pipeLoaderSDXL v1 (Legacy)¶
Documentation¶
- Class name:
ttN pipeLoaderSDXL - Category:
🌏 tinyterra/legacy - Output node:
False
The ttN pipeLoaderSDXL node is designed for loading and initializing the state of a pipeline specifically tailored for SDXL models within the tinyterra ecosystem. It plays a crucial role in setting up the pipeline's initial state, managing data flow, and ensuring that the pipeline is ready for subsequent operations, such as sampling and image generation.
Input types¶
Required¶
ckpt_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
vae_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
lora1_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
lora1_model_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
lora1_clip_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
lora2_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
lora2_model_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
lora2_clip_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
refiner_ckpt_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
refiner_vae_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
refiner_lora1_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
refiner_lora1_model_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
refiner_lora1_clip_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
refiner_lora2_name- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
refiner_lora2_model_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
refiner_lora2_clip_strength- unknown
- Comfy dtype:
FLOAT - Python dtype:
unknown
clip_skip- unknown
- Comfy dtype:
INT - Python dtype:
unknown
positive- unknown
- Comfy dtype:
STRING - Python dtype:
unknown
positive_token_normalization- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
positive_weight_interpretation- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
negative- unknown
- Comfy dtype:
STRING - Python dtype:
unknown
negative_token_normalization- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
negative_weight_interpretation- unknown
- Comfy dtype:
COMBO[STRING] - Python dtype:
unknown
empty_latent_width- unknown
- Comfy dtype:
INT - Python dtype:
unknown
empty_latent_height- unknown
- Comfy dtype:
INT - Python dtype:
unknown
batch_size- unknown
- Comfy dtype:
INT - Python dtype:
unknown
seed- unknown
- Comfy dtype:
INT - Python dtype:
unknown
Output types¶
sdxl_pipe- Comfy dtype:
PIPE_LINE_SDXL - unknown
- Python dtype:
unknown
- Comfy dtype:
model- Comfy dtype:
MODEL - unknown
- Python dtype:
unknown
- Comfy dtype:
positive- Comfy dtype:
CONDITIONING - unknown
- Python dtype:
unknown
- Comfy dtype:
negative- Comfy dtype:
CONDITIONING - unknown
- Python dtype:
unknown
- Comfy dtype:
vae- Comfy dtype:
VAE - unknown
- Python dtype:
unknown
- Comfy dtype:
clip- Comfy dtype:
CLIP - unknown
- Python dtype:
unknown
- Comfy dtype:
refiner_model- Comfy dtype:
MODEL - unknown
- Python dtype:
unknown
- Comfy dtype:
refiner_positive- Comfy dtype:
CONDITIONING - unknown
- Python dtype:
unknown
- Comfy dtype:
refiner_negative- Comfy dtype:
CONDITIONING - unknown
- Python dtype:
unknown
- Comfy dtype:
refiner_vae- Comfy dtype:
VAE - unknown
- Python dtype:
unknown
- Comfy dtype:
refiner_clip- Comfy dtype:
CLIP - unknown
- Python dtype:
unknown
- Comfy dtype:
latent- Comfy dtype:
LATENT - unknown
- Python dtype:
unknown
- Comfy dtype:
seed- Comfy dtype:
INT - unknown
- Python dtype:
unknown
- Comfy dtype:
ui- Provides a user interface component, typically for displaying images generated by the pipeline, depending on the 'image_output' parameter settings.
