Introduction to Private Serverless Models

As mentioned earlier, each fal app runs in an isolated environment that gets voided right after each request (unless keep_alive is set). But for certain use cases, it may be important to persist certain results after the run is over. In such scenarios, you can use the /data volume, which is mounted on each machine and is shared across all your app is running at any point in time linked to your fal account.

import fal
from pathlib import Path

DATA_DIR = Path("/data/mnist")

```py {2,11,21,26}
class FalModel(
    fal.App,
    requirements=["torch>=2.0.0", "torchvision"],

  ):
    machine_type = "GPU"


    @fal.endpoint("/")
    def text_to_image(self, input: Input) -> Output:
    import torch
    from torchvision import datasets

    already_present = DATA_DIR.exists()
    if already_present:
        print("Test data is already downloaded, skipping download!")

    test_data = datasets.FashionMNIST(
        root=DATA_DIR,
        train=False,
        download=not already_present,
    )
    ...

When you invoke this app for the first time, you will notice that Torch downloads the test dataset. However, subsequent invocations - even those not covered by the invocation’s keep_alive - will skip the download and proceed directly to your logic.