Problem tracing binary CNN model after recent tracing optimization about hls4ml HOT 8 OPEN

Maybe it does fix it. I just checked out the PR, and I get no error. I thought that it complained about a string activation having __name__ extracted, but I don't see that problem any more. So I think once the #961 goes in we can maybe close this.

from hls4ml.

I am including @AdrianAlan and @jmduarte because they were involved in the latest tracing updates.

from hls4ml.

Thanks for catching that. I will take a look.

from hls4ml.

There's a fix for this in #961 (that PR needs bit more work to clean up)

from hls4ml.

I am not sure it's a complete fix. When I attempted the fairly straightforward fix of checking that it's not None, I ran into isuses when activation was a string.

from hls4ml.

We should validate the fix on this, as well.

from hls4ml.

The fix LGTM.

@jmitrevs What do you mean when you say you ran into issues? When you add, let's say, activation='relu', in L33 in your example, the output looks all right, but the affected layer is out of order. Is that what you refer to?

from hls4ml.