--- title: orbitquant_unets canonical_url: "https://www.modelscope.cn/models/ApacheOne/orbitquant_unets" md_url: "https://www.modelscope.cn/models/ApacheOne/orbitquant_unets.md" repository: ApacheOne/orbitquant_unets last_updated: 2026-07-11 license: other pipeline_tag: text-to-image-synthesis tasks: - text-to-image-synthesis parameters: 30.7B tensor_type: - I64 - I8 - I32 - BF16 - F32 - U8 - F16 library_name: - safetensors - pytorch frameworks: - pytorch downloads: 148 stars: 0 tags: - sdxl - unet - orbitquant - quantization - w4a4 - w3a4 - w2a4 - custom-runtime-required --- # orbitquant_unets > orbitquant_unets - ApacheOne 在 ModelScope 开源的模型。USE CONVROT INT4 and INT8 for now ApacheOne/orbitquant_unets 是 ModelScope 魔搭社区上的 30.7B 参数text-to-image-synthesis模型，采用 other 许可。 - **Repository**: ApacheOne/orbitquant_unets - **License**: other - **Tasks**: text-to-image-synthesis - **Parameters**: 30.7B - **Tags**: sdxl, unet, orbitquant, quantization, w4a4, w3a4, w2a4, custom-runtime-required - **Downloads**: 148 - **Stars**: 0 - **Last updated**: 2026-07-11 Source: https://www.modelscope.cn/models/ApacheOne/orbitquant_unets --- # USE CONVROT INT4 and INT8 for now # OrbitQuant SDXL Exports Experimental OrbitQuant and ConvRot quantized exports for CreaPrompt Lightning Hyper-SDXL. > **Work in progress:** The native A4 runtime is not public yet. These models cannot currently be used without the custom runtime. A ComfyUI loader node that outputs a model compatible with the core KSampler is still under development. ## W2A4 Recovery Status WIP for W2A4 block-level rollout recovery. ## Model Format These are not standard SDXL checkpoints that can be loaded directly by existing SDXL loaders. Depending on the export, the files can contain: * OrbitQuant-packed SDXL UNet weights * ConvRot-quantized SDXL UNet weights * A4 activation-quantization metadata * Quantized CLIP and text-encoder weights * FP16 VAE weights * `__orbitquant__` metadata * Codebooks * Rotations * Quantization scales * Runtime configuration A custom runtime is required to execute the packed OrbitQuant and ConvRot weights without expanding them back into ordinary FP16 weights. # Examples ## Full OrbitQuant versus Full ConvRot

Full OrbitQuant W4A4 versus Full ConvRot W4A4 — CLIP INT4

## Full-UNet Tests

Full ConvRot W4A4 UNet — CLIP INT4

Full OrbitQuant W4A4 UNet — CLIP INT4

## Mixed OrbitQuant and ConvRot UNet — CLIP FP16

OrbitQuant W4A4 + ConvRot W4A4 — CLIP FP16

OrbitQuant W4A4 + ConvRot W8A8 — CLIP FP16

## Mixed OrbitQuant and ConvRot UNet — CLIP INT4

OrbitQuant W4A4 + ConvRot W4A4 — CLIP INT4

OrbitQuant W4A4 + ConvRot W8A8 — CLIP INT4

## Mixed OrbitQuant and ConvRot UNet — CLIP INT8

OrbitQuant W4A4 + ConvRot W8A8 — CLIP INT8

First Mixed-Quant Model Test — CLIP INT8

## Root Example

OrbitQuant W4A4 + ConvRot W8A8 — CLIP INT4

## Runtime Availability The following components are still required for normal use: * Native packed OrbitQuant weight execution * Native A4 activation quantization * Native ConvRot quantized operators * ComfyUI model loader * Integration with the core ComfyUI KSampler * Runtime validation * Performance tuning Until the custom runtime is released, these files should be considered experimental development exports.