k in conds:
+ if k != "positive":
+ apply_empty_x_to_equal_area(list(filter(lambda c: c.get('control_apply_to_uncond', False) == True, positive)), conds[k], 'control', lambda cond_cnets, x: cond_cnets[x])
+ apply_empty_x_to_equal_area(positive, conds[k], 'gligen', lambda cond_cnets, x: cond_cnets[x])
- extra_args = {"cond":positive, "uncond":negative, "cond_scale": cfg, "model_options": model_options, "seed":seed}
+ return conds
+
+class CFGGuider:
+ def __init__(self, model_patcher):
+ self.model_patcher = model_patcher
+ self.model_options = model_patcher.model_options
+ self.original_conds = {}
+ self.cfg = 1.0
+
+ def set_conds(self, positive, negative):
+ self.inner_set_conds({"positive": positive, "negative": negative})
+
+ def set_cfg(self, cfg):
+ self.cfg = cfg
+
+ def inner_set_conds(self, conds):
+ for k in conds:
+ self.original_conds[k] = comfy.sampler_helpers.convert_cond(conds[k])
+
+ def __call__(self, *args, **kwargs):
+ return self.predict_noise(*args, **kwargs)
+
+ def predict_noise(self, x, timestep, model_options={}, seed=None):
+ return sampling_function(self.inner_model, x, timestep, self.conds.get("negative", None), self.conds.get("positive", None), self.cfg, model_options=model_options, seed=seed)
+
+ def inner_sample(self, noise, latent_image, device, sampler, sigmas, denoise_mask, callback, disable_pbar, seed):
+ if latent_image is not None and torch.count_nonzero(latent_image) > 0: #Don't shift the empty latent image.
+ latent_image = self.inner_model.process_latent_in(latent_image)
+
+ self.conds = process_conds(self.inner_model, noise, self.conds, device, latent_image, denoise_mask, seed)
+
+ extra_args = {"model_options": self.model_options, "seed":seed}
+
+ samples = sampler.sample(self, sigmas, extra_args, callback, noise, latent_image, denoise_mask, disable_pbar)
+ return self.inner_model.process_latent_out(samples.to(torch.float32))
+
+ def sample(self, noise, latent_image, sampler, sigmas, denoise_mask=None, callback=None, disable_pbar=False, seed=None):
+ if sigmas.shape[-1] == 0:
+ return latent_image
+
+ self.conds = {}
+ for k in self.original_conds:
+ self.conds[k] = list(map(lambda a: a.copy(), self.original_conds[k]))
+
+ self.inner_model, self.conds, self.loaded_models = comfy.sampler_helpers.prepare_sampling(self.model_patcher, noise.shape, self.conds)
+ device = self.model_patcher.load_device
+
+ if denoise_mask is not None:
+ denoise_mask = comfy.sampler_helpers.prepare_mask(denoise_mask, noise.shape, device)
+
+ noise = noise.to(device)
+ latent_image = latent_image.to(device)
+ sigmas = sigmas.to(device)
+
+ output = self.inner_sample(noise, latent_image, device, sampler, sigmas, denoise_mask, callback, disable_pbar, seed)
+
+ comfy.sampler_helpers.cleanup_models(self.conds, self.loaded_models)
+ del self.inner_model
+ del self.conds
+ del self.loaded_models
+ return output
+
+
+def sample(model, noise, positive, negative, cfg, device, sampler, sigmas, model_options={}, latent_image=None, denoise_mask=None, callback=None, disable_pbar=False, seed=None):
+ cfg_guider = CFGGuider(model)
+ cfg_guider.set_conds(positive, negative)
+ cfg_guider.set_cfg(cfg)
+ return cfg_guider.sample(noise, latent_image, sampler, sigmas, denoise_mask, callback, disable_pbar, seed)
- samples = sampler.sample(model_wrap, sigmas, extra_args, callback, noise, latent_image, denoise_mask, disable_pbar)
- return model.process_latent_out(samples.to(torch.float32))
SCHEDULER_NAMES = ["normal", "karras", "exponential", "sgm_uniform", "simple", "ddim_uniform"]
SAMPLER_NAMES = KSAMPLER_NAMES + ["ddim", "uni_pc", "uni_pc_bh2"]
-def calculate_sigmas_scheduler(model, scheduler_name, steps):
+def calculate_sigmas(model_sampling, scheduler_name, steps):
if scheduler_name == "karras":
- sigmas = k_diffusion_sampling.get_sigmas_karras(n=steps, sigma_min=float(model.model_sampling.sigma_min), sigma_max=float(model.model_sampling.sigma_max))
+ sigmas = k_diffusion_sampling.get_sigmas_karras(n=steps, sigma_min=float(model_sampling.sigma_min), sigma_max=float(model_sampling.sigma_max))
elif scheduler_name == "exponential":
- sigmas = k_diffusion_sampling.get_sigmas_exponential(n=steps, sigma_min=float(model.model_sampling.sigma_min), sigma_max=float(model.model_sampling.sigma_max))
+ sigmas = k_diffusion_sampling.get_sigmas_exponential(n=steps, sigma_min=float(model_sampling.sigma_min), sigma_max=float(model_sampling.sigma_max))
elif scheduler_name == "normal":
- sigmas = normal_scheduler(model, steps)
+ sigmas = normal_scheduler(model_sampling, steps)
elif scheduler_name == "simple":
- sigmas = simple_scheduler(model, steps)
+ sigmas = simple_scheduler(model_sampling, steps)
elif scheduler_name == "ddim_uniform":
- sigmas = ddim_scheduler(model, steps)
+ sigmas = ddim_scheduler(model_sampling, steps)
elif scheduler_name == "sgm_uniform":
- sigmas = normal_scheduler(model, steps, sgm=True)
+ sigmas = normal_scheduler(model_sampling, steps, sgm=True)
else:
logging.error("error invalid scheduler {}".format(scheduler_name))
return sigmas
@@ -664,7 +720,7 @@ class KSampler:
steps += 1
discard_penultimate_sigma = True
- sigmas = calculate_sigmas_scheduler(self.model, self.scheduler, steps)
+ sigmas = calculate_sigmas(self.model.get_model_object("model_sampling"), self.scheduler, steps)
if discard_penultimate_sigma:
sigmas = torch.cat([sigmas[:-2], sigmas[-1:]])
@@ -675,9 +731,12 @@ class KSampler:
if denoise is None or denoise > 0.9999:
self.sigmas = self.calculate_sigmas(steps).to(self.device)
else:
- new_steps = int(steps/denoise)
- sigmas = self.calculate_sigmas(new_steps).to(self.device)
- self.sigmas = sigmas[-(steps + 1):]
+ if denoise <= 0.0:
+ self.sigmas = torch.FloatTensor([])
+ else:
+ new_steps = int(steps/denoise)
+ sigmas = self.calculate_sigmas(new_steps).to(self.device)
+ self.sigmas = sigmas[-(steps + 1):]
def sample(self, noise, positive, negative, cfg, latent_image=None, start_step=None, last_step=None, force_full_denoise=False, denoise_mask=None, sigmas=None, callback=None, disable_pbar=False, seed=None):
if sigmas is None:
diff --git a/comfy/sd.py b/comfy/sd.py
index 85821120e..9671e4aee 100644
--- a/comfy/sd.py
+++ b/comfy/sd.py
@@ -210,16 +210,26 @@ class VAE:
self.first_stage_model = StageC_coder()
self.downscale_ratio = 32
self.latent_channels = 16
- else:
+ elif "decoder.conv_in.weight" in sd:
#default SD1.x/SD2.x VAE parameters
ddconfig = {'double_z': True, 'z_channels': 4, 'resolution': 256, 'in_channels': 3, 'out_ch': 3, 'ch': 128, 'ch_mult': [1, 2, 4, 4], 'num_res_blocks': 2, 'attn_resolutions': [], 'dropout': 0.0}
- if 'encoder.down.2.downsample.conv.weight' not in sd: #Stable diffusion x4 upscaler VAE
+ if 'encoder.down.2.downsample.conv.weight' not in sd and 'decoder.up.3.upsample.conv.weight' not in sd: #Stable diffusion x4 upscaler VAE
ddconfig['ch_mult'] = [1, 2, 4]
self.downscale_ratio = 4
self.upscale_ratio = 4
- self.first_stage_model = AutoencoderKL(ddconfig=ddconfig, embed_dim=4)
+ self.latent_channels = ddconfig['z_channels'] = sd["decoder.conv_in.weight"].shape[1]
+ if 'quant_conv.weight' in sd:
+ self.first_stage_model = AutoencoderKL(ddconfig=ddconfig, embed_dim=4)
+ else:
+ self.first_stage_model = AutoencodingEngine(regularizer_config={'target': "comfy.ldm.models.autoencoder.DiagonalGaussianRegularizer"},
+ encoder_config={'target': "comfy.ldm.modules.diffusionmodules.model.Encoder", 'params': ddconfig},
+ decoder_config={'target': "comfy.ldm.modules.diffusionmodules.model.Decoder", 'params': ddconfig})
+ else:
+ logging.warning("WARNING: No VAE weights detected, VAE not initalized.")
+ self.first_stage_model = None
+ return
else:
self.first_stage_model = AutoencoderKL(**(config['params']))
self.first_stage_model = self.first_stage_model.eval()
@@ -408,6 +418,8 @@ def load_gligen(ckpt_path):
return comfy.model_patcher.ModelPatcher(model, load_device=model_management.get_torch_device(), offload_device=model_management.unet_offload_device())
def load_checkpoint(config_path=None, ckpt_path=None, output_vae=True, output_clip=True, embedding_directory=None, state_dict=None, config=None):
+ logging.warning("Warning: The load checkpoint with config function is deprecated and will eventually be removed, please use the other one.")
+ model, clip, vae, _ = load_checkpoint_guess_config(ckpt_path, output_vae=output_vae, output_clip=output_clip, output_clipvision=False, embedding_directory=embedding_directory, output_model=True)
#TODO: this function is a mess and should be removed eventually
if config is None:
with open(config_path, 'r') as stream:
@@ -415,81 +427,20 @@ def load_checkpoint(config_path=None, ckpt_path=None, output_vae=True, output_cl
model_config_params = config['model']['params']
clip_config = model_config_params['cond_stage_config']
scale_factor = model_config_params['scale_factor']
- vae_config = model_config_params['first_stage_config']
-
- fp16 = False
- if "unet_config" in model_config_params:
- if "params" in model_config_params["unet_config"]:
- unet_config = model_config_params["unet_config"]["params"]
- if "use_fp16" in unet_config:
- fp16 = unet_config.pop("use_fp16")
- if fp16:
- unet_config["dtype"] = torch.float16
-
- noise_aug_config = None
- if "noise_aug_config" in model_config_params:
- noise_aug_config = model_config_params["noise_aug_config"]
-
- model_type = model_base.ModelType.EPS
if "parameterization" in model_config_params:
if model_config_params["parameterization"] == "v":
- model_type = model_base.ModelType.V_PREDICTION
+ m = model.clone()
+ class ModelSamplingAdvanced(comfy.model_sampling.ModelSamplingDiscrete, comfy.model_sampling.V_PREDICTION):
+ pass
+ m.add_object_patch("model_sampling", ModelSamplingAdvanced(model.model.model_config))
+ model = m
- clip = None
- vae = None
+ layer_idx = clip_config.get("params", {}).get("layer_idx", None)
+ if layer_idx is not None:
+ clip.clip_layer(layer_idx)
- class WeightsLoader(torch.nn.Module):
- pass
-
- if state_dict is None:
- state_dict = comfy.utils.load_torch_file(ckpt_path)
-
- class EmptyClass:
- pass
-
- model_config = comfy.supported_models_base.BASE({})
-
- from . import latent_formats
- model_config.latent_format = latent_formats.SD15(scale_factor=scale_factor)
- model_config.unet_config = model_detection.convert_config(unet_config)
-
- if config['model']["target"].endswith("ImageEmbeddingConditionedLatentDiffusion"):
- model = model_base.SD21UNCLIP(model_config, noise_aug_config["params"], model_type=model_type)
- else:
- model = model_base.BaseModel(model_config, model_type=model_type)
-
- if config['model']["target"].endswith("LatentInpaintDiffusion"):
- model.set_inpaint()
-
- if fp16:
- model = model.half()
-
- offload_device = model_management.unet_offload_device()
- model = model.to(offload_device)
- model.load_model_weights(state_dict, "model.diffusion_model.")
