ComfyUI Node
Hy-Nomas Load Decoder
Loads the PixelUMM video decoder (models/vae). Weights under NVIDIA's One-Way Noncommercial license.
Hy-Nomas Load Decoder
- decoder
◄decoder_name▾►
◄matmulfp8►
CategoryHy-Nomas
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| decoder_name | COMBO | e.g. diffusion_pytorch_model_fp8.safetensors (VAE/PixelUMM_FT_FP8) | |
| matmul | COMBO | fp8 | fp8: FP8 tensor cores (RTX 40). dequant: bf16 matmuls on the dequantized weights, closest to the trained decoder. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| decoder | HYNOMAS_DECODER | — |