Download pre-configured setup files to quick launch an NVIDIA TAO Jupyter Notebook on Azure Machine Learning with the appropriate resources (Compute Cluster and Environment) and fine-tune a TAO pretrained model, the set up is done using the NGC-AzureML Quick Launch Toolkit. azureml-ngc-tools is a CLI toolkit that enables you to launch a fully setup Jupyter notebook environment on AzureML with a combination of NGC assets pre-installed.
These ready-to-use config files work seamlessly with the NGC-AzureML Quick Launch CLI Toolkit (azureml-ngc-tools) to provide you direct access to a fully setup and optimally configured Azure Machine Learning environment with the TAO SDK container and the additional resources required for this example, pre-installed for you.
The two ready-to-use files are:
azure_config.json: Defines the most optimal Azure environment setup for this use-case. Simply insert your Azure credentials in the indicated fields and use right away.
ngc_config.json: Pre-installs the TAO SDK, and several TAO Notebooks that the user could run on AzureML to learn more about TAO. It includes the Object Detection using TAO DetectNet_v2 Notebook that produces a fine tuned model ready to be deployed as an AzureML Triton Inference Endpoint. No modifications are required; use this file as is.
After inserting your Azure credentials in the azure_config.json file downloaded from here, quick launch a ready-to-use Jupyter notebook on AzureML with the following bash commands:
pip install azureml-ngc-tools azureml-ngc-tools --login azure_config.json --app ngc_config.json
The ready-to-use config files, the NGC-AzureML Quick Launch Toolkit (azureml-ngc-tools) and example code was developed by NVIDIA.
The config files listed here when used with the NGC-AzureML Quick Launch CLI toolkit (azureml-ngc-tools) uses the TAO SDK which is to be used in accordance with the End User License Agreement included with it. Licenses are also available along with the model application zip file. By pulling and using the TAO SDK container and downloading models, you accept the terms and conditions of these licenses.