xref: /aosp_15_r20/external/pytorch/benchmarks/README.md (revision da0073e96a02ea20f0ac840b70461e3646d07c45)
1# PyTorch Benchmarks
2
3This folder contains scripts that produce reproducible timings of various PyTorch features.
4
5It also provides mechanisms to compare PyTorch with other frameworks.
6
7## Setup environment
8Make sure you're on a machine with CUDA, torchvision, and pytorch installed. Install in the following order:
9```
10# Install torchvision. It comes with the pytorch stable release binary
11conda install pytorch torchvision -c pytorch
12
13# Install the latest pytorch master from source.
14# It should supersede the installation from the release binary.
15cd $PYTORCH_HOME
16python setup.py build develop
17
18# Check the pytorch installation version
19python -c "import torch; print(torch.__version__)"
20```
21
22## Benchmark List
23
24Please refer to each subfolder to discover each benchmark suite. Links are provided where descriptions exist:
25
26* [Fast RNNs](fastrnns/README.md)
27* [Dynamo](dynamo/README.md)
28* [Functional autograd](functional_autograd_benchmark/README.md)
29* [Instruction counts](instruction_counts/README.md)
30* [Operator](operator_benchmark/README.md)
31* [Overrides](overrides_benchmark/README.md)
32* [Sparse](sparse/README.md)
33* [Tensor expression](tensorexpr/HowToRun.md)
34