batch-shipyard/recipes/Caffe2-GPU

Caffe2-GPU

This recipe shows how to run Caffe2 on a single GPU N-series VM.

Configuration

Please see refer to this set of sample configuration files for this recipe.

Pool Configuration

The pool configuration should enable the following properties:

• vm_size must be one of STANDARD_NC6, STANDARD_NC12, STANDARD_NC24, STANDARD_NV6, STANDARD_NV12, STANDARD_NV24. NC VM instances feature K80 GPUs for GPU compute acceleration while NV VM instances feature M60 GPUs for visualization workloads. Because Caffe2 is a GPU-accelerated compute application, it is best to choose NC VM instances.
• vm_configuration is the VM configuration
  • platform_image specifies to use a platform image
    • publisher should be Canonical or OpenLogic.
    • offer should be UbuntuServer for Canonical or CentOS for OpenLogic.
    • sku should be 16.04-LTS for Ubuntu or 7.3 for CentOS.

Other pool properties such as publisher, offer, sku, vm_size and vm_count should be set to your desired values.

Global Configuration

The global configuration should set the following properties:

• docker_images array must have a reference to a valid Caffe2 GPU-enabled Docker image. The official Caffe2 Docker images can be used for this recipe. The Docker image caffe2ai/caffe2 may be used.

Jobs Configuration

The jobs configuration should set the following properties within the tasks array which should have a task definition containing:

• image should be the name of the Docker image for this container invocation, e.g., caffe2ai/caffe2
• resource_files array should be populated if you want Azure Batch to handle the download of the training file from the web endpoint:
  • file_path is the local file path which should be set to mnist.py
  • blob_source is the remote URL of the file to retrieve: https://raw.githubusercontent.com/Azure/batch-shipyard/master/recipes/Caffe2-CPU/scripts/mnist.py
• command should contain the command to pass to the Docker run invocation. For the caffe2ai/caffe2 Docker image and the sample script above, the command would be: python -u mnist.py --gpu
• gpu must be set to true. This enables invoking the nvidia-docker wrapper.