resnet-hptuning

Introduction

This sample shows how to run hyperparameter tuning jobs on AI Platform with Cloud TPUs using TensorFlow's tf.metrics.

This sample is adapted from the official samples for training ResNet-50 with Cloud TPUs to run on AI Platform.

Requirements

• Install the Google Cloud Platform SDK. The SDK includes the commandline tools gcloud for submitting training jobs to AI Platform.

• Enable Cloud Storage.

• Authorize Cloud TPU to access your project.

Steps

1. Clone the repository.

   git clone https://github.com/GoogleCloudPlatform/cloudml-samples.git
   

2. If you do not already have a Cloud Storage bucket, create one to be used for the training job.

   gsutil mb gs://[YOUR_GCS_BUCKET]
   export GCS_BUCKET="gs://[YOUR_GCS_BUCKET]"
   

3. Run the sample. The included script will train ResNet-50 for 1024 steps using a fake dataset.

   cd cloudml-samples/tpu/hptuning/resnet-hptuning
   bash submit_resnet_hptuning.sh