-
- if output_vae:
- vae_sd = comfy.utils.state_dict_prefix_replace(state_dict, {"first_stage_model.": ""}, filter_keys=True)
- vae = VAE(sd=vae_sd, config=vae_config)
-
- if output_clip:
- w = WeightsLoader()
- clip_target = EmptyClass()
- clip_target.params = clip_config.get("params", {})
- if clip_config["target"].endswith("FrozenOpenCLIPEmbedder"):
- clip_target.clip = sd2_clip.SD2ClipModel
- clip_target.tokenizer = sd2_clip.SD2Tokenizer
- clip = CLIP(clip_target, embedding_directory=embedding_directory)
- w.cond_stage_model = clip.cond_stage_model.clip_h
- elif clip_config["target"].endswith("FrozenCLIPEmbedder"):
- clip_target.clip = sd1_clip.SD1ClipModel
- clip_target.tokenizer = sd1_clip.SD1Tokenizer
- clip = CLIP(clip_target, embedding_directory=embedding_directory)
- w.cond_stage_model = clip.cond_stage_model.clip_l
- load_clip_weights(w, state_dict)
-
- return (comfy.model_patcher.ModelPatcher(model, load_device=model_management.get_torch_device(), offload_device=offload_device), clip, vae)
+ return (model, clip, vae)
def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, output_clipvision=False, embedding_directory=None, output_model=True):
sd = comfy.utils.load_torch_file(ckpt_path)
@@ -535,7 +486,11 @@ def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, o
clip = CLIP(clip_target, embedding_directory=embedding_directory)
m, u = clip.load_sd(clip_sd, full_model=True)
if len(m) > 0:
- logging.warning("clip missing: {}".format(m))
+ m_filter = list(filter(lambda a: ".logit_scale" not in a and ".transformer.text_projection.weight" not in a, m))
+ if len(m_filter) > 0:
+ logging.warning("clip missing: {}".format(m))
+ else:
+ logging.debug("clip missing: {}".format(m))
if len(u) > 0:
logging.debug("clip unexpected {}:".format(u))
@@ -600,7 +555,7 @@ def load_unet(unet_path):
raise RuntimeError("ERROR: Could not detect model type of: {}".format(unet_path))
return model
-def save_checkpoint(output_path, model, clip=None, vae=None, clip_vision=None, metadata=None):
+def save_checkpoint(output_path, model, clip=None, vae=None, clip_vision=None, metadata=None, extra_keys={}):
clip_sd = None
load_models = [model]
if clip is not None:
@@ -610,4 +565,7 @@ def save_checkpoint(output_path, model, clip=None, vae=None, clip_vision=None, m
model_management.load_models_gpu(load_models)
clip_vision_sd = clip_vision.get_sd() if clip_vision is not None else None
sd = model.model.state_dict_for_saving(clip_sd, vae.get_sd(), clip_vision_sd)
+ for k in extra_keys:
+ sd[k] = extra_keys[k]
+
comfy.utils.save_torch_file(sd, output_path, metadata=metadata)
diff --git a/comfy/supported_models.py b/comfy/supported_models.py
index 9bfe3ea1d..b3b69e05b 100644
--- a/comfy/supported_models.py
+++ b/comfy/supported_models.py
@@ -174,6 +174,11 @@ class SDXL(supported_models_base.BASE):
self.sampling_settings["sigma_max"] = 80.0
self.sampling_settings["sigma_min"] = 0.002
return model_base.ModelType.EDM
+ elif "edm_vpred.sigma_max" in state_dict:
+ self.sampling_settings["sigma_max"] = float(state_dict["edm_vpred.sigma_max"].item())
+ if "edm_vpred.sigma_min" in state_dict:
+ self.sampling_settings["sigma_min"] = float(state_dict["edm_vpred.sigma_min"].item())
+ return model_base.ModelType.V_PREDICTION_EDM
elif "v_pred" in state_dict:
return model_base.ModelType.V_PREDICTION
else:
@@ -469,7 +474,7 @@ class SDXL_instructpix2pix(SDXL):
}
def get_model(self, state_dict, prefix="", device=None):
- return model_base.SDXL_instructpix2pix(self, device=device)
+ return model_base.SDXL_instructpix2pix(self, model_type=self.model_type(state_dict, prefix), device=device)
models = [Stable_Zero123, SD15_instructpix2pix, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXL_instructpix2pix, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B, SV3D_u, SV3D_p]
diff --git a/comfy/supported_models_base.py b/comfy/supported_models_base.py
index 6196daabf..cf7cdff34 100644
--- a/comfy/supported_models_base.py
+++ b/comfy/supported_models_base.py
@@ -47,7 +47,8 @@ class BASE:
return self.unet_config["in_channels"] > 4
def __init__(self, unet_config):
- self.unet_config = unet_config
+ self.unet_config = unet_config.copy()
+ self.sampling_settings = self.sampling_settings.copy()
self.latent_format = self.latent_format()
for x in self.unet_extra_config:
self.unet_config[x] = self.unet_extra_config[x]
diff --git a/comfy_extras/nodes_advanced_samplers.py b/comfy_extras/nodes_advanced_samplers.py
new file mode 100644
index 000000000..d973def81
--- /dev/null
+++ b/comfy_extras/nodes_advanced_samplers.py
@@ -0,0 +1,61 @@
+import comfy.samplers
+import comfy.utils
+import torch
+import numpy as np
+from tqdm.auto import trange, tqdm
+import math
+
+
+@torch.no_grad()
+def sample_lcm_upscale(model, x, sigmas, extra_args=None, callback=None, disable=None, total_upscale=2.0, upscale_method="bislerp", upscale_steps=None):
+ extra_args = {} if extra_args is None else extra_args
+
+ if upscale_steps is None:
+ upscale_steps = max(len(sigmas) // 2 + 1, 2)
+ else:
+ upscale_steps += 1
+ upscale_steps = min(upscale_steps, len(sigmas) + 1)
+
+ upscales = np.linspace(1.0, total_upscale, upscale_steps)[1:]
+
+ orig_shape = x.size()
+ s_in = x.new_ones([x.shape[0]])
+ for i in trange(len(sigmas) - 1, disable=disable):
+ denoised = model(x, sigmas[i] * s_in, **extra_args)
+ if callback is not None:
+ callback({'x': x, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i], 'denoised': denoised})
+
+ x = denoised
+ if i < len(upscales):
+ x = comfy.utils.common_upscale(x, round(orig_shape[-1] * upscales[i]), round(orig_shape[-2] * upscales[i]), upscale_method, "disabled")
+
+ if sigmas[i + 1] > 0:
+ x += sigmas[i + 1] * torch.randn_like(x)
+ return x
+
+
+class SamplerLCMUpscale:
+ upscale_methods = ["bislerp", "nearest-exact", "bilinear", "area", "bicubic"]
+
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"scale_ratio": ("FLOAT", {"default": 1.0, "min": 0.1, "max": 20.0, "step": 0.01}),
+ "scale_steps": ("INT", {"default": -1, "min": -1, "max": 1000, "step": 1}),
+ "upscale_method": (s.upscale_methods,),
+ }
+ }
+ RETURN_TYPES = ("SAMPLER",)
+ CATEGORY = "sampling/custom_sampling/samplers"
+
+ FUNCTION = "get_sampler"
+
+ def get_sampler(self, scale_ratio, scale_steps, upscale_method):
+ if scale_steps < 0:
+ scale_steps = None
+ sampler = comfy.samplers.KSAMPLER(sample_lcm_upscale, extra_options={"total_upscale": scale_ratio, "upscale_steps": scale_steps, "upscale_method": upscale_method})
+ return (sampler, )
+
+NODE_CLASS_MAPPINGS = {
+ "SamplerLCMUpscale": SamplerLCMUpscale,
+}
diff --git a/comfy_extras/nodes_align_your_steps.py b/comfy_extras/nodes_align_your_steps.py
new file mode 100644
index 000000000..3ffe53187
--- /dev/null
+++ b/comfy_extras/nodes_align_your_steps.py
@@ -0,0 +1,53 @@
+#from: https://research.nvidia.com/labs/toronto-ai/AlignYourSteps/howto.html
+import numpy as np
+import torch
+
+def loglinear_interp(t_steps, num_steps):
+ """
+ Performs log-linear interpolation of a given array of decreasing numbers.
