Can Ergen

Hi, we are currently still exchanging ideas with lamin and CZI to make the implementation better (and hopefully work towards support throughout all models - currently scVI works). Overall, the...

@ori-kron-wis Can you add tests for all pytorch models (not pyro and not jax) for compile. Can you check speed improvements on your end? You should execute it with: `model.train(accelerator='cuda',...

Needs tests like: `model2.train(accelerator='cuda', batch_size=5000, max_epochs=100, train_size=0.9, plan_kwargs={'n_epochs_kl_warmup': 100, 'compile': True}, datasplitter_kwargs={'drop_last': True})` and then get_elbo, get_reconstruction_loss, get_latent.

To clarify, it runs smoothly with a single GPU. @ilan-gold to me this looks more like an AnnData issue. It’s trying to use multiple file streams to write the file(?)....

There seems to be more legacy. Can your remove all and tag me? I can have another look then.

@ori-kron-wis any update on this?

Hi Ori, scArches also doesn’t work if you add new genes. It just that the prepare function removes those. Adding new proteins is not surgery anymore (small adaptation) but requires...

While I have experienced similar issues in other parts of the code, I don't get why the proposed change fixes the issue. Both the old and new code generate int64...

Closing here as it's not reproducible in my hands. Test below. ``` import scvi from scvi.data import synthetic_iid from scvi.external import SOLO import scanpy as sc import anndata as ad...

Fixed the wrong behavior that we have seen in different models (high significance despite low mean LFC but high standard deviation). It still requires a release note and it would...