Is your feature request related to a problem? Please describe.
In the vision module of bitbots_main, we are currently using tvm compiled *.so modules created with tvm version 0.11.0 of YOEO.
These were created with the convertONNXModelToTVM.py` script.
With version 0.19.0 I've verified that the model still works.
It seems, that since version 0.20.0 a lot of the internal structure of TVM has changed, and I assume that with the full introduction of the relax VM in TVM the API for converting the model also has changed.
Describe the solution you'd like
Verify that the conversion script works with version 0.22.0 of TVM, and potentially add some documentation on how to load the resulting *.so.
If it does not, update the script accordingly.
Is your feature request related to a problem? Please describe.
In the vision module of
bitbots_main, we are currently using tvm compiled*.somodules created with tvm version0.11.0of YOEO.These were created with the convertONNXModelToTVM.py` script.
With version
0.19.0I've verified that the model still works.It seems, that since version 0.20.0 a lot of the internal structure of TVM has changed, and I assume that with the full introduction of the relax VM in TVM the API for converting the model also has changed.
Describe the solution you'd like
Verify that the conversion script works with version
0.22.0of TVM, and potentially add some documentation on how to load the resulting*.so.If it does not, update the script accordingly.