Zoni Qwen360 2512 INT8 deployment bundle
This repository is a deployment-only aggregation of four public model assets used by Zoni's locked Qwen360 ComfyUI graph. The files are copied server-side from their exact source revisions; they are not modified or re-quantized.
RunPod currently permits one cached Hugging Face repository per endpoint. This bundle keeps every large safetensors file outside the Docker image so cold starts do not need to pull roughly 31 GB of model data from the image registry.
| Bundle path | Source repository and revision | Bytes | SHA-256 |
|---|---|---|---|
diffusion_models/qwen-image-2512-int8-tensorwise.safetensors | LAXMAYDAY/Qwen-Image-2512-int8-tensorwise@26e54e4418a2bde66862cc396e2c9ba5fe8471da | 20,555,635,384 | a3c68480c89ea1ff81524230b10ff68e6ccc3cedbe1bb983148797cb919c0e7d |
text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors | Comfy-Org/Qwen-Image_ComfyUI@46839d338df81ce625d5fae27d7e370314c0fbc9 | 9,384,670,680 | cb5636d852a0ea6a9075ab1bef496c0db7aef13c02350571e388aea959c5c0b4 |
vae/qwen_image_vae.safetensors | Comfy-Org/Qwen-Image_ComfyUI@46839d338df81ce625d5fae27d7e370314c0fbc9 | 253,806,246 | a70580f0213e67967ee9c95f05bb400e8fb08307e017a924bf3441223e023d1f |
loras/qwen-360-diffusion-2512-int8-bf16-v2.safetensors | ProGamerGov/qwen-360-diffusion@f63e5de60bd27b7d9af9c64f8c591381775ea9aa | 755,039,336 | 11065421c2a4c8f503ee1be5d6dc0320a9ffb5bfaded2490813046fd2312578d |
The base Qwen and Comfy assets are Apache-2.0 licensed. The Qwen360 LoRA is MIT licensed. Refer to each source repository for its full model card, license, training details, and usage conditions.
Zoni's worker independently checks the Hugging Face content-addressed blob identity and exact byte size of all four files before ComfyUI starts. A missing or changed file causes the worker to fail closed; it never downloads a replacement at runtime.