Nodes/SK-ComfyUI-FolderingLoader/[SK] Checkpoint (Config, Deprecated)
ComfyUI Node

[SK] Checkpoint (Config, Deprecated)

The old yaml-config loader, kept around but not the one you want

By SKFRMSEHF·Created 8 months ago·Updated 8 months ago· 0
[SK] Checkpoint (Config, Deprecated)
    • MODEL
    • CLIP
    • VAE
    config_name
    ckpt

    The author labeled this one "Deprecated" right in its display name, and that's the honest, correct call - you almost certainly don't need this node. It's a checkpoint loader that also takes an explicit .yaml config file, a holdover from the days when ComfyUI (following A1111's lead) needed an external config to tell it a checkpoint's architecture details before it could load correctly. Modern checkpoints - SD1.5 through everything after - self-describe that information in their own file metadata, so the config step became unnecessary years ago. ComfyUI's own core ships the equivalent node for the same reason: legacy compatibility, not a recommendation.

    It's tucked under SK Loader/Advanced, one of eight loaders in SK-ComfyUI-FolderingLoader, a small pack by GitHub author SKFRMSEHF that rebuilds ComfyUI's stock loaders with a folder-tree file picker instead of a flat dropdown. No README shipped with it and there's no discussion of the pack anywhere online, so this is read straight off the node's own schema plus what's generally true about config-based checkpoint loading.

    How it works

    Same underlying operation as any checkpoint loader - split the archive into MODEL, CLIP, and VAE - but with a second input, config_name, that points at a .yaml file describing the model's architecture (things like which UNet variant, which conditioning setup). On genuinely old checkpoints from the SD1.x era that shipped without embedded architecture metadata, that config was how the loader knew what it was looking at. Every checkpoint you're likely to download today carries that information in the file itself, which is exactly why this path got deprecated: there's nothing left for the config to tell the loader that the file doesn't already say.

    The inputs that matter

    • config_name - the .yaml config file, picked from a combo. Unless you're specifically working with an old checkpoint that shipped one, you won't have a reason to touch this.
    • ckpt - the checkpoint file itself, picked from the folder-tree combo, same as every other loader in the pack.

    Outputs are MODEL, CLIP, VAE - identical shape to the non-deprecated SK_CheckpointLoaderSimple.

    Installing it

    No install instructions shipped with the pack, so it's the standard route: ComfyUI Manager → search SK-ComfyUI-FolderingLoader or the class name → install → restart, or manually cd ComfyUI/custom_nodes && git clone https://github.com/SKFRMSEHF/comfyui_SK_Loader then restart. No models bundled - this node only becomes useful at all if you happen to already have an old checkpoint-plus-config pair on disk.

    Common issues

    • You almost never need this node. If you're not sure why you'd want a config file, you don't - use SK_CheckpointLoaderSimple (or ComfyUI's own CheckpointLoaderSimple) instead. This one exists purely so a specific class of ancient checkpoint still loads.
    • Config/checkpoint mismatch. If you are using this deliberately, the config has to match the checkpoint's actual architecture - pairing the wrong yaml with the wrong file produces broken or nonsensical output, and there's no validation catching that mismatch for you.
    • Prefer safetensors over .ckpt if your combo offers both for the same model. .ckpt is Python's pickle format and can execute code on load; safetensors carries no such risk and there's zero quality difference.
    • Empty combos. Both config_name and ckpt only list what's already on disk at ComfyUI startup - drop the files in the right folders and restart rather than expecting a live scan.
    CategorySK Loader/Advanced

    Inputs (2)

    NameTypeDefaultDescription
    config_nameCOMBO11 options: anything_v3.yaml, v1-inference.yaml, v1-inference_clip_skip_2.yaml, v1-inference_clip_skip_2_fp16.yaml, v1-inference_fp16.yaml, v1-inpainting-inference.yaml, +5
    ckptCOMBOSelect file [[SK_TREE::[]]]

    Outputs (3)

    NameTypeDescription
    MODELMODEL
    CLIPCLIP
    VAEVAE