October 2018
Beginner
362 pages
9h 32m
English
When we deploy a TensorFlow SavedModel to the GCP platform, we either need to upload the entire SavedModel directory to a storage location on GCP or train in the cloud as we did previously. Regardless of what method you main use, your TensorFlow model's binaries should be stored in a Google Cloud Storage location.
To start our deployment process, we first need to create a deployed model object. You can create it with the command as follows:
gcloud ml-engine models create "deployed_classifier"
Next, we'll create an environment variable that will let GCP know where our saved model binaries are:
DEPLOYMENT_SOURCE="gs://classifier_bucket123/classifier_model/binaries" ...
Read now
Unlock full access