Torch: print model at log verbosity 3

rwth-i6 / returnn

The RWTH extensible training framework for universal recurrent neural networks

http://returnn.readthedocs.io/

Other

349 stars 130 forks source link

Torch: print model at log verbosity 3 #1575

Open NeoLegends opened 4 months ago

NeoLegends commented 4 months ago

I find log verbosity 3 a reasonable verbosity for "daily" work/trainings, but I'm missing the model structure in the log. This is because it's only printed at v4 level. Is there any good reason the log is not printed at v3?

See https://github.com/rwth-i6/returnn/blob/2c0b75c1c0c61732a12a175ac3cef433f0bea271/returnn/torch/engine.py#L865C53-L865C53

albertz commented 4 months ago

I find the model structure sth which you only really want to see on debugging. I rarely look at it. Or maybe only once when I wrote a new model, i.e. for most trainings, not at all. I actually considered removing it completely, or only printing it for verbosity level 5, or only under a given special flag, which would be disabled by default. Because it totally spams my log, and it's annoying to skip over it when I want to see the other things at startup.