+ """
+ xs = np.linspace(0, 1, len(t_steps))
+ ys = np.log(t_steps[::-1])
+
+ new_xs = np.linspace(0, 1, num_steps)
+ new_ys = np.interp(new_xs, xs, ys)
+
+ interped_ys = np.exp(new_ys)[::-1].copy()
+ return interped_ys
+
+NOISE_LEVELS = {"SD1": [14.6146412293, 6.4745760956, 3.8636745985, 2.6946151520, 1.8841921177, 1.3943805092, 0.9642583904, 0.6523686016, 0.3977456272, 0.1515232662, 0.0291671582],
+ "SDXL":[14.6146412293, 6.3184485287, 3.7681790315, 2.1811480769, 1.3405244945, 0.8620721141, 0.5550693289, 0.3798540708, 0.2332364134, 0.1114188177, 0.0291671582],
+ "SVD": [700.00, 54.5, 15.886, 7.977, 4.248, 1.789, 0.981, 0.403, 0.173, 0.034, 0.002]}
+
+class AlignYourStepsScheduler:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"model_type": (["SD1", "SDXL", "SVD"], ),
+ "steps": ("INT", {"default": 10, "min": 10, "max": 10000}),
+ "denoise": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}),
+ }
+ }
+ RETURN_TYPES = ("SIGMAS",)
+ CATEGORY = "sampling/custom_sampling/schedulers"
+
+ FUNCTION = "get_sigmas"
+
+ def get_sigmas(self, model_type, steps, denoise):
+ total_steps = steps
+ if denoise < 1.0:
+ if denoise <= 0.0:
+ return (torch.FloatTensor([]),)
+ total_steps = round(steps * denoise)
+
+ sigmas = NOISE_LEVELS[model_type][:]
+ if (steps + 1) != len(sigmas):
+ sigmas = loglinear_interp(sigmas, steps + 1)
+
+ sigmas = sigmas[-(total_steps + 1):]
+ sigmas[-1] = 0
+ return (torch.FloatTensor(sigmas), )
+
+NODE_CLASS_MAPPINGS = {
+ "AlignYourStepsScheduler": AlignYourStepsScheduler,
+}
diff --git a/comfy_extras/nodes_attention_multiply.py b/comfy_extras/nodes_attention_multiply.py
new file mode 100644
index 000000000..4747eb395
--- /dev/null
+++ b/comfy_extras/nodes_attention_multiply.py
@@ -0,0 +1,120 @@
+
+def attention_multiply(attn, model, q, k, v, out):
+ m = model.clone()
+ sd = model.model_state_dict()
+
+ for key in sd:
+ if key.endswith("{}.to_q.bias".format(attn)) or key.endswith("{}.to_q.weight".format(attn)):
+ m.add_patches({key: (None,)}, 0.0, q)
+ if key.endswith("{}.to_k.bias".format(attn)) or key.endswith("{}.to_k.weight".format(attn)):
+ m.add_patches({key: (None,)}, 0.0, k)
+ if key.endswith("{}.to_v.bias".format(attn)) or key.endswith("{}.to_v.weight".format(attn)):
+ m.add_patches({key: (None,)}, 0.0, v)
+ if key.endswith("{}.to_out.0.bias".format(attn)) or key.endswith("{}.to_out.0.weight".format(attn)):
+ m.add_patches({key: (None,)}, 0.0, out)
+
+ return m
+
+
+class UNetSelfAttentionMultiply:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required": { "model": ("MODEL",),
+ "q": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "k": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "v": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "out": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ }}
+ RETURN_TYPES = ("MODEL",)
+ FUNCTION = "patch"
+
+ CATEGORY = "_for_testing/attention_experiments"
+
+ def patch(self, model, q, k, v, out):
+ m = attention_multiply("attn1", model, q, k, v, out)
+ return (m, )
+
+class UNetCrossAttentionMultiply:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required": { "model": ("MODEL",),
+ "q": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "k": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "v": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "out": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ }}
+ RETURN_TYPES = ("MODEL",)
+ FUNCTION = "patch"
+
+ CATEGORY = "_for_testing/attention_experiments"
+
+ def patch(self, model, q, k, v, out):
+ m = attention_multiply("attn2", model, q, k, v, out)
+ return (m, )
+
+class CLIPAttentionMultiply:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required": { "clip": ("CLIP",),
+ "q": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "k": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "v": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "out": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ }}
+ RETURN_TYPES = ("CLIP",)
+ FUNCTION = "patch"
+
+ CATEGORY = "_for_testing/attention_experiments"
+
+ def patch(self, clip, q, k, v, out):
+ m = clip.clone()
+ sd = m.patcher.model_state_dict()
+
+ for key in sd:
+ if key.endswith("self_attn.q_proj.weight") or key.endswith("self_attn.q_proj.bias"):
+ m.add_patches({key: (None,)}, 0.0, q)
+ if key.endswith("self_attn.k_proj.weight") or key.endswith("self_attn.k_proj.bias"):
+ m.add_patches({key: (None,)}, 0.0, k)
+ if key.endswith("self_attn.v_proj.weight") or key.endswith("self_attn.v_proj.bias"):
+ m.add_patches({key: (None,)}, 0.0, v)
+ if key.endswith("self_attn.out_proj.weight") or key.endswith("self_attn.out_proj.bias"):
+ m.add_patches({key: (None,)}, 0.0, out)
+ return (m, )
+
+class UNetTemporalAttentionMultiply:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required": { "model": ("MODEL",),
+ "self_structural": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "self_temporal": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "cross_structural": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ "cross_temporal": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
+ }}
+ RETURN_TYPES = ("MODEL",)
+ FUNCTION = "patch"
+
+ CATEGORY = "_for_testing/attention_experiments"
+
+ def patch(self, model, self_structural, self_temporal, cross_structural, cross_temporal):
+ m = model.clone()
+ sd = model.model_state_dict()
+
+ for k in sd:
+ if (k.endswith("attn1.to_out.0.bias") or k.endswith("attn1.to_out.0.weight")):
+ if '.time_stack.' in k:
+ m.add_patches({k: (None,)}, 0.0, self_temporal)
+ else:
+ m.add_patches({k: (None,)}, 0.0, self_structural)
+ elif (k.endswith("attn2.to_out.0.bias") or k.endswith("attn2.to_out.0.weight")):
+ if '.time_stack.' in k:
+ m.add_patches({k: (None,)}, 0.0, cross_temporal)
+ else:
+ m.add_patches({k: (None,)}, 0.0, cross_structural)
+ return (m, )
+
+NODE_CLASS_MAPPINGS = {
+ "UNetSelfAttentionMultiply": UNetSelfAttentionMultiply,
+ "UNetCrossAttentionMultiply": UNetCrossAttentionMultiply,
+ "CLIPAttentionMultiply": CLIPAttentionMultiply,
+ "UNetTemporalAttentionMultiply": UNetTemporalAttentionMultiply,
+}
diff --git a/comfy_extras/nodes_canny.py b/comfy_extras/nodes_canny.py
index 8138b5f73..fab2ab7ac 100644
--- a/comfy_extras/nodes_canny.py
+++ b/comfy_extras/nodes_canny.py
@@ -1,4 +1,3 @@
-#From https://github.com/kornia/kornia
import math
import torch
diff --git a/comfy_extras/nodes_clip_sdxl.py b/comfy_extras/nodes_clip_sdxl.py
index dcf8859fa..3087b917b 100644
--- a/comfy_extras/nodes_clip_sdxl.py
+++ b/comfy_extras/nodes_clip_sdxl.py
@@ -8,7 +8,7 @@ class CLIPTextEncodeSDXLRefiner:
"ascore": ("FLOAT", {"default": 6.0, "min": 0.0, "max": 1000.0, "step": 0.01}),
"width": ("INT", {"default": 1024.0, "min": 0, "max": MAX_RESOLUTION}),
"height": ("INT", {"default": 1024.0, "min": 0, "max": MAX_RESOLUTION}),
- "text": ("STRING", {"multiline": True}), "clip": ("CLIP", ),
+ "text": ("STRING", {"multiline": True, "dynamicPrompts": True}), "clip": ("CLIP", ),
}}
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "encode"
@@ -30,8 +30,8 @@ class CLIPTextEncodeSDXL:
"crop_h": ("INT", {"default": 0, "min": 0, "max": MAX_RESOLUTION}),
"target_width": ("INT", {"default": 1024.0, "min": 0, "max": MAX_RESOLUTION}),
"target_height": ("INT", {"default": 1024.0, "min": 0, "max": MAX_RESOLUTION}),
- "text_g": ("STRING", {"multiline": True, "default": "CLIP_G"}), "clip": ("CLIP", ),
- "text_l": ("STRING", {"multiline": True, "default": "CLIP_L"}), "clip": ("CLIP", ),
+ "text_g": ("STRING", {"multiline": True, "dynamicPrompts": True}), "clip": ("CLIP", ),
+ "text_l": ("STRING", {"multiline": True, "dynamicPrompts": True}), "clip": ("CLIP", ),
}}
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "encode"
diff --git a/comfy_extras/nodes_cond.py b/comfy_extras/nodes_cond.py
index 646fefa17..4c3a1d5bf 100644
--- a/comfy_extras/nodes_cond.py
+++ b/comfy_extras/nodes_cond.py
@@ -3,7 +3,7 @@
class CLIPTextEncodeControlnet:
@classmethod
def INPUT_TYPES(s):
- return {"required": {"clip": ("CLIP", ), "conditioning": ("CONDITIONING", ), "text": ("STRING", {"multiline": True})}}
+ return {"required": {"clip": ("CLIP", ), "conditioning": ("CONDITIONING", ), "text": ("STRING", {"multiline": True, "dynamicPrompts": True})}}
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "encode"
diff --git a/comfy_extras/nodes_custom_sampler.py b/comfy_extras/nodes_custom_sampler.py
index 72ff7957f..47f08bf60 100644
--- a/comfy_extras/nodes_custom_sampler.py
+++ b/comfy_extras/nodes_custom_sampler.py
@@ -4,6 +4,7 @@ from comfy.k_diffusion import sampling as k_diffusion_sampling
import latent_preview
import torch
import comfy.utils
+import node_helpers
class BasicScheduler:
@@ -24,10 +25,11 @@ class BasicScheduler:
def get_sigmas(self, model, scheduler, steps, denoise):
total_steps = steps
if denoise < 1.0:
+ if denoise <= 0.0:
+ return (torch.FloatTensor([]),)
total_steps = int(steps/denoise)
- comfy.model_management.load_models_gpu([model])
- sigmas = comfy.samplers.calculate_sigmas_scheduler(model.model, scheduler, total_steps).cpu()
+ sigmas = comfy.samplers.calculate_sigmas(model.get_model_object("model_sampling"), scheduler, total_steps).cpu()
sigmas = sigmas[-(steps + 1):]
return (sigmas, )
@@ -37,8 +39,8 @@ class KarrasScheduler:
def INPUT_TYPES(s):
return {"required":
{"steps": ("INT", {"default": 20, "min": 1, "max": 10000}),
- "sigma_max": ("FLOAT", {"default": 14.614642, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}),
- "sigma_min": ("FLOAT", {"default": 0.0291675, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}),
+ "sigma_max": ("FLOAT", {"default": 14.614642, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}),
+ "sigma_min": ("FLOAT", {"default": 0.0291675, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}),
"rho": ("FLOAT", {"default": 7.0, "min": 0.0, "max": 100.0, "step":0.01, "round": False}),
}
}
@@ -56,8 +58,8 @@ class ExponentialScheduler:
def INPUT_TYPES(s):
return {"required":
{"steps": ("INT", {"default": 20, "min": 1, "max": 10000}),
