[Frontend][MXNet] Importer Missing Operators

[sxjscience]

Feature Request

Recently, TVM has supported the incoming MXNet 2.0 frontend via a series of PRs: #6054, #6699. This enables the new GluonNLP 1.0, which has been upgraded from MXNet 1.x to MXNet 2.0 to convert to TVM graph (as added in dmlc/gluon-nlp#1390).

However, not all operators in MXNet support conversion to TVM. The following is the list of operators that need to be supported.

• _npi_subtract_scalar BART model
• _npi_stack, BART model
• _npi_advanced_indexing_multiple, BART model. This is triggered when we call a[idx1, idx2]. Also see the MXNet-side implementation.

[sxjscience]

@junrushao1994 I created this new issue. Also, I just figured out that I cannot attach the "help wanted" label so would you help with that?

[insop]

For _npi_subtract_scalar BART model: #7191 is submitted.

[insop]

For _npi_stack : #7209 is submitted.

[insop]

For _npi_advanced_indexing_multiple is at #7230 (review), PTAL.

[sxjscience]

Also, there is another issue that I observed in GluonNLP: dmlc/gluon-nlp#1425 . The initial investigation suggests it's related to mixed-dtype support in the MXNet --> TVM converter.

[insop]

For _npi_advanced_indexing_multiple is at #7230 (review), PTAL.

@sxjscience
Any help on the above PR will be appreciated.
Thank you.