An example serverless ludwig ML model for determining which movie a particular excerpt is from.
- specify your data in data.csv
- change the example model ludwig definition in ludwig-model.yml
ludwig experiment --data_csv data.csv --model_definition_file ludwig-model.yml
If you change the ludwig model, you will need to update the python code to reflect your features and outputs.
Deploy to GCP with:
- Move your ludwig model to ./model
- Ignore irrelevant files using .gcloudignore (see below)
gcloud config set project my-projectgcloud beta functions deploy predict --runtime python37 --memory=1024MB --trigger-http
Due to GCP size limitations, you may need to remove or ignore unnecessary ludwig model files before deploying to GCP. You should only need to keep the following files:
- model_hyperparameters.json
- model_weights.data-00000-of-00001
- model_weights.index
- model_weights.meta
- train_set_metadata.json
With thanks to Gilbert Tanner's post about hosting a ludwig model using Flask.