Modules that have backward hooks assigned cannot be compiled

Trainer
1

When I’m calling the .fit function on the lightning module’s object, I’m getting this error.
I’ve tried using a simple single convolution layer and I’m not using any forward or backward hooks.

I was trying to pinpoint the origin of this problem and found out its realted to the log_graph of tensorboard.
TensorBoardLogger(hparams.tblog_path, name=hparams.run_name, log_graph=True)
If I run the same code with log_graph as False, then I’m not getting such an error.

Any idea what could be causing this issue?
Thanks :slight_smile:

2

hey @soumickmj !

can you share a reproducible script, and open an issue here?

Also, we have moved the discussions to GitHub Discussions. You might want to check that out instead to get a quick response. The forums will be marked read-only soon.

Thank you