(Prototype) Use iOS GPU in PyTorch

Author: Tao Xu

Introduction

This tutorial introduces the steps to run your models on iOS GPU. We’ll be using the mobilenetv2 model as an example. Since the mobile GPU features are currently in the prototype stage, you’ll need to build a custom pytorch binary from source. For the time being, only a limited number of operators are supported, and certain client side APIs are subject to change in the future versions.

Model Preparation

Since GPUs consume weights in a different order, the first step we need to do is to convert our TorchScript model to a GPU compatible model. This step is also known as 《prepacking》.

PyTorch with Metal

To do that, we’ll install a pytorch nightly binary that includes the Metal backend. Go ahead run the command below

conda install pytorch -c pytorch-nightly
// or
pip3 install --pre torch -f https://download.pytorch.org/whl/nightly/cpu/torch_nightly.html

Also, you can build a custom pytorch binary from source that includes the Metal backend. Just checkout the pytorch source code from github and run the command below

cd PYTORCH_ROOT
USE_PYTORCH_METAL_EXPORT=ON python setup.py install --cmake

The command above will build a custom pytorch binary from master. The install argument simply tells setup.py to override the existing PyTorch on your desktop. Once the build finished, open another terminal to check the PyTorch version to see if the installation was successful. As the time of writing of this recipe, the version is 1.8.0a0+41237a4. You might be seeing different numbers depending on when you check out the code from master, but it should be greater than 1.7.0.

import torch
torch.__version__ #1.8.0a0+41237a4

Metal Compatible Model

The next step is going to be converting the mobilenetv2 torchscript model to a Metal compatible model. We’ll be leveraging the optimize_for_mobile API from the torch.utils module. As shown below

import torch
import torchvision
from torch.utils.mobile_optimizer import optimize_for_mobile

model = torchvision.models.mobilenet_v2(pretrained=True)
scripted_model = torch.jit.script(model)
optimized_model = optimize_for_mobile(scripted_model, backend='metal')
print(torch.jit.export_opnames(optimized_model))
optimized_model._save_for_lite_interpreter('./mobilenetv2_metal.pt')

Note that the torch.jit.export_opnames(optimized_model) is going to dump all the optimized operators from the optimized_mobile. If everything works well, you should be able to see the following ops being printed out from the console

['aten::adaptive_avg_pool2d',
'aten::add.Tensor',
'aten::addmm',
'aten::reshape',
'aten::size.int',
'metal::copy_to_host',
'metal_prepack::conv2d_run']

Those are all the ops we need to run the mobilenetv2 model on iOS GPU. Cool! Now that you have the mobilenetv2_metal.pt saved on your disk, let’s move on to the iOS part.

Use PyTorch iOS library with Metal

The PyTorch iOS library with Metal support LibTorch-Lite-Nightly is available in Cocoapods. You can read the Using the Nightly PyTorch iOS Libraries in CocoaPods section from the iOS tutorial for more detail about its usage.

We also have the HelloWorld-Metal example that shows how to conect all pieces together.

Note that if you run the HelloWorld-Metal example, you may notice that the results are slighly different from the results we got from the CPU model as shown in the iOS tutorial.

- timber wolf, grey wolf, gray wolf, Canis lupus
- malamute, malemute, Alaskan malamute
- Eskimo dog, husky

This is because by default Metal uses fp16 rather than fp32 to compute. The precision loss is expected.

Use LibTorch-Lite Built from Source

You can also build a custom LibTorch-Lite from Source and use it to run GPU models on iOS Metal. In this section, we’ll be using the HelloWorld example to demonstrate this process.

First, make sure you have deleted the build folder from the 《Model Preparation》 step in PyTorch root directory. Then run the command below

IOS_ARCH=arm64 USE_PYTORCH_METAL=1 ./scripts/build_ios.sh

Note IOS_ARCH tells the script to build a arm64 version of Libtorch-Lite. This is because in PyTorch, Metal is only available for the iOS devices that support the Apple A9 chip or above. Once the build finished, follow the Build PyTorch iOS libraries from source section from the iOS tutorial to setup the XCode settings properly. Don’t forget to copy the ./mobilenetv2_metal.pt to your XCode project and modify the model file path accordingly.

Next we need to make some changes in TorchModule.mm

...
// #import <Libtorch-Lite/Libtorch-Lite.h>
// If it's built from source with Xcode, comment out the line above
// and use following headers
#include <torch/csrc/jit/mobile/import.h>
#include <torch/csrc/jit/mobile/module.h>
#include <torch/script.h>
...

- (NSArray<NSNumber*>*)predictImage:(void*)imageBuffer {
  c10::InferenceMode mode;
  at::Tensor tensor = torch::from_blob(imageBuffer, {1, 3, 224, 224}, at::kFloat).metal();
  auto outputTensor = _impl.forward({tensor}).toTensor().cpu();
  ...
}
...

As you can see, we simply just call .metal() to move our input tensor from CPU to GPU, and then call .cpu() to move the result back. Internally, .metal() will copy the input data from the CPU buffer to a GPU buffer with a GPU compatible memory format. When .cpu() is invoked, the GPU command buffer will be flushed and synced. After forward finished, the final result will then be copied back from the GPU buffer back to a CPU buffer.

The last step we have to do is to add the Accelerate.framework and the MetalPerformanceShaders.framework to your xcode project (Open your project via XCode, go to your project target’s 《General》 tab, locate the 《Frameworks, Libraries and Embedded Content》 section and click the 《+》 button).

If everything works fine, you should be able to see the inference results on your phone.

Conclusion

In this tutorial, we demonstrated how to convert a mobilenetv2 model to a GPU compatible model. We walked through a HelloWorld example to show how to use the C++ APIs to run models on iOS GPU. Please be aware of that GPU feature is still under development, new operators will continue to be added. APIs are subject to change in the future versions.

Thanks for reading! As always, we welcome any feedback, so please create an issue here if you have any.

Learn More

• The Mobilenetv2 from Torchvision

• To learn more about how to use optimize_for_mobile, please refer to the Mobile Perf Recipe