Fix compiler crash on non-square convolutions

tensil-ai / tensil

Open source machine learning accelerators

https://www.tensil.ai

Other

348 stars 28 forks source link

Fix compiler crash on non-square convolutions #29

Closed petrohi closed 2 years ago

petrohi commented 2 years ago

Fix ONNX frontend parsing pads argument in Conv operation;
Simplify/unify padding, strides and kernel size parsing in frontends;
Test 2D convolutions with non-square kernels and activations;
Test ONNX frontend with all convolutions cases.

shortcut-integration[bot] commented 2 years ago

This pull request has been linked to Shortcut Story #433: Compiler crashes on non-square convolutions.

petrohi commented 2 years ago

The case with differing number of channels is generally covered with ResNet20 tests, but I agree we ideally want an isolated case for single convolution. Will add a note to add this in the future.

tdb-alcorn commented 2 years ago

Makes sense, please go ahead and merge.