Skip to content

Latest commit

 

History

History
40 lines (27 loc) · 1.83 KB

aml_configuration.md

File metadata and controls

40 lines (27 loc) · 1.83 KB
Raw

Azure ML Configuration

This requires the following steps:

  1. Configure AML workspace
  2. Create an Azure Service Principal
  3. Upload the data to the default datastore of your workspace

Configure AML workspace

First step is to attach to an AML workspace.

For you convenience, we recommend you start by moving the file config/config_sample.json to config.json (in root of repo). All you need to fill out is your subscription id. You can then execute the file create_workspace.py to create your workspace. Do make sure to pay attention to the output when running the script, as it may include further instructions or error messages.

See documentation for more info.

Create Azure Service Principal

This is necessary for non-interactive authentication. Create a service principal and give it Contributor access to your workspace (see documentation).

Store the information in a config.json file in the root directory of this repository.

Once you have this info, you can add it to your config.json file by adding these three lines:

"service_principal_id": "",
"service_principal_password": "",
"tenant_id": "",

Upload the data to the default datastore of your workspace

We upload the training data so that it can be mounted as remote drives in the aml compute targets. You can use the method upload_data in utils.py for that.

For example:

conda activate prednet
python
from utils import upload_data
upload_data()