Training RetinaNet on Cloud TPU

This document describes an implementation of the RetinaNet object detection model. The code is available on GitHub.

The instructions below assume you are already familiar with running a model on Cloud TPU. If you are new to Cloud TPU, you can refer to the Quickstart for a basic introduction.

If you plan to train on a TPU Pod slice, review Training on TPU Pods to understand parameter changes required for Pod slices.

Warning: This tutorial uses a third-party dataset. Google provides no representation, warranty, or other guarantees about the validity, or any other aspects of this dataset.

Objectives

• Create a Cloud Storage bucket to hold your dataset and model output
• Prepare the COCO dataset
• Set up a Compute Engine VM and Cloud TPU node for training and evaluation
• Run training and evaluation on a single Cloud TPU or a Cloud TPU Pod

Costs

This tutorial uses billable components of Google Cloud, including:

• Compute Engine
• Cloud TPU
• Cloud Storage

Use the pricing calculator to generate a cost estimate based on your projected usage. New Google Cloud users might be eligible for a free trial.

Before you begin

This section provides information on setting up Cloud Storage bucket and a Compute Engine VM.

Important: Set up your Compute Engine VM, your Cloud TPU node and your Cloud Storage bucket in the same region/zone to reduce network latency and network costs.

1. Open a Cloud Shell window.

   Open Cloud Shell

2. Create a variable for your project's ID.

   export PROJECT_ID=project-id
   

3. Configure the gcloud command-line tool to use the project where you want to create Cloud TPU.

   gcloud config set project ${PROJECT_ID}
   

   The first time you run this command in a new Cloud Shell VM, an Authorize Cloud Shell page is displayed. Click Authorize at the bottom of the page to allow gcloud to make GCP API calls with your credentials.

4. Create a Service Account for the Cloud TPU project.

   gcloud beta services identity create --service tpu.googleapis.com --project $PROJECT_ID
   

   The command returns a Cloud TPU Service Account with following format:

   [e-mail]
   

5. Create a Cloud Storage bucket using the following command:

   Note: In the following command, replace bucket-name with the name you want to assign to your bucket.

   gsutil mb -p ${PROJECT_ID} -c standard -l europe-west4 -b on gs://bucket-name
   

   This Cloud Storage bucket stores the data you use to train your model and the training results. The ctpu up tool used in this tutorial sets up default permissions for the Cloud TPU Service Account. If you want finer-grain permissions, review the access level permissions.

   The bucket location must be in the same region as your virtual machine (VM) and your TPU node. VMs and TPU nodes are located in specific zones, which are subdivisions within a region.

6. Launch a Compute Engine VM using the ctpu up command.

   Note: This tutorial uses TensorFlow 1.15.5 which is based on the Debian 10 image and defaults to Python 3.6. Earlier versions of TensorFlow 1.15.x were based on Debian 9 and default to Python 3.5.

   $ ctpu up --project=${PROJECT_ID} \
    --zone=europe-west4-a \
    --vm-only \
    --disk-size-gb=300 \
    --machine-type=n1-standard-8 \
    --tf-version=1.15.5 \
    --name=retinanet-tutorial
   

   Command flag descriptions

   project

   Your GCP project ID

   zone

   The zone where you plan to create your Cloud TPU.

   vm-only

   Creates the VM without creating a Cloud TPU. By default the ctpu up command creates a VM and a Cloud TPU.

   disk-size-gb

   The size of the hard disk in GB of the VM created by the ctpu up command.

   machine-type

   The machine type of the Compute Engine VM to create.

   tf-version

   The version of Tensorflow ctpu installs on the VM.

   name

   The name of the Cloud TPU to create.

7. The configuration you specified appears. Enter y to approve or n to cancel.

8. When the ctpu up command has finished executing, verify that your shell prompt has changed from username@projectname to username@vm-name. This change shows that you are now logged into your Compute Engine VM.

   Note: The first time you run ctpu up on a project it may take several minutes to perform startup tasks such as SSH key propagation and API turnup. Occasionally, during the SSH key propagation, an error message (ERROR: (gcloud.compute.ssh) [/usr/bin/ssh] exited with return code [255]) is returned. If that happens, run the following command to resolve the problem.

   gcloud compute ssh retinanet-tutorial --zone=europe-west4-a
   

   As you continue these instructions, run each command that begins with (vm)$ in your VM session window.

   When the ctpu command launches a Compute Engine virtual machine (VM), it automatically places the RetinaNet model files from TensorFlow branch in the /usr/share/tpu/models/official/detection/ directory.

