This is the direct GGUF conversion of black forest-labs/FLUX.1-dev. The binary format can be used with low VRAM, theoretically even with 6GB. ... Reposted from city96: https://huggingface.co/city96/FLUX.1-dev-gguf

If you attempt to convert it yourself, please refer to the GGUF documentation: https://github.com/intel/neural-speed/blob/main/neural_speed/convert/convert_chatglm.py#L148

As this is a quantized model, not fine-tuned, all original model restrictions/licensing terms still apply.

The model files need to be used together with the ComfyUI-GGUF custom nodes.

Place the model files in ComfyUI/models/unet - see the GitHub readme for further installation instructions  https://github.com/city96/ComfyUI-GGUF.

Refer to this chart for a basic overview of quantization types: https://github.com/ggerganov/llama.cpp/blob/master/examples/perplexity/README.md#llama-3-8b-scoreboard.