- "sigma_max": ("FLOAT", {"default": 14.614642, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}),
- "sigma_min": ("FLOAT", {"default": 0.0291675, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}),
+ "sigma_max": ("FLOAT", {"default": 14.614642, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}),
+ "sigma_min": ("FLOAT", {"default": 0.0291675, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}),
}
}
RETURN_TYPES = ("SIGMAS",)
@@ -74,8 +76,8 @@ class PolyexponentialScheduler:
def INPUT_TYPES(s):
return {"required":
{"steps": ("INT", {"default": 20, "min": 1, "max": 10000}),
- "sigma_max": ("FLOAT", {"default": 14.614642, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}),
- "sigma_min": ("FLOAT", {"default": 0.0291675, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}),
+ "sigma_max": ("FLOAT", {"default": 14.614642, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}),
+ "sigma_min": ("FLOAT", {"default": 0.0291675, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}),
"rho": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 100.0, "step":0.01, "round": False}),
}
}
@@ -105,8 +107,7 @@ class SDTurboScheduler:
def get_sigmas(self, model, steps, denoise):
start_step = 10 - int(10 * denoise)
timesteps = torch.flip(torch.arange(1, 11) * 100 - 1, (0,))[start_step:start_step + steps]
- comfy.model_management.load_models_gpu([model])
- sigmas = model.model.model_sampling.sigma(timesteps)
+ sigmas = model.get_model_object("model_sampling").sigma(timesteps)
sigmas = torch.cat([sigmas, sigmas.new_zeros([1])])
return (sigmas, )
@@ -115,8 +116,8 @@ class VPScheduler:
def INPUT_TYPES(s):
return {"required":
{"steps": ("INT", {"default": 20, "min": 1, "max": 10000}),
- "beta_d": ("FLOAT", {"default": 19.9, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}), #TODO: fix default values
- "beta_min": ("FLOAT", {"default": 0.1, "min": 0.0, "max": 1000.0, "step":0.01, "round": False}),
+ "beta_d": ("FLOAT", {"default": 19.9, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}), #TODO: fix default values
+ "beta_min": ("FLOAT", {"default": 0.1, "min": 0.0, "max": 5000.0, "step":0.01, "round": False}),
"eps_s": ("FLOAT", {"default": 0.001, "min": 0.0, "max": 1.0, "step":0.0001, "round": False}),
}
}
@@ -138,6 +139,7 @@ class SplitSigmas:
}
}
RETURN_TYPES = ("SIGMAS","SIGMAS")
+ RETURN_NAMES = ("high_sigmas", "low_sigmas")
CATEGORY = "sampling/custom_sampling/sigmas"
FUNCTION = "get_sigmas"
@@ -147,6 +149,27 @@ class SplitSigmas:
sigmas2 = sigmas[step:]
return (sigmas1, sigmas2)
+class SplitSigmasDenoise:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"sigmas": ("SIGMAS", ),
+ "denoise": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}),
+ }
+ }
+ RETURN_TYPES = ("SIGMAS","SIGMAS")
+ RETURN_NAMES = ("high_sigmas", "low_sigmas")
+ CATEGORY = "sampling/custom_sampling/sigmas"
+
+ FUNCTION = "get_sigmas"
+
+ def get_sigmas(self, sigmas, denoise):
+ steps = max(sigmas.shape[-1] - 1, 0)
+ total_steps = round(steps * denoise)
+ sigmas1 = sigmas[:-(total_steps)]
+ sigmas2 = sigmas[-(total_steps + 1):]
+ return (sigmas1, sigmas2)
+
class FlipSigmas:
@classmethod
def INPUT_TYPES(s):
@@ -160,6 +183,9 @@ class FlipSigmas:
FUNCTION = "get_sigmas"
def get_sigmas(self, sigmas):
+ if len(sigmas) == 0:
+ return (sigmas,)
+
sigmas = sigmas.flip(0)
if sigmas[0] == 0:
sigmas[0] = 0.0001
@@ -310,6 +336,24 @@ class SamplerDPMAdaptative:
"s_noise":s_noise })
return (sampler, )
+class Noise_EmptyNoise:
+ def __init__(self):
+ self.seed = 0
+
+ def generate_noise(self, input_latent):
+ latent_image = input_latent["samples"]
+ return torch.zeros(latent_image.shape, dtype=latent_image.dtype, layout=latent_image.layout, device="cpu")
+
+
+class Noise_RandomNoise:
+ def __init__(self, seed):
+ self.seed = seed
+
+ def generate_noise(self, input_latent):
+ latent_image = input_latent["samples"]
+ batch_inds = input_latent["batch_index"] if "batch_index" in input_latent else None
+ return comfy.sample.prepare_noise(latent_image, self.seed, batch_inds)
+
class SamplerCustom:
@classmethod
def INPUT_TYPES(s):
@@ -337,10 +381,9 @@ class SamplerCustom:
latent = latent_image
latent_image = latent["samples"]
if not add_noise:
- noise = torch.zeros(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, device="cpu")
+ noise = Noise_EmptyNoise().generate_noise(latent)
else:
- batch_inds = latent["batch_index"] if "batch_index" in latent else None
- noise = comfy.sample.prepare_noise(latent_image, noise_seed, batch_inds)
+ noise = Noise_RandomNoise(noise_seed).generate_noise(latent)
noise_mask = None
if "noise_mask" in latent:
@@ -361,6 +404,207 @@ class SamplerCustom:
out_denoised = out
return (out, out_denoised)
+class Guider_Basic(comfy.samplers.CFGGuider):
+ def set_conds(self, positive):
+ self.inner_set_conds({"positive": positive})
+
+class BasicGuider:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"model": ("MODEL",),
+ "conditioning": ("CONDITIONING", ),
+ }
+ }
+
+ RETURN_TYPES = ("GUIDER",)
+
+ FUNCTION = "get_guider"
+ CATEGORY = "sampling/custom_sampling/guiders"
+
+ def get_guider(self, model, conditioning):
+ guider = Guider_Basic(model)
+ guider.set_conds(conditioning)
+ return (guider,)
+
+class CFGGuider:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"model": ("MODEL",),
+ "positive": ("CONDITIONING", ),
+ "negative": ("CONDITIONING", ),
+ "cfg": ("FLOAT", {"default": 8.0, "min": 0.0, "max": 100.0, "step":0.1, "round": 0.01}),
+ }
+ }
+
+ RETURN_TYPES = ("GUIDER",)
+
+ FUNCTION = "get_guider"
+ CATEGORY = "sampling/custom_sampling/guiders"
+
+ def get_guider(self, model, positive, negative, cfg):
+ guider = comfy.samplers.CFGGuider(model)
+ guider.set_conds(positive, negative)
+ guider.set_cfg(cfg)
+ return (guider,)
+
+class Guider_DualCFG(comfy.samplers.CFGGuider):
+ def set_cfg(self, cfg1, cfg2):
+ self.cfg1 = cfg1
+ self.cfg2 = cfg2
+
+ def set_conds(self, positive, middle, negative):
+ middle = node_helpers.conditioning_set_values(middle, {"prompt_type": "negative"})
+ self.inner_set_conds({"positive": positive, "middle": middle, "negative": negative})
+
+ def predict_noise(self, x, timestep, model_options={}, seed=None):
+ negative_cond = self.conds.get("negative", None)
+ middle_cond = self.conds.get("middle", None)
+
+ out = comfy.samplers.calc_cond_batch(self.inner_model, [negative_cond, middle_cond, self.conds.get("positive", None)], x, timestep, model_options)
+ return comfy.samplers.cfg_function(self.inner_model, out[1], out[0], self.cfg2, x, timestep, model_options=model_options, cond=middle_cond, uncond=negative_cond) + (out[2] - out[1]) * self.cfg1
+
+class DualCFGGuider:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"model": ("MODEL",),
+ "cond1": ("CONDITIONING", ),
+ "cond2": ("CONDITIONING", ),
+ "negative": ("CONDITIONING", ),
+ "cfg_conds": ("FLOAT", {"default": 8.0, "min": 0.0, "max": 100.0, "step":0.1, "round": 0.01}),
+ "cfg_cond2_negative": ("FLOAT", {"default": 8.0, "min": 0.0, "max": 100.0, "step":0.1, "round": 0.01}),
+ }
+ }
+
+ RETURN_TYPES = ("GUIDER",)
+
+ FUNCTION = "get_guider"
+ CATEGORY = "sampling/custom_sampling/guiders"
+
+ def get_guider(self, model, cond1, cond2, negative, cfg_conds, cfg_cond2_negative):
+ guider = Guider_DualCFG(model)
+ guider.set_conds(cond1, cond2, negative)
+ guider.set_cfg(cfg_conds, cfg_cond2_negative)
+ return (guider,)
+
+class DisableNoise:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":{
+ }
+ }
+
+ RETURN_TYPES = ("NOISE",)
+ FUNCTION = "get_noise"
+ CATEGORY = "sampling/custom_sampling/noise"
+
+ def get_noise(self):
+ return (Noise_EmptyNoise(),)
+
+
+class RandomNoise(DisableNoise):
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":{
+ "noise_seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}),
+ }
+ }
+
+ def get_noise(self, noise_seed):
+ return (Noise_RandomNoise(noise_seed),)
+
+
+class SamplerCustomAdvanced:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"noise": ("NOISE", ),
+ "guider": ("GUIDER", ),
+ "sampler": ("SAMPLER", ),
+ "sigmas": ("SIGMAS", ),
+ "latent_image": ("LATENT", ),
+ }
+ }
+
+ RETURN_TYPES = ("LATENT","LATENT")
+ RETURN_NAMES = ("output", "denoised_output")
+
+ FUNCTION = "sample"
+
+ CATEGORY = "sampling/custom_sampling"
+
+ def sample(self, noise, guider, sampler, sigmas, latent_image):
+ latent = latent_image
+ latent_image = latent["samples"]
+
+ noise_mask = None
+ if "noise_mask" in latent:
+ noise_mask = latent["noise_mask"]
+
+ x0_output = {}
+ callback = latent_preview.prepare_callback(guider.model_patcher, sigmas.shape[-1] - 1, x0_output)
+
+ disable_pbar = not comfy.utils.PROGRESS_BAR_ENABLED
+ samples = guider.sample(noise.generate_noise(latent), latent_image, sampler, sigmas, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=noise.seed)
+ samples = samples.to(comfy.model_management.intermediate_device())
+
+ out = latent.copy()
+ out["samples"] = samples
+ if "x0" in x0_output:
+ out_denoised = latent.copy()
+ out_denoised["samples"] = guider.model_patcher.model.process_latent_out(x0_output["x0"].cpu())
+ else:
+ out_denoised = out
+ return (out, out_denoised)
+
+class AddNoise:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"model": ("MODEL",),
+ "noise": ("NOISE", ),
+ "sigmas": ("SIGMAS", ),
+ "latent_image": ("LATENT", ),
+ }
+ }
+
+ RETURN_TYPES = ("LATENT",)
+
+ FUNCTION = "add_noise"
+
+ CATEGORY = "_for_testing/custom_sampling/noise"
+
+ def add_noise(self, model, noise, sigmas, latent_image):
+ if len(sigmas) == 0:
+ return latent_image
+
+ latent = latent_image
+ latent_image = latent["samples"]
+
+ noisy = noise.generate_noise(latent)
+
+ model_sampling = model.get_model_object("model_sampling")
+ process_latent_out = model.get_model_object("process_latent_out")
+ process_latent_in = model.get_model_object("process_latent_in")
+
+ if len(sigmas) > 1:
+ scale = torch.abs(sigmas[0] - sigmas[-1])
+ else:
+ scale = sigmas[0]
+
+ if torch.count_nonzero(latent_image) > 0: #Don't shift the empty latent image.
