I am currently trying to implement a custom **GRU/LSTM `OnnxOpConverter`** 
inside `tvm/relax/frontend/onnx`.

However, I’ve run into many issues. I noticed that there is an LSTM 
implementation under `topi/nn` which uses `te.scan`, and I was planning to 
reuse it inside `OnnxOpConverter`. Unfortunately, I always fail.

My main problem is: when I try to call it with something like `bb.emit_te(lstm, 
...)`, I keep getting various syntax errors. I am not very familiar with **TVM 
Relax**, so I am struggling here.

Could anyone help me? Ideally, I would like to have an `LSTM(OnnxOpConverter)` 
implementation that can reuse the existing LSTM in `topi`.

Thanks a lot in advance!





---
[Visit 
Topic](https://discuss.tvm.apache.org/t/how-to-use-topi-lstm-function/18606/1) 
to respond.

You are receiving this because you enabled mailing list mode.

To unsubscribe from these emails, [click 
here](https://discuss.tvm.apache.org/email/unsubscribe/03489ba3ff2086978115e24364bd0349460c3503fe9744d62407d664ac5112d4).

Reply via email to