Nodes/Foundation-1/Foundation-1 Model Loader
ComfyUI Node

Foundation-1 Model Loader

Loads a Foundation-1 model from a local .safetensors checkpoint. Auto-downloads from HuggingFace on first run if no model is present. The model is cached and only reloads when the checkpoint, device, or attention type changes.

By Saganaki22·Created 5 months ago·Updated 5 months ago· 105
Foundation-1 Model Loader
    • model
    model
    attentionauto
    Categoryaudio/Foundation-1

    Inputs (2)

    NameTypeDefaultDescription
    modelCOMBOFoundation-1 .safetensors checkpoint. If no model is found it will be downloaded automatically from huggingface.co/RoyalCities/Foundation-1 on first run.
    attentionCOMBOauto'auto' uses SageAttention if installed, otherwise SDPA. 'sdpa' enables all PyTorch SDPA backends (uses Flash internally if available). 'flash_attention_2' forces the Flash SDP backend only (CUDA only). 'sageattention' monkey-patches F.sdpa with SageAttention (CUDA only — requires: pip install sageattention). Changing this setting unloads and reloads the model.

    Outputs (1)

    NameTypeDescription
    modelFOUNDATION1_MODEL