+ latent_image = process_latent_in(latent_image)
+ noisy = model_sampling.noise_scaling(scale, noisy, latent_image)
+ noisy = process_latent_out(noisy)
+ noisy = torch.nan_to_num(noisy, nan=0.0, posinf=0.0, neginf=0.0)
+
+ out = latent.copy()
+ out["samples"] = noisy
+ return (out,)
+
+
NODE_CLASS_MAPPINGS = {
"SamplerCustom": SamplerCustom,
"BasicScheduler": BasicScheduler,
@@ -377,5 +621,14 @@ NODE_CLASS_MAPPINGS = {
"SamplerDPMPP_SDE": SamplerDPMPP_SDE,
"SamplerDPMAdaptative": SamplerDPMAdaptative,
"SplitSigmas": SplitSigmas,
+ "SplitSigmasDenoise": SplitSigmasDenoise,
"FlipSigmas": FlipSigmas,
+
+ "CFGGuider": CFGGuider,
+ "DualCFGGuider": DualCFGGuider,
+ "BasicGuider": BasicGuider,
+ "RandomNoise": RandomNoise,
+ "DisableNoise": DisableNoise,
+ "AddNoise": AddNoise,
+ "SamplerCustomAdvanced": SamplerCustomAdvanced,
}
diff --git a/comfy_extras/nodes_ip2p.py b/comfy_extras/nodes_ip2p.py
new file mode 100644
index 000000000..c2e70a84c
--- /dev/null
+++ b/comfy_extras/nodes_ip2p.py
@@ -0,0 +1,45 @@
+import torch
+
+class InstructPixToPixConditioning:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required": {"positive": ("CONDITIONING", ),
+ "negative": ("CONDITIONING", ),
+ "vae": ("VAE", ),
+ "pixels": ("IMAGE", ),
+ }}
+
+ RETURN_TYPES = ("CONDITIONING","CONDITIONING","LATENT")
+ RETURN_NAMES = ("positive", "negative", "latent")
+ FUNCTION = "encode"
+
+ CATEGORY = "conditioning/instructpix2pix"
+
+ def encode(self, positive, negative, pixels, vae):
+ x = (pixels.shape[1] // 8) * 8
+ y = (pixels.shape[2] // 8) * 8
+
+ if pixels.shape[1] != x or pixels.shape[2] != y:
+ x_offset = (pixels.shape[1] % 8) // 2
+ y_offset = (pixels.shape[2] % 8) // 2
+ pixels = pixels[:,x_offset:x + x_offset, y_offset:y + y_offset,:]
+
+ concat_latent = vae.encode(pixels)
+
+ out_latent = {}
+ out_latent["samples"] = torch.zeros_like(concat_latent)
+
+ out = []
+ for conditioning in [positive, negative]:
+ c = []
+ for t in conditioning:
+ d = t[1].copy()
+ d["concat_latent_image"] = concat_latent
+ n = [t[0], d]
+ c.append(n)
+ out.append(c)
+ return (out[0], out[1], out_latent)
+
+NODE_CLASS_MAPPINGS = {
+ "InstructPixToPixConditioning": InstructPixToPixConditioning,
+}
diff --git a/comfy_extras/nodes_model_downscale.py b/comfy_extras/nodes_model_downscale.py
index 48bcc6892..58b5073ec 100644
--- a/comfy_extras/nodes_model_downscale.py
+++ b/comfy_extras/nodes_model_downscale.py
@@ -20,8 +20,9 @@ class PatchModelAddDownscale:
CATEGORY = "_for_testing"
def patch(self, model, block_number, downscale_factor, start_percent, end_percent, downscale_after_skip, downscale_method, upscale_method):
- sigma_start = model.model.model_sampling.percent_to_sigma(start_percent)
- sigma_end = model.model.model_sampling.percent_to_sigma(end_percent)
+ model_sampling = model.get_model_object("model_sampling")
+ sigma_start = model_sampling.percent_to_sigma(start_percent)
+ sigma_end = model_sampling.percent_to_sigma(end_percent)
def input_block_patch(h, transformer_options):
if transformer_options["block"][1] == block_number:
diff --git a/comfy_extras/nodes_model_merging.py b/comfy_extras/nodes_model_merging.py
index a25b73ca7..2a431f65d 100644
--- a/comfy_extras/nodes_model_merging.py
+++ b/comfy_extras/nodes_model_merging.py
@@ -2,7 +2,9 @@ import comfy.sd
import comfy.utils
import comfy.model_base
import comfy.model_management
+import comfy.model_sampling
+import torch
import folder_paths
import json
import os
@@ -189,6 +191,13 @@ def save_checkpoint(model, clip=None, vae=None, clip_vision=None, filename_prefi
# "stable-diffusion-v2-768-v", "stable-diffusion-v2-unclip-l", "stable-diffusion-v2-unclip-h",
# "v2-inpainting"
+ extra_keys = {}
+ model_sampling = model.get_model_object("model_sampling")
+ if isinstance(model_sampling, comfy.model_sampling.ModelSamplingContinuousEDM):
+ if isinstance(model_sampling, comfy.model_sampling.V_PREDICTION):
+ extra_keys["edm_vpred.sigma_max"] = torch.tensor(model_sampling.sigma_max).float()
+ extra_keys["edm_vpred.sigma_min"] = torch.tensor(model_sampling.sigma_min).float()
+
if model.model.model_type == comfy.model_base.ModelType.EPS:
metadata["modelspec.predict_key"] = "epsilon"
elif model.model.model_type == comfy.model_base.ModelType.V_PREDICTION:
@@ -203,7 +212,7 @@ def save_checkpoint(model, clip=None, vae=None, clip_vision=None, filename_prefi
output_checkpoint = f"{filename}_{counter:05}_.safetensors"
output_checkpoint = os.path.join(full_output_folder, output_checkpoint)
- comfy.sd.save_checkpoint(output_checkpoint, model, clip, vae, clip_vision, metadata=metadata)
+ comfy.sd.save_checkpoint(output_checkpoint, model, clip, vae, clip_vision, metadata=metadata, extra_keys=extra_keys)
class CheckpointSave:
def __init__(self):
diff --git a/comfy_extras/nodes_model_merging_model_specific.py b/comfy_extras/nodes_model_merging_model_specific.py
new file mode 100644
index 000000000..f2d008d8b
--- /dev/null
+++ b/comfy_extras/nodes_model_merging_model_specific.py
@@ -0,0 +1,60 @@
+import comfy_extras.nodes_model_merging
+
+class ModelMergeSD1(comfy_extras.nodes_model_merging.ModelMergeBlocks):
+ CATEGORY = "advanced/model_merging/model_specific"
+ @classmethod
+ def INPUT_TYPES(s):
+ arg_dict = { "model1": ("MODEL",),
+ "model2": ("MODEL",)}
+
+ argument = ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01})
+
+ arg_dict["time_embed."] = argument
+ arg_dict["label_emb."] = argument
+
+ for i in range(12):
+ arg_dict["input_blocks.{}.".format(i)] = argument
+
+ for i in range(3):
+ arg_dict["middle_block.{}.".format(i)] = argument
+
+ for i in range(12):
+ arg_dict["output_blocks.{}.".format(i)] = argument
+
+ arg_dict["out."] = argument
+
+ return {"required": arg_dict}
+
+
+class ModelMergeSDXL(comfy_extras.nodes_model_merging.ModelMergeBlocks):
+ CATEGORY = "advanced/model_merging/model_specific"
+
+ @classmethod
+ def INPUT_TYPES(s):
+ arg_dict = { "model1": ("MODEL",),
+ "model2": ("MODEL",)}
+
+ argument = ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01})
+
+ arg_dict["time_embed."] = argument
+ arg_dict["label_emb."] = argument
+
+ for i in range(9):
+ arg_dict["input_blocks.{}".format(i)] = argument
+
+ for i in range(3):
+ arg_dict["middle_block.{}".format(i)] = argument
+
+ for i in range(9):
+ arg_dict["output_blocks.{}".format(i)] = argument
+
+ arg_dict["out."] = argument
+
+ return {"required": arg_dict}
+
+
+NODE_CLASS_MAPPINGS = {
+ "ModelMergeSD1": ModelMergeSD1,
+ "ModelMergeSD2": ModelMergeSD1, #SD1 and SD2 have the same blocks
+ "ModelMergeSDXL": ModelMergeSDXL,
+}
diff --git a/comfy_extras/nodes_pag.py b/comfy_extras/nodes_pag.py
new file mode 100644
index 000000000..63f43fd62
--- /dev/null
+++ b/comfy_extras/nodes_pag.py
@@ -0,0 +1,56 @@
+#Modified/simplified version of the node from: https://github.com/pamparamm/sd-perturbed-attention
+#If you want the one with more options see the above repo.
+
+#My modified one here is more basic but has less chances of breaking with ComfyUI updates.
+
+import comfy.model_patcher
+import comfy.samplers
+
+class PerturbedAttentionGuidance:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {
+ "required": {
+ "model": ("MODEL",),
+ "scale": ("FLOAT", {"default": 3.0, "min": 0.0, "max": 100.0, "step": 0.1, "round": 0.01}),
+ }
+ }
+
+ RETURN_TYPES = ("MODEL",)
+ FUNCTION = "patch"
+
+ CATEGORY = "_for_testing"
+
+ def patch(self, model, scale):
+ unet_block = "middle"
+ unet_block_id = 0
+ m = model.clone()
+
+ def perturbed_attention(q, k, v, extra_options, mask=None):
+ return v
+
+ def post_cfg_function(args):
+ model = args["model"]
+ cond_pred = args["cond_denoised"]
+ cond = args["cond"]
+ cfg_result = args["denoised"]
+ sigma = args["sigma"]
+ model_options = args["model_options"].copy()
+ x = args["input"]
+
+ if scale == 0:
+ return cfg_result
+
+ # Replace Self-attention with PAG
+ model_options = comfy.model_patcher.set_model_options_patch_replace(model_options, perturbed_attention, "attn1", unet_block, unet_block_id)
+ (pag,) = comfy.samplers.calc_cond_batch(model, [cond], x, sigma, model_options)
+
+ return cfg_result + (cond_pred - pag) * scale
+
+ m.set_model_sampler_post_cfg_function(post_cfg_function)
+
+ return (m,)
+
+NODE_CLASS_MAPPINGS = {
+ "PerturbedAttentionGuidance": PerturbedAttentionGuidance,
+}
diff --git a/comfy_extras/nodes_perpneg.py b/comfy_extras/nodes_perpneg.py
index 9e8a218f9..306cf9cd0 100644
--- a/comfy_extras/nodes_perpneg.py
+++ b/comfy_extras/nodes_perpneg.py
@@ -1,10 +1,20 @@
import torch
import comfy.model_management
-import comfy.sample
+import comfy.sampler_helpers
import comfy.samplers
import comfy.utils
+import node_helpers
+def perp_neg(x, noise_pred_pos, noise_pred_neg, noise_pred_nocond, neg_scale, cond_scale):
+ pos = noise_pred_pos - noise_pred_nocond
+ neg = noise_pred_neg - noise_pred_nocond
+ perp = neg - ((torch.mul(neg, pos).sum())/(torch.norm(pos)**2)) * pos
+ perp_neg = perp * neg_scale
+ cfg_result = noise_pred_nocond + cond_scale*(pos - perp_neg)
+ return cfg_result
+
+#TODO: This node should be removed, it has been replaced with PerpNegGuider
class PerpNeg:
@classmethod
def INPUT_TYPES(s):
@@ -19,7 +29,7 @@ class PerpNeg:
def patch(self, model, empty_conditioning, neg_scale):
m = model.clone()
- nocond = comfy.sample.convert_cond(empty_conditioning)
+ nocond = comfy.sampler_helpers.convert_cond(empty_conditioning)
def cfg_function(args):
model = args["model"]
@@ -33,12 +43,7 @@ class PerpNeg:
