Qwen Image Distill — INT8 ConvRot
INT8 + ConvRot quantization of qwen_image_distill_full_bf16.safetensors.
Quantization
- Format: INT8 + ConvRot
- ConvRot group size: 256
- Quantization: per-channel INT8
- Clipping: MSE-optimal
- Quantized layers: 840
- Quantized parameters: ~20.38B
- Non-eligible layers: kept in BF16
- Quantizer: Comfy-Org
comfy-model-tools
Original model
This is a quantized version of the original Qwen Image Distill BF16 model.
ComfyUI
Place the .safetensors file in:
ComfyUI/models/diffusion_models/
Then select it as the diffusion model.
Disclaimer
This repository contains a quantized derivative of the original model. Refer to the original model's license and terms for usage restrictions.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for sverinn/Qwen-Image_ComfyUI-int8-ConvRot
Base model
DiffSynth-Studio/Qwen-Image-Distill-Full Finetuned
Comfy-Org/Qwen-Image_ComfyUI