TPU training (Basic)

Audience: Users looking to train on single or multiple TPU cores.

Lightning supports running on TPUs. At this moment, TPUs are available on Google Cloud (GCP), Google Colab and Kaggle Environments. For more information on TPUs watch this video.

What is a TPU?

Tensor Processing Unit (TPU) is an AI accelerator application-specific integrated circuit (ASIC) developed by Google specifically for neural networks.

A TPU has 8 cores where each core is optimized for 128x128 matrix multiplies. In general, a single TPU is about as fast as 5 V100 GPUs!

A TPU pod hosts many TPUs on it. Currently, TPU v3 Pod has up to 2048 TPU cores and 32 TiB of memory! You can request a full pod from Google cloud or a “slice” which gives you some subset of those 2048 cores.

Run on 1 TPU core

Enable the following Trainer arguments to run on 1 TPU.

trainer = Trainer(accelerator="tpu", devices=1)

Run on multiple TPU cores

For multiple TPU cores, change the value of the devices flag.

trainer = Trainer(accelerator="tpu", devices=8)

Run on a specific TPU core

To run on a specific core, specify the index of the TPU core.

trainer = pl.Trainer(accelerator="tpu", devices=[5])

This example runs on the 5th core, not on five cores.

How to access TPUs

To access TPUs, there are three main ways.

Google Colab

Colab is like a jupyter notebook with a free GPU or TPU hosted on GCP.

To get a TPU on colab, follow these steps:

  1. Go to Google Colab.

  2. Click “new notebook” (bottom right of pop-up).

  3. Click runtime > change runtime settings. Select Python 3, and hardware accelerator “TPU”. This will give you a TPU with 8 cores.

  4. Next, insert this code into the first cell and execute. This will install the xla library that interfaces between PyTorch and the TPU.

    !pip install cloud-tpu-client==0.10
  5. Once the above is done, install PyTorch Lightning.

    !pip install pytorch-lightning
  6. Then set up your LightningModule as normal.

Google Cloud (GCP)

You could refer to this page for getting started with Cloud TPU resources on GCP.


For starting Kaggle projects with TPUs, refer to this kernel.

Optimize Performance

The TPU was designed for specific workloads and operations to carry out large volumes of matrix multiplication, convolution operations and other commonly used ops in applied deep learning. The specialization makes it a strong choice for NLP tasks, sequential convolutional networks, and under low precision operation. There are cases in which training on TPUs is slower when compared with GPUs, for possible reasons listed:

  • Too small batch size.

  • Explicit evaluation of tensors during training, e.g. tensor.item()

  • Tensor shapes (e.g. model inputs) change often during training.

  • Limited resources when using TPU’s with PyTorch Link

  • XLA Graph compilation during the initial steps Reference

  • Some tensor ops are not fully supported on TPU, or not supported at all. These operations will be performed on CPU (context switch).

  • PyTorch integration is still experimental. Some performance bottlenecks may simply be the result of unfinished implementation.

The official PyTorch XLA performance guide has more detailed information on how PyTorch code can be optimized for TPU. In particular, the metrics report allows one to identify operations that lead to context switching.

© Copyright Copyright (c) 2018-2023, Lightning AI et al...

Built with Sphinx using a theme provided by Read the Docs.