Examples Gallery

A couple of examples to illustrate different implementation of dot product (see also sphinx-gallery).

Getting started

pytorch nightly build should be installed, see Start Locally.

git clone https://github.com/sdpython/experimental-experiment.git
pip install onnxruntime-gpu pynvml
pip install -r requirements-dev.txt
export PYTHONPATH=$PYTHONPATH:<this folder>

Compare torch exporters

The script evaluates the memory peak, the computation time of the exporters. It also compares the exported models when run through onnxruntime. The full script takes around 20 minutes to complete. It stores on disk all the graphs, the data used to draw them, and the models.

python _doc/examples/plot_torch_export.py -s large

See pytorch and onnx for an organized version of this page.

102: Examples with onnxscript

102: Examples with onnxscript

101: Graph Optimization

101: Graph Optimization

102: Tweak onnx export

102: Tweak onnx export

101: Profile an existing model with onnxruntime

101: Profile an existing model with onnxruntime

101: Onnx Model Rewriting

101: Onnx Model Rewriting

101: Linear Regression and export to ONNX

101: Linear Regression and export to ONNX

101: A custom backend for torch

101: A custom backend for torch

101: Some dummy examples with torch.export.export

101: Some dummy examples with torch.export.export

102: Convolution and Matrix Multiplication

102: Convolution and Matrix Multiplication

102: Fuse kernels in a small Llama Model

102: Fuse kernels in a small Llama Model

301: Compares LLAMA exporters

301: Compares LLAMA exporters

301: Compares LLAMA exporters for onnxrt backend

301: Compares LLAMA exporters for onnxrt backend

102: Measure LLAMA speed

102: Measure LLAMA speed

201: Evaluate DORT Training

201: Evaluate DORT Training

201: Evaluate DORT

201: Evaluate DORT

201: Evaluate different ways to export a torch model to ONNX

201: Evaluate different ways to export a torch model to ONNX