Skip to content

unCLIPCheckpointLoader

Documentation

  • Class name: unCLIPCheckpointLoader
  • Category: loaders
  • Output node: False

The unCLIPCheckpointLoader node is designed for loading checkpoints specifically tailored for unCLIP models. It facilitates the retrieval and initialization of models, CLIP vision modules, and VAEs from a specified checkpoint, streamlining the setup process for further operations or analyses.

Input types

Required

  • ckpt_name
    • The 'ckpt_name' parameter specifies the name of the checkpoint to be loaded. It is crucial for identifying and retrieving the correct checkpoint file from a predefined directory of checkpoints, thereby determining the models and configurations to be initialized.
    • Comfy dtype: COMBO[STRING]
    • Python dtype: str

Output types

  • model
    • Comfy dtype: MODEL
    • Represents the primary model loaded from the checkpoint.
    • Python dtype: torch.nn.Module
  • clip
    • Comfy dtype: CLIP
    • Represents the CLIP module loaded from the checkpoint, if available.
    • Python dtype: torch.nn.Module
  • vae
    • Comfy dtype: VAE
    • Represents the VAE module loaded from the checkpoint, if available.
    • Python dtype: torch.nn.Module
  • clip_vision
    • Comfy dtype: CLIP_VISION
    • Represents the CLIP vision module loaded from the checkpoint, if available.
    • Python dtype: torch.nn.Module

Usage tips

  • Infra type: GPU
  • Common nodes: unknown

Source code

class unCLIPCheckpointLoader:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "ckpt_name": (folder_paths.get_filename_list("checkpoints"), ),
                             }}
    RETURN_TYPES = ("MODEL", "CLIP", "VAE", "CLIP_VISION")
    FUNCTION = "load_checkpoint"

    CATEGORY = "loaders"

    def load_checkpoint(self, ckpt_name, output_vae=True, output_clip=True):
        ckpt_path = folder_paths.get_full_path("checkpoints", ckpt_name)
        out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, output_clipvision=True, embedding_directory=folder_paths.get_folder_paths("embeddings"))
        return out