[MODEL REQUEST] FLUX.1-dev

[BrickDesignerNL]

Is your feature request related to a problem? Please describe.
Good model for generating images

Details of model being requested

• Model name: FLUX.1-dev
• Source repo link: https://huggingface.co/black-forest-labs/FLUX.1-dev
• Research paper link [If applicable]: https://blackforestlabs.ai/announcing-black-forest-labs/
• Model use case: integration with https://github.com/lllyasviel/stable-diffusion-webui-forge

Additional context for requested model
Easier to get good results than default SD model.

[mestrona-3]

Hi @BrickDesignerNL, thank you for the feature request! We'll add it to our list of requested models. In the meantime, we'd encourage you to try it out via our BYOM (bring your own model) approach to https://app.aihub.qualcomm.com/jobs/. Please feel free to post in Slack if the job fails.

[BrickDesignerNL]

@mestrona-3 I love to, but the instructions there are limited. I do not see BYOM in the link you provide, I see the LLAMA and Whisper that I earlier 'created' via this repo.
But find it hard to understand what I should change and how in https://github.com/quic/ai-hub-models/tree/main/qai_hub_models/models/stable_diffusion_v2_1_quantized

To make https://app.aihub.qualcomm.com/jobs/?type=compile&ownerKind=user&ownerName=@
work with https://huggingface.co/black-forest-labs/FLUX.1-schnell/tree/main or https://huggingface.co/Comfy-Org/stable-diffusion-3.5-fp8 to convert it to int8 and make it work with https://github.com/quic/wos-ai-plugins/tree/main/plugins/gimp/stable-diffusion

Or even directly let safetensor files run on the NPU.
Is there a way to convert them to int8 for the NPU and keep them in the safetensor format (so not binary)?
I see someone did:

https://huggingface.co/Disty0/FLUX.1-dev-qint8

Would be nice to understand how to get it integrated into https://github.com/quic/wos-ai-plugins/tree/main/plugins/gimp/stable-diffusion and let it run on the NPU.