philschmid

from autogluon import ObjectDetection as task

epochs = 20
detector = task.fit(dataset_train,
                    num_trials=3,
                    epochs=epochs,
                    lr=ag.Categorical(5e-4, 1e-4, 3e-4),
                                        ngpus_per_trial=1)

philschmid

# download the data
root = './'
filename_zip = ag.download('https://philschmid-datasets.s3.eu-central-1.amazonaws.com/tiny_fruit.zip',
                        path=root)
# unzip data
filename = ag.unzip(filename_zip, root=root)

# create Dataset
data_root = os.path.join(root, filename)
# train dataset

philschmid

test_map = detector.evaluate(dataset_test)
print(f"The mAP on the test dataset is {test_map[1][1]}")

philschmid

image = 'mixed_10.jpg'
image_path = os.path.join(data_root, 'JPEGImages', image)

ind, prob, loc = detector.predict(image_path)

philschmid

savefile = 'model.pkl'
detector.save(savefile)

philschmid

from autogluon import Detector
new_detector = Detector.load(savefile)

image = 'mixed_17.jpg'
image_path = os.path.join(data_root, 'JPEGImages', image)

detector.predict(image_path)

philschmid

serverless create --template aws-python3 --path scale-machine-learning-w-pytorch 

philschmid

import json


def hello(event, context):
    body = {
        "message": "Go Serverless v1.0! Your function executed successfully!",
        "input":event
    }
    response = {
        "statusCode": 200,

philschmid

serverless plugin install -n serverless-python-requirements

philschmid

https://download.pytorch.org/whl/cpu/torch-1.5.0%2Bcpu-cp38-cp38-linux_x86_64.whl
torchvision==0.6.0
requests_toolbelt