9. Use the export command to set these environment variables.

   Note: In the following commands, replace bucket-name with the name of your bucket.

   (vm)$ export STORAGE_BUCKET=gs://bucket-name
   

   (vm)$ export TPU_NAME=retinanet-tutorial
   (vm)$ export DATA_DIR=${STORAGE_BUCKET}/coco
   

10. Install extra packages

    The RetinaNet training application requires several extra packages. Install them now:

    (vm)$ sudo apt-get install -y python3-tk
    

    (vm)$ pip3 install --user Cython matplotlib opencv-python-headless pyyaml Pillow
    

    (vm)$ pip3 install --user 'git+https://github.com/cocodataset/cocoapi#egg=pycocotools&subdirectory=PythonAPI'
    

    (vm)$ pip3 install --user -U gast==0.2.2
    

Prepare the COCO dataset

1. Run the download_and_preprocess_coco.sh script to convert the COCO dataset into a set of TFRecords (*.tfrecord) that the training application expects.

   (vm)$ sudo bash /usr/share/tpu/tools/datasets/download_and_preprocess_coco.sh ./data/dir/coco
   

   This installs the required libraries and then runs the preprocessing script. It outputs a number of *.tfrecord files in your local data directory.

   Note: The COCO download and conversion script takes approximately 1 hour to complete.

2. After you convert the data into TFRecords, copy them from local storage to your Cloud Storage bucket using the gsutil command. You must also copy the annotation files. These files help validate the model's performance:

   (vm)$ gsutil -m cp ./data/dir/coco/*.tfrecord ${DATA_DIR}
   

   (vm)$ gsutil cp ./data/dir/coco/raw-data/annotations/*.json ${DATA_DIR}
   

Set up the training environment

1. Run the following command to create your Cloud TPU.

   (vm)$ ctpu up --project=${PROJECT_ID} \
     --tpu-only \
     --tf-version=1.15.5 \
     --name=retinanet-tutorial
   

   Command flag descriptions

   project

   Your GCP project ID

   tpu-only

   Create a Cloud TPU only. By default the ctpu up command creates a VM and a Cloud TPU.

   tf-version

   The version of Tensorflow ctpu installs on the VM.

   name

   The name of the Cloud TPU to create.

2. The configuration you specified appears. Enter y to approve or n to cancel.

   You will see a message: Operation success; not ssh-ing to Compute Engine VM due to --tpu-only flag. Since you previously completed SSH key propagation, you can ignore this message.

3. Update the keepalive values for your VM connection.

   This tutorial requires a long-lived connection to the Compute Engine instance. To ensure you aren't disconnected from the instance, run the following command:

   (vm)$ sudo /sbin/sysctl \
     -w net.ipv4.tcp_keepalive_time=120 \
     net.ipv4.tcp_keepalive_intvl=120 \
     net.ipv4.tcp_keepalive_probes=5
   

4. You are now ready to run the model on the preprocessed COCO data. First, add the top-level /models folder to the Python path with the command:

   (vm)$ export PYTHONPATH=${PYTHONPATH}:/usr/share/tpu/models
   

Training and evaluation require TensorFlow 1.13 or a later version.

Note: If you want to monitor the model's output and performance, follow the guide to setting up TensorBoard.

Single Cloud TPU device training

Note: This script trains for 22,500 steps and takes approximately 3 hours to train on a Cloud TPU v2-8 and approximately 1 1/2 hours to train on a Cloud TPU v3-8.

1. Set up the following environment variables:

   (vm)$ export MODEL_DIR=${STORAGE_BUCKET}/retinanet-model-train
   (vm)$ export RESNET_CHECKPOINT=gs://cloud-tpu-checkpoints/retinanet/resnet50-checkpoint-2018-02-07
   (vm)$ export TRAIN_FILE_PATTERN=${DATA_DIR}/train-*
   (vm)$ export EVAL_FILE_PATTERN=${DATA_DIR}/val-*
   (vm)$ export VAL_JSON_FILE=${DATA_DIR}/instances_val2017.json
   

2. Run the training script:

   (vm)$ python3 /usr/share/tpu/models/official/detection/main.py \
   --use_tpu=True \
   --tpu=${TPU_NAME} \
   --num_cores=8 \
   --model_dir=${MODEL_DIR} \
   --mode="train" \
   --eval_after_training=True \
   --params_override="{ type: retinanet, train: { checkpoint: { path: ${RESNET_CHECKPOINT}, prefix: resnet50/ }, train_file_pattern: ${TRAIN_FILE_PATTERN} }, eval: { val_json_file: ${VAL_JSON_FILE}, eval_file_pattern: ${EVAL_FILE_PATTERN}, eval_samples: 5000 } }"
   