(noise_pred_nocond,) = comfy.samplers.calc_cond_batch(model, [nocond_processed], x, sigma, model_options)
- pos = noise_pred_pos - noise_pred_nocond
- neg = noise_pred_neg - noise_pred_nocond
- perp = neg - ((torch.mul(neg, pos).sum())/(torch.norm(pos)**2)) * pos
- perp_neg = perp * neg_scale
- cfg_result = noise_pred_nocond + cond_scale*(pos - perp_neg)
- cfg_result = x - cfg_result
+ cfg_result = x - perp_neg(x, noise_pred_pos, noise_pred_neg, noise_pred_nocond, neg_scale, cond_scale)
return cfg_result
m.set_model_sampler_cfg_function(cfg_function)
@@ -46,10 +51,52 @@ class PerpNeg:
return (m, )
+class Guider_PerpNeg(comfy.samplers.CFGGuider):
+ def set_conds(self, positive, negative, empty_negative_prompt):
+ empty_negative_prompt = node_helpers.conditioning_set_values(empty_negative_prompt, {"prompt_type": "negative"})
+ self.inner_set_conds({"positive": positive, "empty_negative_prompt": empty_negative_prompt, "negative": negative})
+
+ def set_cfg(self, cfg, neg_scale):
+ self.cfg = cfg
+ self.neg_scale = neg_scale
+
+ def predict_noise(self, x, timestep, model_options={}, seed=None):
+ positive_cond = self.conds.get("positive", None)
+ negative_cond = self.conds.get("negative", None)
+ empty_cond = self.conds.get("empty_negative_prompt", None)
+
+ out = comfy.samplers.calc_cond_batch(self.inner_model, [negative_cond, positive_cond, empty_cond], x, timestep, model_options)
+ return perp_neg(x, out[1], out[0], out[2], self.neg_scale, self.cfg)
+
+class PerpNegGuider:
+ @classmethod
+ def INPUT_TYPES(s):
+ return {"required":
+ {"model": ("MODEL",),
+ "positive": ("CONDITIONING", ),
+ "negative": ("CONDITIONING", ),
+ "empty_conditioning": ("CONDITIONING", ),
+ "cfg": ("FLOAT", {"default": 8.0, "min": 0.0, "max": 100.0, "step":0.1, "round": 0.01}),
+ "neg_scale": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 100.0, "step": 0.01}),
+ }
+ }
+
+ RETURN_TYPES = ("GUIDER",)
+
+ FUNCTION = "get_guider"
+ CATEGORY = "_for_testing"
+
+ def get_guider(self, model, positive, negative, empty_conditioning, cfg, neg_scale):
+ guider = Guider_PerpNeg(model)
+ guider.set_conds(positive, negative, empty_conditioning)
+ guider.set_cfg(cfg, neg_scale)
+ return (guider,)
+
NODE_CLASS_MAPPINGS = {
"PerpNeg": PerpNeg,
+ "PerpNegGuider": PerpNegGuider,
}
NODE_DISPLAY_NAME_MAPPINGS = {
- "PerpNeg": "Perp-Neg",
+ "PerpNeg": "Perp-Neg (DEPRECATED by PerpNegGuider)",
}
diff --git a/comfy_extras/nodes_photomaker.py b/comfy_extras/nodes_photomaker.py
index 90130142b..29d127d74 100644
--- a/comfy_extras/nodes_photomaker.py
+++ b/comfy_extras/nodes_photomaker.py
@@ -141,7 +141,7 @@ class PhotoMakerEncode:
return {"required": { "photomaker": ("PHOTOMAKER",),
"image": ("IMAGE",),
"clip": ("CLIP", ),
- "text": ("STRING", {"multiline": True, "default": "photograph of photomaker"}),
+ "text": ("STRING", {"multiline": True, "dynamicPrompts": True, "default": "photograph of photomaker"}),
}}
RETURN_TYPES = ("CONDITIONING",)
diff --git a/comfy_extras/nodes_post_processing.py b/comfy_extras/nodes_post_processing.py
index 0110b472f..68f6ef51e 100644
--- a/comfy_extras/nodes_post_processing.py
+++ b/comfy_extras/nodes_post_processing.py
@@ -5,6 +5,7 @@ from PIL import Image
import math
import comfy.utils
+import comfy.model_management
class Blend:
@@ -102,6 +103,7 @@ class Blur:
if blur_radius == 0:
return (image,)
+ image = image.to(comfy.model_management.get_torch_device())
batch_size, height, width, channels = image.shape
kernel_size = blur_radius * 2 + 1
@@ -112,7 +114,7 @@ class Blur:
blurred = F.conv2d(padded_image, kernel, padding=kernel_size // 2, groups=channels)[:,:,blur_radius:-blur_radius, blur_radius:-blur_radius]
blurred = blurred.permute(0, 2, 3, 1)
- return (blurred,)
+ return (blurred.to(comfy.model_management.intermediate_device()),)
class Quantize:
def __init__(self):
@@ -225,6 +227,7 @@ class Sharpen:
return (image,)
batch_size, height, width, channels = image.shape
+ image = image.to(comfy.model_management.get_torch_device())
kernel_size = sharpen_radius * 2 + 1
kernel = gaussian_kernel(kernel_size, sigma, device=image.device) * -(alpha*10)
@@ -239,7 +242,7 @@ class Sharpen:
result = torch.clamp(sharpened, 0, 1)
- return (result,)
+ return (result.to(comfy.model_management.intermediate_device()),)
class ImageScaleToTotalPixels:
upscale_methods = ["nearest-exact", "bilinear", "area", "bicubic", "lanczos"]
diff --git a/comfy_extras/nodes_upscale_model.py b/comfy_extras/nodes_upscale_model.py
index 2b5e49a55..52c95df23 100644
--- a/comfy_extras/nodes_upscale_model.py
+++ b/comfy_extras/nodes_upscale_model.py
@@ -37,9 +37,14 @@ class ImageUpscaleWithModel:
def upscale(self, upscale_model, image):
device = model_management.get_torch_device()
+
+ memory_required = model_management.module_size(upscale_model)
+ memory_required += (512 * 512 * 3) * image.element_size() * max(upscale_model.scale, 1.0) * 384.0 #The 384.0 is an estimate of how much some of these models take, TODO: make it more accurate
+ memory_required += image.nelement() * image.element_size()
+ model_management.free_memory(memory_required, device)
+
upscale_model.to(device)
in_img = image.movedim(-1,-3).to(device)
- free_memory = model_management.get_free_memory(device)
tile = 512
overlap = 32
diff --git a/execution.py b/execution.py
index 35bdb77a6..47d58b9d7 100644
--- a/execution.py
+++ b/execution.py
@@ -176,7 +176,7 @@ def recursive_execute(server, prompt, outputs, current_item, extra_data, execute
for node_id, node_outputs in outputs.items():
output_data_formatted[node_id] = [[format_value(x) for x in l] for l in node_outputs]
- logging.error("!!! Exception during processing !!!")
+ logging.error(f"!!! Exception during processing!!! {ex}")
logging.error(traceback.format_exc())
error_details = {
diff --git a/folder_paths.py b/folder_paths.py
index a84524d9f..489795002 100644
--- a/folder_paths.py
+++ b/folder_paths.py
@@ -2,7 +2,7 @@ import os
import time
import logging
-supported_pt_extensions = set(['.ckpt', '.pt', '.bin', '.pth', '.safetensors'])
+supported_pt_extensions = set(['.ckpt', '.pt', '.bin', '.pth', '.safetensors', '.pkl'])
folder_names_and_paths = {}
@@ -181,6 +181,8 @@ def get_full_path(folder_name, filename):
full_path = os.path.join(x, filename)
if os.path.isfile(full_path):
return full_path
+ elif os.path.islink(full_path):
+ logging.warning("WARNING path {} exists but doesn't link anywhere, skipping.".format(full_path))
return None
diff --git a/main.py b/main.py
index b3a3ebea8..a374f2b12 100644
--- a/main.py
+++ b/main.py
@@ -243,11 +243,11 @@ if __name__ == "__main__":
call_on_start = None
if args.auto_launch:
- def startup_server(address, port):
+ def startup_server(scheme, address, port):
import webbrowser
if os.name == 'nt' and address == '0.0.0.0':
address = '127.0.0.1'
- webbrowser.open(f"http://{address}:{port}")
+ webbrowser.open(f"{scheme}://{address}:{port}")
call_on_start = startup_server
try:
diff --git a/node_helpers.py b/node_helpers.py
new file mode 100644
index 000000000..43b9e829f
--- /dev/null
+++ b/node_helpers.py
@@ -0,0 +1,24 @@
+from PIL import ImageFile, UnidentifiedImageError
+
+def conditioning_set_values(conditioning, values={}):
+ c = []
+ for t in conditioning:
+ n = [t[0], t[1].copy()]
+ for k in values:
+ n[1][k] = values[k]
+ c.append(n)
+
+ return c
+
+def pillow(fn, arg):
+ prev_value = None
+ try:
+ x = fn(arg)
+ except (OSError, UnidentifiedImageError, ValueError): #PIL issues #4472 and #2445, also fixes ComfyUI issue #3416
+ prev_value = ImageFile.LOAD_TRUNCATED_IMAGES
+ ImageFile.LOAD_TRUNCATED_IMAGES = True
+ x = fn(arg)
+ finally:
+ if prev_value is not None:
+ ImageFile.LOAD_TRUNCATED_IMAGES = prev_value
+ return x
diff --git a/nodes.py b/nodes.py
index 5cf020df2..488afd577 100644
--- a/nodes.py
+++ b/nodes.py
@@ -10,14 +10,14 @@ import time
import random
import logging
-from PIL import Image, ImageOps, ImageSequence
+from PIL import Image, ImageOps, ImageSequence, ImageFile
from PIL.PngImagePlugin import PngInfo
+
import numpy as np
import safetensors.torch
sys.path.insert(0, os.path.join(os.path.dirname(os.path.realpath(__file__)), "comfy"))
-
import comfy.diffusers_load
import comfy.samplers
import comfy.sample
@@ -34,6 +34,7 @@ import importlib
import folder_paths
import latent_preview
+import node_helpers
def before_node_execution():
comfy.model_management.throw_exception_if_processing_interrupted()
@@ -46,7 +47,7 @@ MAX_RESOLUTION=16384
class CLIPTextEncode:
@classmethod
def INPUT_TYPES(s):
- return {"required": {"text": ("STRING", {"multiline": True}), "clip": ("CLIP", )}}
+ return {"required": {"text": ("STRING", {"multiline": True, "dynamicPrompts": True}), "clip": ("CLIP", )}}
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "encode"
@@ -151,13 +152,9 @@ class ConditioningSetArea:
CATEGORY = "conditioning"
def append(self, conditioning, width, height, x, y, strength):
- c = []
- for t in conditioning:
- n = [t[0], t[1].copy()]
- n[1]['area'] = (height // 8, width // 8, y // 8, x // 8)
- n[1]['strength'] = strength
- n[1]['set_area_to_bounds'] = False
- c.append(n)
+ c = node_helpers.conditioning_set_values(conditioning, {"area": (height // 8, width // 8, y // 8, x // 8),
+ "strength": strength,
+ "set_area_to_bounds": False})
return (c, )
class ConditioningSetAreaPercentage:
@@ -176,13 +173,9 @@ class ConditioningSetAreaPercentage:
CATEGORY = "conditioning"
def append(self, conditioning, width, height, x, y, strength):
- c = []
- for t in conditioning:
- n = [t[0], t[1].copy()]
