Experimenting with Flux.1 Dev GGUF Quants and Text Preserving CLIP-L Encoder

1. Flux.1 Dev Q8 GGUF quantization with standard CLIP-L text encoder

2. Flux.1 Dev Q8 GGUF quantization with ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only text encoder

3. Flux.1 Dev Q6 K GGUF quantization with standard CLIP-L text encoder

4. Flux.1 Dev Q6 K GGUF quantization with ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only text encoder

5. Flux.1 Dev Q4 KS GGUF quantization with standard CLIP-L text encoder

6. Flux.1 Dev Q4 KS GGUF quantization with ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only text encoder

7. Flux.1 Dev Q2 K GGUF quantization with standard CLIP-L text encoder

8. Flux.1 Dev Q2 K GGUF quantization with ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only text encoder

9. Flux.1 Dev Q5 KS GGUF quantization with ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only text encoder

10. Flux.1 Dev Q3 KS GGUF quantization with ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only text encoder

There is a full write up on Medium.