monster/Qwen-Image-Layered-MRP-MLX
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add small README note on pre-quantized models on Huggingface

Browse Source
This commit is contained in:
filipstrand 2025-12-17 21:20:22 +01:00
parent 0ade9a504c
commit d70c8edb8c
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
README.md
View File

@ -861,9 +861,13 @@ mflux-generate-fibo \
--width 1024 \
--height 1024 \
--steps 20 \
--seed 42 \
--seed 42
```
⚠️ * Note: As of MFLUX v.0.13, some internal changes have been made which breaks compatibility with older pre-quantized models.
Newer ones will be uploaded, but in the meantime, you can always save a new quantized version from the original weights using the [mflux-save](#-saving-a-quantized-version-to-disk) command.*
To save disk space, you can delete the original full 16-bit model from the Huggingface cache after saving the quantized version.
---
### 💽 Running a model directly from disk