Current CI status:

Note: PyTorch/XLA r2.1 will be the last release with XRT available as a legacy runtime. Our main release build will not include XRT, but it will be available in a separate package.

PyTorch/XLA is a Python package that uses the XLA deep learning compiler to connect the PyTorch deep learning framework and Cloud TPUs. You can try it right now, for free, on a single Cloud TPU VM with Kaggle!

Take a look at one of our Kaggle notebooks to get started:

• Stable Diffusion with PyTorch/XLA 2.0
• Distributed PyTorch/XLA Basics

To install PyTorch/XLA a new VM:

pip install torch~=2.0.0 https://storage.googleapis.com/tpu-pytorch/wheels/tpuvm/torch_xla-2.0-cp38-cp38-linux_x86_64.whl

To update your existing training loop, make the following changes:

-import torch.multiprocessing as mp
 import torch_xla.core.xla_model as xm
 import torch_xla.distributed.parallel_loader as pl
 import torch_xla.distributed.xla_multiprocessing as xmp

 def _mp_fn(index):
   ...

   # Move the model paramters to your XLA device
   model.to(xm.xla_device())
 
   # MpDeviceLoader preloads data to the XLA device
   xla_train_loader = pl.MpDeviceLoader(train_loader, xm.xla_device())

-  for inputs, labels in train_loader:
   for inputs, labels in xla_train_loader:
     optimizer.zero_grad()
     outputs = model(inputs)
     loss = loss_fn(outputs, labels)
     loss.backward()
-    optimizer.step()
 
     # `xm.optimizer_step` combines gradients across replicas
     xm.optimizer_step()

 if __name__ == '__main__':
-  mp.spawn(_mp_fn, args=(), nprocs=world_size)
   # xmp.spawn automatically selects the correct world size
   xmp.spawn(_mp_fn, args=())

If you're using DistributedDataParallel, make the following changes:

 import torch.distributed as dist
-import torch.multiprocessing as mp
 import torch_xla.core.xla_model as xm
 import torch_xla.distributed.parallel_loader as pl
 import torch_xla.distributed.xla_multiprocessing as xmp
 import torch_xla.distributed.xla_backend

 def _mp_fn(rank, world_size):
   ...

-  os.environ['MASTER_ADDR'] = 'localhost'
-  os.environ['MASTER_PORT'] = '12355'
-  dist.init_process_group("gloo", rank=rank, world_size=world_size)
   # Rank and world size are inferred from the XLA device runtime
   dist.init_process_group("xla", init_method='xla://')
 
   model.to(xm.xla_device())
   # `gradient_as_bucket_view=tpu` required for XLA
   ddp_model = DDP(model, gradient_as_bucket_view=True)

-  model = model.to(rank)
-  ddp_model = DDP(model, device_ids=[rank])
   xla_train_loader = pl.MpDeviceLoader(train_loader, xm.xla_device())

-  for inputs, labels in train_loader:
   for inputs, labels in xla_train_loader:
     optimizer.zero_grad()
     outputs = ddp_model(inputs)
     loss = loss_fn(outputs, labels)
     loss.backward()
     optimizer.step()

 if __name__ == '__main__':
-  mp.spawn(_mp_fn, args=(), nprocs=world_size)
   xmp.spawn(_mp_fn, args=())

Additional information on PyTorch/XLA, including a description of its semantics and functions, is available at PyTorch.org. See the API Guide for best practices when writing networks that run on XLA devices (TPU, GPU, CPU and...).

Our comprehensive user guides are available at:

Documentation for the latest release

Documentation for master branch

• Cloud TPU VM quickstart
• Cloud TPU Pod slice quickstart
• Profiling on TPU VM
• GPU guide

Note: For TPU Pod customers using XRT (our legacy runtime), we have custom wheels for torch, torchvision, and torch_xla at https://storage.googleapis.com/tpu-pytorch/wheels/xrt.

You can also add yyyymmdd after torch_xla-nightly to get the nightly wheel of a specified date. To get the companion pytorch and torchvision nightly wheel, replace the torch_xla with torch or torchvision on above wheel links.

For PyTorch/XLA release r2.0 and older and when developing PyTorch/XLA, install the libtpu pip package with the following command:

pip3 install torch_xla[tpuvm]

This is only required on Cloud TPU VMs.

To run on compute instances with GPUs.

If PyTorch/XLA isn't performing as expected, see the troubleshooting guide, which has suggestions for debugging and optimizing your network(s).

The PyTorch/XLA team is always happy to hear from users and OSS contributors! The best way to reach out is by filing an issue on this Github. Questions, bug reports, feature requests, build issues, etc. are all welcome!

See the contribution guide.

This repository is jointly operated and maintained by Google, Facebook and a number of individual contributors listed in the CONTRIBUTORS file. For questions directed at Facebook, please send an email to [email protected]. For questions directed at Google, please send an email to [email protected]. For all other questions, please open up an issue in this repository here.

You can find additional useful reading materials in

• Performance debugging on Cloud TPU VM
• Lazy tensor intro
• Scaling deep learning workloads with PyTorch / XLA and Cloud TPU VM
• Scaling PyTorch models on Cloud TPUs with FSDP