Gökçen Eraslan

The default behavior is the same, if `key` is not given, everything will be stored under adata.uns['paga'] as before and the plotting functions will also look for adata.uns['paga'] unless a...

> I see... but as it's always `adata.uns['paga']` what would happen if I have an old object and then run a new `sc.tl.paga()` with `key` set to something. Then you'll...

I never get adjustText to work without numerous rounds of parameter optimization, so yeah, I agree.

Maybe we should make a new function called `sc.pp.sample`?

Any news on this bug? We also rely on captum for our research, would appreciate any fix.

Are there similar issues in shap.DeepExplainer too?

Thanks! Looking forward to it.

It seems you forgot to add another file in which you defined `tgl_file_callback` global struct and its methods such as - `get_state_filename()` - `get_auth_key_filename()` - `get_secret_chat_filename()`

Seems like it resets in every 100 batch, so not after each epoch.

I think `latent_dim` represents how many categorical variables there is in the model, while `categorical_dim` denotes the number of categories in each latent categorical variable. This is why the "true"...