Usage tips¶
- Infra type:
GPU - Common nodes: unknown
Source code¶
class ttN_pipeLoaderSDXL:
version = '1.1.2'
@classmethod
def INPUT_TYPES(cls):
return {"required": {
"ckpt_name": (folder_paths.get_filename_list("checkpoints"), ),
"vae_name": (["Baked VAE"] + folder_paths.get_filename_list("vae"),),
"lora1_name": (["None"] + folder_paths.get_filename_list("loras"),),
"lora1_model_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"lora1_clip_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"lora2_name": (["None"] + folder_paths.get_filename_list("loras"),),
"lora2_model_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"lora2_clip_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"refiner_ckpt_name": (["None"] + folder_paths.get_filename_list("checkpoints"), ),
"refiner_vae_name": (["Baked VAE"] + folder_paths.get_filename_list("vae"),),
"refiner_lora1_name": (["None"] + folder_paths.get_filename_list("loras"),),
"refiner_lora1_model_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"refiner_lora1_clip_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"refiner_lora2_name": (["None"] + folder_paths.get_filename_list("loras"),),
"refiner_lora2_model_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"refiner_lora2_clip_strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"clip_skip": ("INT", {"default": -2, "min": -24, "max": 0, "step": 1}),
"positive": ("STRING", {"default": "Positive","multiline": True}),
"positive_token_normalization": (["none", "mean", "length", "length+mean"],),
"positive_weight_interpretation": (["comfy", "A1111", "compel", "comfy++", "down_weight"],),
"negative": ("STRING", {"default": "Negative", "multiline": True}),
"negative_token_normalization": (["none", "mean", "length", "length+mean"],),
"negative_weight_interpretation": (["comfy", "A1111", "compel", "comfy++", "down_weight"],),
"empty_latent_width": ("INT", {"default": 1024, "min": 64, "max": MAX_RESOLUTION, "step": 8}),
"empty_latent_height": ("INT", {"default": 1024, "min": 64, "max": MAX_RESOLUTION, "step": 8}),
"batch_size": ("INT", {"default": 1, "min": 1, "max": 64}),
"seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}),
},
"hidden": {"prompt": "PROMPT", "ttNnodeVersion": ttN_pipeLoaderSDXL.version, "my_unique_id": "UNIQUE_ID"}}
RETURN_TYPES = ("PIPE_LINE_SDXL" ,"MODEL", "CONDITIONING", "CONDITIONING", "VAE", "CLIP", "MODEL", "CONDITIONING", "CONDITIONING", "VAE", "CLIP", "LATENT", "INT",)
RETURN_NAMES = ("sdxl_pipe","model", "positive", "negative", "vae", "clip", "refiner_model", "refiner_positive", "refiner_negative", "refiner_vae", "refiner_clip", "latent", "seed",)
FUNCTION = "adv_pipeloader"
CATEGORY = "🌏 tinyterra/legacy"
def adv_pipeloader(self, ckpt_name, vae_name,
lora1_name, lora1_model_strength, lora1_clip_strength,
lora2_name, lora2_model_strength, lora2_clip_strength,
refiner_ckpt_name, refiner_vae_name,
refiner_lora1_name, refiner_lora1_model_strength, refiner_lora1_clip_strength,
refiner_lora2_name, refiner_lora2_model_strength, refiner_lora2_clip_strength,
clip_skip,
positive, positive_token_normalization, positive_weight_interpretation,
negative, negative_token_normalization, negative_weight_interpretation,
empty_latent_width, empty_latent_height, batch_size, seed, prompt=None, my_unique_id=None):
def SDXL_loader(ckpt_name, vae_name,
lora1_name, lora1_model_strength, lora1_clip_strength,
lora2_name, lora2_model_strength, lora2_clip_strength,
positive, positive_token_normalization, positive_weight_interpretation,
negative, negative_token_normalization, negative_weight_interpretation,):
model: ModelPatcher | None = None
clip: CLIP | None = None
vae: VAE | None = None
# Load models
model, clip, vae = loader.load_checkpoint(ckpt_name)
if lora1_name != "None":
model, clip = loader.load_lora(lora1_name, model, clip, lora1_model_strength, lora1_clip_strength)
if lora2_name != "None":
model, clip = loader.load_lora(lora2_name, model, clip, lora2_model_strength, lora2_clip_strength)
# Check for custom VAE
if vae_name not in ["Baked VAE", "Baked-VAE"]:
