Launching a Habana DLAMI

The latest DLAMI is ready to use with Habana Gaudi accelerators. Use the following steps to launch your Habana DLAMI and ensure that your Python and framework-specific resources are active. For additional setup resources, see the Habana Gaudi Setup and Installation respository.

Select a Habana DLAMI

Launch a DL1 instance with the Habana DLAMI of your choice.

For step-by-step instructions on launching a DLAMI, see Launching and Configuring a DLAMI.

For a list of the most recent Habana DLAMIs, see the Release Notes for DLAMI.

Activate Python Environment

Connect to your DL1 instance and activate the recommended Python environment for your Habana DLAMI. To check your recommended Python environment, select your DLAMI in the Release Notes.

Import Machine Learning Framework

Instances with Habana accelerators are pre-integrated with popular machine learning frameworks such as TensorFlow and PyTorch. Import the machine learning framework of your choice.

Import TensorFlow

To use TensorFlow on your Habana DLAMI, navigate to the folder of the Python environment that you activated and import TensorFlow.

/usr/bin/$PYTHON_VERSION
import tensorflow
tensorflow.__version__

To check the TensorFlow version compatible with your Habana DLAMI, select your DLAMI in the Release Notes.

Import PyTorch

To use PyTorch on your Habana DLAMI, navigate to the folder of the Python environment that you activated and import the appropriate PyTorch version.

/usr/bin/$PYTHON_VERSION
import torch
torch.__version__

To check the PyTorch version compatible with your Habana DLAMI, select your DLAMI in the Release Notes.

For more information on how to run and train machine learning models in TensorFlow and PyTorch using your Habana DLAMI, see the Habana Model References GitHub repository. For additional resources on working with your Habana DLAMI, visit the Habana Gaudi documentation.