Azure Machine Learning Examples
Blogposts:
Examples
Simple Example
Train
python run_on_azureml.py \
--experiment test-exp \
--script simple_example/run.py \
--message Howdy!
Deploy
python deploy_on_azureml.py \
--run_id test-exp_1618105880_601b57cd \
--experiment_name test-exp \
--model_artifact_path logs/message.txt \
--score_file simple_example/score.py \
--model_name message-model \
--service_name message-service
Predict on Endpoint
python simple_example/score.py --endpoint <YOUR ENDPOINT>
Keras
Train
python run_on_azureml.py \
--experiment test-exp \
--script keras_mnist_example/train.py \
--max_epochs 10 \
--batch_size 64
Deploy
In the Azure ML Portal, navigate to the run you're happy with, and copy its Run ID. Note this is not just an int like "
Run 5", but a longer unique identifier. Pass this to the --run_id flag of deploy_on_azure.py. Your Run ID should
look similar to the one below, but will not be the same.
python deploy_on_azureml.py \
--run_id test-exp_1618103328_5e55046d \
--experiment_name test-exp \
--model_artifact_path logs/saved_model \
--score_file keras_mnist_example/score.py \
--model_name test-keras-model \
--service_name test-keras-service
Predict from Endpoint
python keras_mnist_example/score.py --endpoint <YOUR ENDPOINT>
PyTorch Lightning Example
Train
python run_on_azureml.py \
--experiment test-exp \
--script lightning_mnist_example/train.py \
--max_epochs 10 \
--batch_size 64 \
--default_root_dir logs/
Deploy
python deploy_on_azureml.py \
--run_id test-exp_1618104960_e34305b3 \
--experiment_name test-exp \
--model_artifact_path logs/lightning_logs/version_0/checkpoints/epoch=9-step=8599.ckpt \
--score_file lightning_mnist_example/score.py \
--model_name test-lit-model \
--service_name test-lit-service
Predict from Endpoint
python lightning_mnist_example/score.py --endpoint <YOUR ENDPOINT>