vae = loader.load_vae(vae_name)
# CLIP skip
if not clip:
raise Exception("No CLIP found")
clipped = clip.clone()
if clip_skip != 0:
clipped.clip_layer(clip_skip)
positive = loader.nsp_parse(positive, seed, title="pipeLoaderSDXL positive", my_unique_id=my_unique_id)
positive_embeddings_final, positive_pooled = advanced_encode(clipped, positive, positive_token_normalization, positive_weight_interpretation, w_max=1.0, apply_to_pooled='enable')
positive_embeddings_final = [[positive_embeddings_final, {"pooled_output": positive_pooled}]]
negative = loader.nsp_parse(negative, seed)
negative_embeddings_final, negative_pooled = advanced_encode(clipped, negative, negative_token_normalization, negative_weight_interpretation, w_max=1.0, apply_to_pooled='enable')
negative_embeddings_final = [[negative_embeddings_final, {"pooled_output": negative_pooled}]]
return model, positive_embeddings_final, negative_embeddings_final, vae, clip
# Create Empty Latent
latent = sampler.emptyLatent(None, batch_size, empty_latent_width, empty_latent_height)
samples = {"samples":latent}
model, positive_embeddings, negative_embeddings, vae, clip = SDXL_loader(ckpt_name, vae_name,
lora1_name, lora1_model_strength, lora1_clip_strength,
lora2_name, lora2_model_strength, lora2_clip_strength,
positive, positive_token_normalization, positive_weight_interpretation,
negative, negative_token_normalization, negative_weight_interpretation)
if refiner_ckpt_name != "None":
refiner_model, refiner_positive_embeddings, refiner_negative_embeddings, refiner_vae, refiner_clip = SDXL_loader(refiner_ckpt_name, refiner_vae_name,
refiner_lora1_name, refiner_lora1_model_strength, refiner_lora1_clip_strength,
refiner_lora2_name, refiner_lora2_model_strength, refiner_lora2_clip_strength,
positive, positive_token_normalization, positive_weight_interpretation,
negative, negative_token_normalization, negative_weight_interpretation)
else:
refiner_model, refiner_positive_embeddings, refiner_negative_embeddings, refiner_vae, refiner_clip = None, None, None, None, None
image = ttNsampler.pil2tensor(Image.new('RGB', (1, 1), (0, 0, 0)))
pipe = {"model": model,
"positive": positive_embeddings,
"negative": negative_embeddings,
"vae": vae,
"clip": clip,
"refiner_model": refiner_model,
"refiner_positive": refiner_positive_embeddings,
"refiner_negative": refiner_negative_embeddings,
"refiner_vae": refiner_vae,
"refiner_clip": refiner_clip,
"samples": samples,
"images": image,
"seed": seed,
"loader_settings": {"ckpt_name": ckpt_name,
"vae_name": vae_name,
"lora1_name": lora1_name,
"lora1_model_strength": lora1_model_strength,
"lora1_clip_strength": lora1_clip_strength,
"lora2_name": lora2_name,
"lora2_model_strength": lora2_model_strength,
"lora2_clip_strength": lora2_clip_strength,
"lora3_name": None,
"lora3_model_strength": None,
"lora3_clip_strength": None,
"refiner_ckpt_name": refiner_ckpt_name,
"refiner_vae_name": refiner_vae_name,
"refiner_lora1_name": refiner_lora1_name,
"refiner_lora1_model_strength": refiner_lora1_model_strength,
"refiner_lora1_clip_strength": refiner_lora1_clip_strength,
"refiner_lora2_name": refiner_lora2_name,
"refiner_lora2_model_strength": refiner_lora2_model_strength,
"refiner_lora2_clip_strength": refiner_lora2_clip_strength,
"clip_skip": clip_skip,
"positive_balance": None,
"positive": positive,
"positive_l": None,
"positive_g": None,
"positive_token_normalization": positive_token_normalization,
"positive_weight_interpretation": positive_weight_interpretation,
"negative_balance": None,
"negative": negative,
"negative_l": None,
"negative_g": None,
"negative_token_normalization": negative_token_normalization,
"negative_weight_interpretation": negative_weight_interpretation,
"empty_latent_width": empty_latent_width,
"empty_latent_height": empty_latent_height,
"batch_size": batch_size,
"seed": seed,
"empty_samples": samples,}
}
return (pipe, model, positive_embeddings, negative_embeddings, vae, clip, refiner_model, refiner_positive_embeddings, refiner_negative_embeddings, refiner_vae, refiner_clip, samples, seed)