Flash in Production¶

Flash Serve¶

Flash Serve makes model deployment simple.

Server Side¶

from flash.image import SemanticSegmentation
from flash.image.segmentation.output import SegmentationLabelsOutput

model = SemanticSegmentation.load_from_checkpoint(
    "https://flash-weights.s3.amazonaws.com/0.9.0/semantic_segmentation_model.pt"
)
model.output = SegmentationLabelsOutput(visualize=False)
model.serve()

Client Side¶

import base64
from pathlib import Path

import flash
import requests

with (Path(flash.ASSETS_ROOT) / "road.png").open("rb") as f:
    imgstr = base64.b64encode(f.read()).decode("UTF-8")

body = {"session": "UUID", "payload": {"inputs": {"data": imgstr}}}
resp = requests.post("http://127.0.0.1:8000/predict", json=body)
print(resp.json())

Credits to @rlizzo, @hhsecond, @lantiga, @luiscape for building the Flash Serve Engine. Read all about it here.