- n[1]['area'] = ("percentage", height, width, y, x)
- n[1]['strength'] = strength
- n[1]['set_area_to_bounds'] = False
- c.append(n)
+ c = node_helpers.conditioning_set_values(conditioning, {"area": ("percentage", height, width, y, x),
+ "strength": strength,
+ "set_area_to_bounds": False})
return (c, )
class ConditioningSetAreaStrength:
@@ -197,11 +190,7 @@ class ConditioningSetAreaStrength:
CATEGORY = "conditioning"
def append(self, conditioning, strength):
- c = []
- for t in conditioning:
- n = [t[0], t[1].copy()]
- n[1]['strength'] = strength
- c.append(n)
+ c = node_helpers.conditioning_set_values(conditioning, {"strength": strength})
return (c, )
@@ -219,19 +208,15 @@ class ConditioningSetMask:
CATEGORY = "conditioning"
def append(self, conditioning, mask, set_cond_area, strength):
- c = []
set_area_to_bounds = False
if set_cond_area != "default":
set_area_to_bounds = True
if len(mask.shape) < 3:
mask = mask.unsqueeze(0)
- for t in conditioning:
- n = [t[0], t[1].copy()]
- _, h, w = mask.shape
- n[1]['mask'] = mask
- n[1]['set_area_to_bounds'] = set_area_to_bounds
- n[1]['mask_strength'] = strength
- c.append(n)
+
+ c = node_helpers.conditioning_set_values(conditioning, {"mask": mask,
+ "set_area_to_bounds": set_area_to_bounds,
+ "mask_strength": strength})
return (c, )
class ConditioningZeroOut:
@@ -266,13 +251,8 @@ class ConditioningSetTimestepRange:
CATEGORY = "advanced/conditioning"
def set_range(self, conditioning, start, end):
- c = []
- for t in conditioning:
- d = t[1].copy()
- d['start_percent'] = start
- d['end_percent'] = end
- n = [t[0], d]
- c.append(n)
+ c = node_helpers.conditioning_set_values(conditioning, {"start_percent": start,
+ "end_percent": end})
return (c, )
class VAEDecode:
@@ -413,13 +393,8 @@ class InpaintModelConditioning:
out = []
for conditioning in [positive, negative]:
- c = []
- for t in conditioning:
- d = t[1].copy()
- d["concat_latent_image"] = concat_latent
- d["concat_mask"] = mask
- n = [t[0], d]
- c.append(n)
+ c = node_helpers.conditioning_set_values(conditioning, {"concat_latent_image": concat_latent,
+ "concat_mask": mask})
out.append(c)
return (out[0], out[1], out_latent)
@@ -608,8 +583,8 @@ class LoraLoader:
return {"required": { "model": ("MODEL",),
"clip": ("CLIP", ),
"lora_name": (folder_paths.get_filename_list("loras"), ),
- "strength_model": ("FLOAT", {"default": 1.0, "min": -20.0, "max": 20.0, "step": 0.01}),
- "strength_clip": ("FLOAT", {"default": 1.0, "min": -20.0, "max": 20.0, "step": 0.01}),
+ "strength_model": ("FLOAT", {"default": 1.0, "min": -100.0, "max": 100.0, "step": 0.01}),
+ "strength_clip": ("FLOAT", {"default": 1.0, "min": -100.0, "max": 100.0, "step": 0.01}),
}}
RETURN_TYPES = ("MODEL", "CLIP")
FUNCTION = "load_lora"
@@ -642,7 +617,7 @@ class LoraLoaderModelOnly(LoraLoader):
def INPUT_TYPES(s):
return {"required": { "model": ("MODEL",),
"lora_name": (folder_paths.get_filename_list("loras"), ),
- "strength_model": ("FLOAT", {"default": 1.0, "min": -20.0, "max": 20.0, "step": 0.01}),
+ "strength_model": ("FLOAT", {"default": 1.0, "min": -100.0, "max": 100.0, "step": 0.01}),
}}
RETURN_TYPES = ("MODEL",)
FUNCTION = "load_lora_model_only"
@@ -991,7 +966,7 @@ class GLIGENTextBoxApply:
return {"required": {"conditioning_to": ("CONDITIONING", ),
"clip": ("CLIP", ),
"gligen_textbox_model": ("GLIGEN", ),
- "text": ("STRING", {"multiline": True}),
+ "text": ("STRING", {"multiline": True, "dynamicPrompts": True}),
"width": ("INT", {"default": 64, "min": 8, "max": MAX_RESOLUTION, "step": 8}),
"height": ("INT", {"default": 64, "min": 8, "max": MAX_RESOLUTION, "step": 8}),
"x": ("INT", {"default": 0, "min": 0, "max": MAX_RESOLUTION, "step": 8}),
@@ -1481,11 +1456,14 @@ class LoadImage:
FUNCTION = "load_image"
def load_image(self, image):
image_path = folder_paths.get_annotated_filepath(image)
- img = Image.open(image_path)
+
+ img = node_helpers.pillow(Image.open, image_path)
+
output_images = []
output_masks = []
for i in ImageSequence.Iterator(img):
- i = ImageOps.exif_transpose(i)
+ i = node_helpers.pillow(ImageOps.exif_transpose, i)
+
if i.mode == 'I':
i = i.point(lambda i: i * (1 / 255))
image = i.convert("RGB")
@@ -1540,8 +1518,8 @@ class LoadImageMask:
FUNCTION = "load_image"
def load_image(self, image, channel):
image_path = folder_paths.get_annotated_filepath(image)
- i = Image.open(image_path)
- i = ImageOps.exif_transpose(i)
+ i = node_helpers.pillow(Image.open, image_path)
+ i = node_helpers.pillow(ImageOps.exif_transpose, i)
if i.getbands() != ("R", "G", "B", "A"):
if i.mode == 'I':
i = i.point(lambda i: i * (1 / 255))
@@ -1965,6 +1943,12 @@ def init_custom_nodes():
"nodes_morphology.py",
"nodes_stable_cascade.py",
"nodes_differential_diffusion.py",
+ "nodes_ip2p.py",
+ "nodes_model_merging_model_specific.py",
+ "nodes_pag.py",
+ "nodes_align_your_steps.py",
+ "nodes_attention_multiply.py",
+ "nodes_advanced_samplers.py",
]
import_failed = []
diff --git a/server.py b/server.py
index 5642bd5e2..bab3b0600 100644
--- a/server.py
+++ b/server.py
@@ -11,6 +11,7 @@ import urllib
import json
import glob
import struct
+import ssl
from PIL import Image, ImageOps
from PIL.PngImagePlugin import PngInfo
from io import BytesIO
@@ -623,14 +624,22 @@ class PromptServer():
async def start(self, address, port, verbose=True, call_on_start=None):
runner = web.AppRunner(self.app, access_log=None)
await runner.setup()
- site = web.TCPSite(runner, address, port)
+ ssl_ctx = None
+ scheme = "http"
+ if args.tls_keyfile and args.tls_certfile:
+ ssl_ctx = ssl.SSLContext(protocol=ssl.PROTOCOL_TLS_SERVER, verify_mode=ssl.CERT_NONE)
+ ssl_ctx.load_cert_chain(certfile=args.tls_certfile,
+ keyfile=args.tls_keyfile)
+ scheme = "https"
+
+ site = web.TCPSite(runner, address, port, ssl_context=ssl_ctx)
await site.start()
if verbose:
logging.info("Starting server\n")
- logging.info("To see the GUI go to: http://{}:{}".format(address, port))
+ logging.info("To see the GUI go to: {}://{}:{}".format(scheme, address, port))
if call_on_start is not None:
- call_on_start(address, port)
+ call_on_start(scheme, address, port)
def add_on_prompt_handler(self, handler):
self.on_prompt_handlers.append(handler)
diff --git a/tests-ui/utils/ezgraph.js b/tests-ui/utils/ezgraph.js
index 8a55246ee..97be7aa72 100644
--- a/tests-ui/utils/ezgraph.js
+++ b/tests-ui/utils/ezgraph.js
@@ -204,13 +204,17 @@ export class EzWidget {
convertToWidget() {
if (!this.isConvertedToInput)
throw new Error(`Widget ${this.widget.name} cannot be converted as it is already a widget.`);
- this.node.menu[`Convert ${this.widget.name} to widget`].call();
+ var menu = this.node.menu["Convert Input to Widget"].item.submenu.options;
+ var index = menu.findIndex(a => a.content == `Convert ${this.widget.name} to widget`);
+ menu[index].callback.call();
}
convertToInput() {
if (this.isConvertedToInput)
throw new Error(`Widget ${this.widget.name} cannot be converted as it is already an input.`);
- this.node.menu[`Convert ${this.widget.name} to input`].call();
+ var menu = this.node.menu["Convert Widget to Input"].item.submenu.options;
+ var index = menu.findIndex(a => a.content == `Convert ${this.widget.name} to input`);
+ menu[index].callback.call();
}
}
diff --git a/web/extensions/core/colorPalette.js b/web/extensions/core/colorPalette.js
index b8d83613d..02546782f 100644
--- a/web/extensions/core/colorPalette.js
+++ b/web/extensions/core/colorPalette.js
@@ -20,6 +20,10 @@ const colorPalettes = {
"MODEL": "#B39DDB", // light lavender-purple
"STYLE_MODEL": "#C2FFAE", // light green-yellow
"VAE": "#FF6E6E", // bright red
+ "NOISE": "#B0B0B0", // gray
+ "GUIDER": "#66FFFF", // cyan
+ "SAMPLER": "#ECB4B4", // very soft red
+ "SIGMAS": "#CDFFCD", // soft lime green
"TAESD": "#DCC274", // cheesecake
},
"litegraph_base": {
diff --git a/web/extensions/core/dynamicPrompts.js b/web/extensions/core/dynamicPrompts.js
index 599a9e685..7417361ba 100644
--- a/web/extensions/core/dynamicPrompts.js
+++ b/web/extensions/core/dynamicPrompts.js
@@ -17,7 +17,7 @@ app.registerExtension({
// Locate dynamic prompt text widgets
// Include any widgets with dynamicPrompts set to true, and customtext
const widgets = node.widgets.filter(
- (n) => (n.type === "customtext" && n.dynamicPrompts !== false) || n.dynamicPrompts
+ (n) => n.dynamicPrompts
);
for (const widget of widgets) {
// Override the serialization of the value to resolve dynamic prompts for all widgets supporting it in this node
diff --git a/web/extensions/core/maskeditor.js b/web/extensions/core/maskeditor.js
index 4f69ac760..36f7496e7 100644
--- a/web/extensions/core/maskeditor.js
+++ b/web/extensions/core/maskeditor.js
@@ -164,6 +164,41 @@ class MaskEditorDialog extends ComfyDialog {
return divElement;
}
+ createOpacitySlider(self, name, callback) {
+ const divElement = document.createElement('div');
+ divElement.id = "maskeditor-opacity-slider";
+ divElement.style.cssFloat = "left";
+ divElement.style.fontFamily = "sans-serif";
+ divElement.style.marginRight = "4px";
+ divElement.style.color = "var(--input-text)";
+ divElement.style.backgroundColor = "var(--comfy-input-bg)";
+ divElement.style.borderRadius = "8px";
+ divElement.style.borderColor = "var(--border-color)";
+ divElement.style.borderStyle = "solid";
+ divElement.style.fontSize = "15px";
+ divElement.style.height = "21px";
+ divElement.style.padding = "1px 6px";
+ divElement.style.display = "flex";
+ divElement.style.position = "relative";
+ divElement.style.top = "2px";
+ divElement.style.pointerEvents = "auto";
