Nodes/ComfyUI_agilly1989_motorway/CheckpointLoader_motorway_edition
ComfyUI Node

CheckpointLoader_motorway_edition

For legacy workflows only

By agilly1989·Created 2 years ago·Updated about a year ago· 7
CheckpointLoader_motorway_edition
  • MOTORWAY 🚌💨
  • MOTORWAY 🚌💨
config_name
ckpt_name
OUTPUT_MODEL_keyMODEL
OUTPUT_CLIP_keyCLIP
OUTPUT_VAE_keyVAE

Here's a piece of ComfyUI archaeology: way back, loading a checkpoint meant picking both the model and a YAML config file that described its architecture. CheckpointLoader is that old dual-picker node, and it still exists in core ComfyUI under advanced/loaders. This _motorway_edition wraps it for agilly1989's Motorway pipe - which means, in a twist, this is one of the few clones where nothing comes out of the Motorway on the input side: you pick the model and config, and the loaded MODEL, CLIP and VAE go into the Motorway under keys you name.

How it works

CheckpointLoader does what every loader does - reads a .safetensors/.ckpt from your models/checkpoints folder and splits it into the three artifacts the pipeline actually uses: the denoising MODEL, the text encoder CLIP, and the VAE. The inputs that matter:

  • config_name - the dropdown of 11 YAML configs (v1-inference.yaml, v2-inference-v.yaml, etc.). This is the legacy bit: you're telling ComfyUI which architecture the file is.
  • ckpt_name - your checkpoint, dynamically populated from the models folder.
  • OUTPUT_MODEL_key - key for the model, default MODEL.
  • OUTPUT_CLIP_key - key for the CLIP, default CLIP.
  • OUTPUT_VAE_key - key for the VAE, default VAE.

The wrapper instantiates the real core node, runs it with your config + checkpoint choices, then writes each returned artifact into the Motorway under your chosen keys. The only visible output is the MOTORWAY 🚌💨. Downstream, CLIPTextEncode_motorway_edition reads the CLIP key, the sampler's INPUT_model_key reads MODEL, and a VAE decode reads VAE - the defaults are chosen so that works without you touching anything.

The catch

For 99% of workflows, CheckpointLoaderSimple_motorway_edition is the one you want. It guesses the config from the file and has a single dropdown. This node exists for the cases where the guess is wrong - usually very old SD 1.5 or SD 2.x-era checkpoints that genuinely need a specific YAML, or workflows that came from that era and refuse to die. If you don't know why you need to pick a config, you probably don't.

Like its siblings with no connection inputs, the MOTORWAY 🚌💨 here is marked optional in the schema. Ignore that. The wrapper still writes outputs and clones the Motorway, so feed it a live one from a MotorwayStart or it dies on 'NoneType' object has no attribute 'update'.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway

Restart, or find "ComfyUI_agilly1989_motorway" in ComfyUI Manager. No pip dependencies, no extra downloads - the pack is just the Motorway class and generated wrappers.

Fine print

The pack is an explicitly-beta side project (README opens "BIG BROKEN WITH ASYNC WILL FIX WHEN I GET THE TIME/MOTIVATION"), and the current hotfix release has the _motorway_edition clones commented out of __init__.py - so on the newest main these nodes may not load at all, which is exactly what you'll see if an old workflow references them and they come up missing. Keys overwrite as they flow down the Motorway, and a key that was never written throws the signature 'MotorwayClass' object has no attribute 'hash_' error. The author takes bug reports on GitHub; Discord DMs are, by his own words, ignored.

Categoryagilly1989 Nodes/Motorway-ed/advanced/loaders

Inputs (6)

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
ckpt_nameCOMBO0 options:
OUTPUT_MODEL_keySTRINGMODEL
OUTPUT_CLIP_keySTRINGCLIP
OUTPUT_VAE_keySTRINGVAE
MOTORWAY 🚌💨optMOTORWAY 🚌💨

Outputs (1)

NameTypeDescription
MOTORWAY 🚌💨MOTORWAY 🚌💨