Repository navigation
[Quantization] Add support for Comfy Quants backend #14705
Description
Add support for the Comfy Quant quantization toolkit
Supported formats: FP8 (E4M3/E5M2), INT8 (W8A8, tensorwise), MXFP8, NVFP4, INT4 (SVDQuant W4A4, AWQ W4A16).
comfy-quants is export-only. The inference needs to run with comfy-kitchen
Proposed approach
Add a ComfyQuantConfig / ComfyQuantizer that uses comfy-kitchen to wrap weights as a QuantizedTensor with the appropriate layout.
from diffusers import FluxTransformer2DModel, ComfyQuantConfig
config = ComfyQuantConfig(compute_dtype=torch.bfloat16)
model = FluxTransformer2DModel.from_single_file(
"path/to/comfy_quant_checkpoint.safetensors",
quantization_config=config,
)Related
• comfy-quants https://github.com/Comfy-Org/comfy-quants
• comfy-kitchen https://github.com/Comfy-Org/comfy-kitchen
Metadata
Metadata
Assignees
Labels
No labels
Type
Projects
- Status
Activity
PrakshaaleJain commentedon Sep 4, 2026
Hi @DN6, can I take this work up?
DN6 commentedon Sep 7, 2026
@PrakshaaleJain Sure thing 👍🏽 Assigning to you.
PrakshaaleJain commentedon Sep 9, 2026
Hi @DN6, I have created a PR regarding the issue, please have a review of it.