Functions

In addition to fal Apps, fal also supports defining serverless functions that can be called from your code just like any other Python function, but they are executed on a remote machine.

Functions are useful for one-off jobs that have a natural end, such as copying data to the /data volume or training a model.

import fal


@fal.function(machine_type="GPU", requirements=["diffusers", "torch", "transformers"])
def generate_image(prompt: str):
    import torch
    from diffusers import FluxPipeline
    from io import BytesIO
    pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=torch.bfloat16)
    pipe.enable_model_cpu_offload()

    out = pipe(
        prompt=prompt,
        guidance_scale=0.,
        height=768,
        width=1360,
        num_inference_steps=4,
        max_sequence_length=256,
    ).images[0]

    buffer = BytesIO()
    out.save(buffer, format="PNG")
    image_bytes = buffer.getvalue()
    return image_bytes

image_bytes = generate_image("A cat holding a sign that says hello world")
with open("image.png", "wb") as f:
    f.write(image_bytes)