Daniel Galvez

Okay. It would be helpful to me if you pushed your local work to your personal remote branch occasionally. I would like to see your python script for creating 3D...

> Could you please tell me what is the default input/output vectorization order (e.g. yzx , xyz ) of the cudnnConvolution component ? The default vectorization order for both input...

> For some reason, it's always around 250000. I don't think arc_idx values over 40 million are normal either, but I do not really have a feel for it. If...

Some thoughts from an initial glance: The cuda context could be the culprit. A possibility is that something isn't being destroyed properly in our destructors, so it is retaining something...

One last thing. Try setting `BatchedThreadedNnet3CudaOnlinePipelineConfig::num_decoder_copy_threads` to 0 to verify that you cannot reproduce the issues when this is the case. If I am reading your messages correctly, you have...

FWIW I did create a custom type caster here: https://github.com/nvidia-riva/riva-asrlib-decoder/blob/main/include/riva/asrlib/decoder/pybind11_dlpack_caster.h I am going through legal approval to license it uner pybind11's BSD 3-clause license to I can open a PR....

Hi @steven-johnson , I am using dlpack for a few things. The first thing was bindings for a CUDA-based beam search decoder for non-autoregressive CTC speech recognition models, for which...

The nb::tensor interface is kind of confusing to read in the nanobind source code because its mixed in with a lot of implementation details, but this page shows some of...

I'm fairly certain that you are using a tensorflow version that is too new. See here: https://github.com/tensorflow/tensorflow/issues/30175 Tensorflow changed the name of its shared object file in version 1.14. Maybe...

@danpovey Can you bring this to the attention of Shuo Wang, whose github username I don't know? Thanks.