+ self.opacity_slider_input = document.createElement('input');
+ self.opacity_slider_input.setAttribute('type', 'range');
+ self.opacity_slider_input.setAttribute('min', '0.1');
+ self.opacity_slider_input.setAttribute('max', '1.0');
+ self.opacity_slider_input.setAttribute('step', '0.01')
+ self.opacity_slider_input.setAttribute('value', '0.7');
+ const labelElement = document.createElement("label");
+ labelElement.textContent = name;
+
+ divElement.appendChild(labelElement);
+ divElement.appendChild(self.opacity_slider_input);
+
+ self.opacity_slider_input.addEventListener("input", callback);
+
+ return divElement;
+ }
+
setlayout(imgCanvas, maskCanvas) {
const self = this;
@@ -203,6 +238,13 @@ class MaskEditorDialog extends ComfyDialog {
self.updateBrushPreview(self, null, null);
});
+ this.brush_opacity_slider = this.createOpacitySlider(self, "Opacity", (event) => {
+ self.brush_opacity = event.target.value;
+ if (self.brush_color_mode !== "negative") {
+ self.maskCanvas.style.opacity = self.brush_opacity;
+ }
+ });
+
this.colorButton = this.createLeftButton(this.getColorButtonText(), () => {
if (self.brush_color_mode === "black") {
self.brush_color_mode = "white";
@@ -237,6 +279,7 @@ class MaskEditorDialog extends ComfyDialog {
bottom_panel.appendChild(this.saveButton);
bottom_panel.appendChild(cancelButton);
bottom_panel.appendChild(this.brush_size_slider);
+ bottom_panel.appendChild(this.brush_opacity_slider);
bottom_panel.appendChild(this.colorButton);
imgCanvas.style.position = "absolute";
@@ -472,7 +515,7 @@ class MaskEditorDialog extends ComfyDialog {
else {
return {
mixBlendMode: "initial",
- opacity: "0.7",
+ opacity: this.brush_opacity,
};
}
}
@@ -538,6 +581,7 @@ class MaskEditorDialog extends ComfyDialog {
this.maskCtx.putImageData(maskData, 0, 0);
}
+ brush_opacity = 0.7;
brush_size = 10;
brush_color_mode = "black";
drawing_mode = false;
diff --git a/web/extensions/core/widgetInputs.js b/web/extensions/core/widgetInputs.js
index 23f51d812..f1a1d22cd 100644
--- a/web/extensions/core/widgetInputs.js
+++ b/web/extensions/core/widgetInputs.js
@@ -256,8 +256,18 @@ export function mergeIfValid(output, config2, forceUpdate, recreateWidget, confi
return { customConfig };
}
+let useConversionSubmenusSetting;
app.registerExtension({
name: "Comfy.WidgetInputs",
+ init() {
+ useConversionSubmenusSetting = app.ui.settings.addSetting({
+ id: "Comfy.NodeInputConversionSubmenus",
+ name: "Node widget/input conversion sub-menus",
+ tooltip: "In the node context menu, place the entries that convert between input/widget in sub-menus.",
+ type: "boolean",
+ defaultValue: true,
+ });
+ },
async beforeRegisterNodeDef(nodeType, nodeData, app) {
// Add menu options to conver to/from widgets
const origGetExtraMenuOptions = nodeType.prototype.getExtraMenuOptions;
@@ -292,12 +302,31 @@ app.registerExtension({
}
}
}
+
+ //Convert.. main menu
if (toInput.length) {
- options.push(...toInput, null);
+ if (useConversionSubmenusSetting.value) {
+ options.push({
+ content: "Convert Widget to Input",
+ submenu: {
+ options: toInput,
+ },
+ });
+ } else {
+ options.push(...toInput, null);
+ }
}
-
if (toWidget.length) {
- options.push(...toWidget, null);
+ if (useConversionSubmenusSetting.value) {
+ options.push({
+ content: "Convert Input to Widget",
+ submenu: {
+ options: toWidget,
+ },
+ });
+ } else {
+ options.push(...toWidget, null);
+ }
}
}
diff --git a/web/lib/litegraph.core.js b/web/lib/litegraph.core.js
index 4ff05ae81..427a62b59 100644
--- a/web/lib/litegraph.core.js
+++ b/web/lib/litegraph.core.js
@@ -7247,7 +7247,7 @@ LGraphNode.prototype.executeAction = function(action)
//create links
for (var i = 0; i < clipboard_info.links.length; ++i) {
var link_info = clipboard_info.links[i];
- var origin_node;
+ var origin_node = undefined;
var origin_node_relative_id = link_info[0];
if (origin_node_relative_id != null) {
origin_node = nodes[origin_node_relative_id];
diff --git a/web/scripts/app.js b/web/scripts/app.js
index 77f29b8e5..a3105c275 100644
--- a/web/scripts/app.js
+++ b/web/scripts/app.js
@@ -953,6 +953,12 @@ export class ComfyApp {
const origProcessMouseDown = LGraphCanvas.prototype.processMouseDown;
LGraphCanvas.prototype.processMouseDown = function(e) {
+ // prepare for ctrl+shift drag: zoom start
+ if(e.ctrlKey && e.shiftKey && e.buttons) {
+ self.zoom_drag_start = [e.x, e.y, this.ds.scale];
+ return;
+ }
+
const res = origProcessMouseDown.apply(this, arguments);
this.selected_group_moving = false;
@@ -973,6 +979,26 @@ export class ComfyApp {
const origProcessMouseMove = LGraphCanvas.prototype.processMouseMove;
LGraphCanvas.prototype.processMouseMove = function(e) {
+ // handle ctrl+shift drag
+ if(e.ctrlKey && e.shiftKey && self.zoom_drag_start) {
+ // stop canvas zoom action
+ if(!e.buttons) {
+ self.zoom_drag_start = null;
+ return;
+ }
+
+ // calculate delta
+ let deltaY = e.y - self.zoom_drag_start[1];
+ let startScale = self.zoom_drag_start[2];
+
+ let scale = startScale - deltaY/100;
+
+ this.ds.changeScale(scale, [this.ds.element.width/2, this.ds.element.height/2]);
+ this.graph.change();
+
+ return;
+ }
+
const orig_selected_group = this.selected_group;
if (this.selected_group && !this.selected_group_resizing && !this.selected_group_moving) {
@@ -1059,6 +1085,20 @@ export class ComfyApp {
// Trigger onPaste
return true;
}
+
+ if((e.key === '+') && e.altKey) {
+ block_default = true;
+ let scale = this.ds.scale * 1.1;
+ this.ds.changeScale(scale, [this.ds.element.width/2, this.ds.element.height/2]);
+ this.graph.change();
+ }
+
+ if((e.key === '-') && e.altKey) {
+ block_default = true;
+ let scale = this.ds.scale * 1 / 1.1;
+ this.ds.changeScale(scale, [this.ds.element.width/2, this.ds.element.height/2]);
+ this.graph.change();
+ }
}
this.graph.change();
diff --git a/web/scripts/pnginfo.js b/web/scripts/pnginfo.js
index 169609209..7132fb60f 100644
--- a/web/scripts/pnginfo.js
+++ b/web/scripts/pnginfo.js
@@ -170,9 +170,12 @@ export async function importA1111(graph, parameters) {
const opts = parameters
.substr(p)
.split("\n")[1]
- .split(",")
+ .match(new RegExp("\\s*([^:]+:\\s*([^\"\\{].*?|\".*?\"|\\{.*?\\}))\\s*(,|$)", "g"))
.reduce((p, n) => {
const s = n.split(":");
+ if (s[1].endsWith(',')) {
+ s[1] = s[1].substr(0, s[1].length -1);
+ }
p[s[0].trim().toLowerCase()] = s[1].trim();
return p;
}, {});
@@ -191,6 +194,7 @@ export async function importA1111(graph, parameters) {
const vaeLoaderNode = LiteGraph.createNode("VAELoader");
const saveNode = LiteGraph.createNode("SaveImage");
let hrSamplerNode = null;
+ let hrSteps = null;
const ceil64 = (v) => Math.ceil(v / 64) * 64;
@@ -290,6 +294,9 @@ export async function importA1111(graph, parameters) {
model(v) {
setWidgetValue(ckptNode, "ckpt_name", v, true);
},
+ "vae"(v) {
+ setWidgetValue(vaeLoaderNode, "vae_name", v, true);
+ },
"cfg scale"(v) {
setWidgetValue(samplerNode, "cfg", +v);
},
@@ -316,6 +323,7 @@ export async function importA1111(graph, parameters) {
const h = ceil64(+wxh[1]);
const hrUp = popOpt("hires upscale");
const hrSz = popOpt("hires resize");
+ hrSteps = popOpt("hires steps");
let hrMethod = popOpt("hires upscaler");
setWidgetValue(imageNode, "width", w);
@@ -398,7 +406,7 @@ export async function importA1111(graph, parameters) {
}
if (hrSamplerNode) {
- setWidgetValue(hrSamplerNode, "steps", getWidget(samplerNode, "steps").value);
+ setWidgetValue(hrSamplerNode, "steps", hrSteps? +hrSteps : getWidget(samplerNode, "steps").value);
setWidgetValue(hrSamplerNode, "cfg", getWidget(samplerNode, "cfg").value);
setWidgetValue(hrSamplerNode, "scheduler", getWidget(samplerNode, "scheduler").value);
setWidgetValue(hrSamplerNode, "sampler_name", getWidget(samplerNode, "sampler_name").value);
@@ -415,7 +423,7 @@ export async function importA1111(graph, parameters) {
graph.arrange();
- for (const opt of ["model hash", "ensd"]) {
+ for (const opt of ["model hash", "ensd", "version", "vae hash", "ti hashes", "lora hashes", "hashes"]) {
delete opts[opt];
}
diff --git a/web/scripts/ui.js b/web/scripts/ui.js
index 5ca6214eb..d0fa46efb 100644
--- a/web/scripts/ui.js
+++ b/web/scripts/ui.js
@@ -90,12 +90,15 @@ function dragElement(dragEl, settings) {
}).observe(dragEl);
function ensureInBounds() {
- if (dragEl.classList.contains("comfy-menu-manual-pos")) {
+ try {
newPosX = Math.min(document.body.clientWidth - dragEl.clientWidth, Math.max(0, dragEl.offsetLeft));
newPosY = Math.min(document.body.clientHeight - dragEl.clientHeight, Math.max(0, dragEl.offsetTop));
positionElement();
}
+ catch(exception){
+ // robust
+ }
}
function positionElement() {
diff --git a/web/scripts/widgets.js b/web/scripts/widgets.js
index 678b1b8ec..6a6899705 100644
--- a/web/scripts/widgets.js
+++ b/web/scripts/widgets.js
@@ -229,7 +229,11 @@ function createIntWidget(node, inputName, inputData, app, isSeedInput) {
val,
function (v) {
const s = this.options.step / 10;
- this.value = Math.round(v / s) * s;
+ let sh = this.options.min % s;
+ if (isNaN(sh)) {
+ sh = 0;
+ }
+ this.value = Math.round((v - sh) / s) * s + sh;
},
config
),
@@ -307,7 +311,9 @@ export const ComfyWidgets = {
return { widget: node.addWidget(widgetType, inputName, val,
function (v) {
if (config.round) {
- this.value = Math.round(v/config.round)*config.round;
+ this.value = Math.round((v + Number.EPSILON)/config.round)*config.round;
+ if (this.value > config.max) this.value = config.max;
+ if (this.value < config.min) this.value = config.min;
} else {
this.value = v;
}