   Parameter	Description
   --use_tpu	Train the model on a single Cloud TPU.
   --tpu	Specifies the name of the Cloud TPU. This is set by specifying the environment variable (TPU_NAME).
   --num_cores	Specifies the number of cores on the Cloud TPU.
   --model_dir	Specifies the directory where checkpoints and summaries are stored during model training. If the folder is missing, the program creates one. When using a Cloud TPU, the model_dir must be a Cloud Storage path (`gs://...`). You can reuse an existing folder to load current checkpoint data and to store additional checkpoints as long as the previous checkpoints were created using TPU of the same size and TensorFlow version.
   --mode	Specifies the mode in which to run the model. Valid values are: train and eval

Single Cloud TPU device evaluation

The following procedure uses the COCO evaluation data. It takes about 10 minutes to run through the evaluation steps.

1. Set up the following environment variables:

   (vm)$ export EVAL_SAMPLES=5000
   

2. Run the evaluation script:

     (vm)$ python3 /usr/share/tpu/models/official/detection/main.py \
       --use_tpu=True \
       --tpu=${TPU_NAME} \
       --num_cores=8 \
       --model_dir=${MODEL_DIR} \
       --mode="eval" \
       --params_override="{ type: retinanet, eval: { val_json_file: ${VAL_JSON_FILE}, eval_file_pattern: ${EVAL_FILE_PATTERN}, eval_samples: ${EVAL_SAMPLES} } }"
   

   Parameter	Description
   --use_tpu	Evaluate the model on a single Cloud TPU.
   --tpu	Specifies the name of the Cloud TPU. This is set by specifying the environment variable (TPU_NAME).
   --num_cores	Specifies the number of cores on the Cloud TPU.
   model_dir	Specifies the directory where checkpoints and summaries are stored during model training. If the folder is missing, the program creates one. When using a Cloud TPU, the model_dir must be a Cloud Storage path (`gs://...`). You can reuse an existing folder to load current checkpoint data and to store additional checkpoints as long as the previous checkpoints were created using TPU of the same size and TensorFlow version.
   --mode	Specifies the mode in which to run the model. Valid values are: train and eval.

At this point, you can either conclude this tutorial and clean up your GCP resources, or you can further explore running the model on Cloud TPU Pods.

Scaling your model with Cloud TPU Pods

You can get results faster by scaling your model with Cloud TPU Pods. The fully supported RetinaNet model can work with the following Pod slices:

• v2-32
• v3-32

Caution: Scaling to larger Pod slices are experimental with this model. The training scripts in this tutorial were optimized for a v3-32 TPU node. They will not run as-is on a v2 configuration because they require additional memory.

Note: The script trains for 2109 steps. It takes approximately 30 minutes to train on a v3-32 TPU type and 10 minutes to train on a v3-128 TPU type.

1. Delete the Cloud TPU resource you created for training the model on a single device.

   (vm)$ ctpu delete --project=${PROJECT_ID} \
    --tpu-only \
    --zone=europe-west4-a \
    --name=retinanet-tutorial

2. Run the ctpu up command, using the tpu-size parameter to specify the Pod slice you want to use. For example, the following command uses a v3-32 Pod slice.

   (vm)$ ctpu up --project=${PROJECT_ID} \
     --tpu-only \
     --tpu-size=v3-32 \
     --zone=europe-west4-a \
     --tf-version=1.15.5 \
     --name=retinanet-tutorial-pod
    

3. The configuration you specified appears. Enter y to approve or n to cancel.

   You will see a message: Operation success; not ssh-ing to Compute Engine VM due to --tpu-only flag. Since you previously completed SSH key propagation, you can ignore this message.

