Reducing Flux Schnell installation requirements?

[StephanosPSteer]

I am far from an expert in AI generated images but I am trying to use Flux Schnell in an app I am creating and I am using a GGUF version which is 6.3GB and it all works fine on my own machine.

The python code is using the diffusers library

ckpt_path = os.path.join(app.root_path, 'models', 'flux1-schnell-Q4_K_S.gguf')

transformer = FluxTransformer2DModel.from_single_file(

ckpt_path,

quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),

torch_dtype=torch.bfloat16,)

if not torch.cuda.is_available():

print("CUDA unavailable, using CPU")

pipe = FluxPipeline.from_pretrained(

os.path.join(app.root_path, 'models', 'FLUX.1-schnell'),

transformer=transformer,

torch_dtype=torch.bfloat16,

)

else:

pipe = FluxPipeline.from_pretrained(

os.path.join(app.root_path, 'models', 'FLUX.1-schnell'),

transformer=transformer,

torch_dtype=torch.bfloat16,

)

pipe.enable_model_cpu_offload()

prompt = prompt #"A drawing of a man walking his dog with a poster on the wall in the background which has text that says 'dog walking prohibited'"

image = pipe(

prompt,

guidance_scale=0.0,

width=1216,

height=704,

num_inference_steps=4,

max_sequence_length=256,

generator=torch.Generator("cpu").manual_seed(0)).images[0]

image.save(image_path)

I think 10GB is ok for a download, however in order for this to work the Flux schnell repo is referred to for various models. I have it locally and that can go up to anything around 100GB. Which seems excessive. Is there a way to reduce the amount or size of files that are used in image generation so any install package is not 100GB plus.