Kroma v0.3 INT8 ConvRot

INT8 ConvRot quantization of lodestones/Kroma, made from kroma-v0.3-base.safetensors.

Quantization

Format: ComfyUI-compatible INT8
Scaling: per-row
Rotation: ConvRot
ConvRot group size: 256
Quantized 2D weights: 224
Sensitive/skipped weights: 40
Output tensors: 878
File size: 14,135,109,560 bytes
SHA256: 887e22bbbab4a39d8689f61c91e87cdd97d72ef219b9a2aa5b952d13c48746c0

Created with silveroxides/convert_to_quant v1.3.3 using INT8, row scaling, ConvRot group size 256, the krea2 filter, Comfy quant metadata, simple mode, and low-memory mode.

Use with a current ComfyUI build that supports Comfy quantized safetensors. This repository contains converted model weights only. The original Kroma model license and usage terms continue to apply; see the source repository.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Hiccup1234/Kroma-v0.3-INT8-ConvRot

Base model

lodestones/Kroma

Finetuned

(2)

this model