4. Set up the following environment variables:

   (vm)$ export MODEL_DIR=${STORAGE_BUCKET}/retinanet-model-pod
   (vm)$ export TPU_NAME=retinanet-tutorial-pod
   

5. Run the Pod training script on a v3-32 TPU node:

   (vm)$ python3 /usr/share/tpu/models/official/detection/main.py \
   --use_tpu=True \
   --tpu=${TPU_NAME} \
   --num_cores=32 \
   --model_dir=${MODEL_DIR} \
   --mode="train" \
   --eval_after_training=False \
   --params_override="{ type: retinanet, train: { train_batch_size: 1024, total_steps: 2109, learning_rate: { warmup_steps: 820, init_learning_rate: 0.64, learning_rate_levels: [0.064, 0.0064], learning_rate_steps: [1641, 1992] }, checkpoint: { path: ${RESNET_CHECKPOINT}, prefix: resnet50/ }, train_file_pattern: ${TRAIN_FILE_PATTERN} }, resnet: { batch_norm: { batch_norm_momentum: 0.9 }}, fpn: { batch_norm: { batch_norm_momentum: 0.9 }}, retinanet_head: { batch_norm: { batch_norm_momentum: 0.9 }} }"
   

   Parameter	Description
   --use_tpu	Train the model on a Cloud TPU pod.
   --tpu	Specifies the name of the Cloud TPU. This is set by specifying the environment variable (TPU_NAME).
   --num_cores	Specifies the number of cores on the Cloud TPU.
   --model_dir	Specifies the directory where checkpoints and summaries are stored during model training. If the folder is missing, the program creates one. When using a Cloud TPU, the model_dir must be a Cloud Storage path (`gs://...`). You can reuse an existing folder to load current checkpoint data and to store additional checkpoints as long as the previous checkpoints were created using TPU of the same size and TensorFlow version.
   --mode	Specifies the mode in which to run the model.
   --eval_after_training	Set to True to evaluate the model after training,
   --params_override	Override model parameters with specified values.

Cleaning up

To avoid incurring charges to your Google Cloud account for the resources used in this tutorial, either delete the project that contains the resources, or keep the project and delete the individual resources.

1. Disconnect from the Compute Engine instance, if you have not already done so:

   (vm)$ exit
   

   Your prompt should now be username@projectname, showing you are in the Cloud Shell.

2. In your Cloud Shell, run ctpu delete with the --zone flag you used when you set up the Cloud TPU to delete your Compute Engine VM and your Cloud TPU:

   $ ctpu delete --project=${PROJECT_ID} \
     --zone=europe-west4-a \
     --name=retinanet-tutorial
   

   Important: If you set the TPU resources name when you ran ctpu up, you must specify that name with the --name flag when you run ctpu delete in order to shut down your TPU resources.

3. Run the following command to verify the Compute Engine VM and Cloud TPU have been shut down:

   $ ctpu status --project=${PROJECT_ID} \
     --zone=europe-west4-a
   

   The deletion might take several minutes. A response like the one below indicates there are no more allocated instances:

   2018/04/28 16:16:23 WARNING: Setting zone to "europe-west4-a"
   No instances currently exist.
    Compute Engine VM:     --
    Cloud TPU:             --
   

   Caution: All training data will be lost when you delete your bucket, so only do this step when you are finished running the tutorial.

4. Run gsutil as shown, replacing bucket-name with the name of the Cloud Storage bucket you created for this tutorial:

   $ gsutil rm -r gs://bucket-name
   

What's next

In this tutorial you have trained the RetinaNet model using a sample dataset. The results of this training are (in most cases) not usable for inference. To use a model for inference you can train the data on a publicly available dataset or your own data set. Models trained on Cloud TPUs require datasets to be in TFRecord format.

You can use the dataset conversion tool sample to convert an image classification dataset into TFRecord format. If you are not using an image classification model, you will have to convert your dataset to TFRecord format yourself. For more information, see TFRecord and tf.Example

Hyperparameter tuning

To improve the model's performance with your dataset, you can tune the model's hyperparameters. You can find information about hyperparameters common to all TPU supported models on GitHub. Information about model-specific hyperparameters can be found in the source code for each model. For more information on hyperparameter tuning, see Overview of hyperparameter tuning, Using the Hyperparameter tuning service and Tune hyperparameters.

Inference

Once you have trained your model you can use it for inference (also called prediction). AI Platform is a cloud-based solution for developing, training, and deploying machine learning models. Once a model is deployed, you can use the AI Platform Prediction service.

Train with different image sizes

You can explore using a larger neural network (for example, ResNet-101 instead of ResNet-50). A larger input image and a more powerful neural network will yield a slower but more precise model.

Use a different basis

Alternatively, you can explore pre-training a ResNet model on your own dataset and using it as a basis for your RetinaNet model. With some work, you can also swap in an alternative neural network in place of ResNet. Finally, if you are interested in implementing your own object detection models, this network may be a good basis for further experimentation.