Evaluating soNNia likelihood along with qm.norm_productive and qm.Z

[psathyrella]

I'm evaluating likelihoods using the module approach since I need qm.norm_productive and qm.Z, which I think are not available with the command line approach. This has been working well for me with SONIA, schematically:

qm = Sonia(ppost_model="humanIGH")
...
qs, pgens, pposts = qm.evaluate_seqs([...])

[...] =  qm.norm_productive, qm.Z

But when I try the same approach with soNNia (just changing qm = SoNNia(ppost_model="humanIGH")) it crashes because it can't find an .h5 file in the default model dir (indeed there don't seem to be any .h5 files in default_models/. Is this expected? It would be nice if I didn't need to train a soNNia model from scratch.

thanks!

[giulioisac]

Hello!

You can find the default deep model for human_B_heavy here. If you need it for other chains, please let me know. By the way, you can also extract norm_productive and Z directly from the log.txt file in each model directory.

Best,
Giulio

[psathyrella]

This works great, thank you!

[psathyrella]

Actually, yes, could you add the light chain models to that repo as well?

thanks!