qonnx
qonnx copied to clipboard
fastmachinelearning
QONNX: Arbitrary-Precision Quantized Neural Networks in ONNX
Was debugging something and noticed a commented `pdb` that might've meant to be removed, commented so no impact, feel free to close this PR if its just noise 😄
Fixes the issue reported here #86, by adding a missing test for the producer node being a fork node. Added some (probably too extensive) documentation and a new check to...
Adds the cleanup transformation requested in #84 Closes #84
I extended channels last converted in order to work with models with branches. I tested with a dummy model composed by a single fork node and with a complete UNet...
I had recently tried building a larger model (~517 nodes) in FINN and got stuck at CreateDataflowPartition for a long time. I did some timing measurements and tracked the longest...
It seems like pads should default to zero in a conv layer if it's not set and auto_pads is either not set or set to 'NOTSET'. It seems like we...
For `LowerConvsToMatMul`, instead of erroring out, assume no padding when padding in Conv nodes is not specified.
This pull request introduces some additional rounding modes, and provides a table, that more accurately describes their behavior. Concretely, the following table has been added to _docs/qonnx-custom-ops/quant_op.md_: | **Number \...
Metadata
Owner
Metadata
QONNX: Arbitrary-Precision Quantized Neural Networks in ONNX