scvi-tools
scvi-tools copied to clipboard
scverse
Deep probabilistic analysis of single-cell and spatial omics data
Proposal to implement Prescient. 1. Code: https://github.com/gifford-lab/prescient 2. Paper: https://www.nature.com/articles/s41467-021-23518-w Level of difficulty: Intermediate, the model is already written with PyTorch. A custom training plan is probably necessary.
We need to write a script that runs all the notebooks. At the very least so it can be done locally, but ideally on github (perhaps with paying AWS or...
Apply this suggestion to our code where/if possible: https://twitter.com/PyTorch/status/1437838231505096708?s=20
Dear scvi-tools Team, My name is Thomas Chaton and I am part of the PyTorch Lightning Team. I dived into this framework while investigating speed drop (here is the tracking...
We already merged #1103 but I'm not noticing that there are speed issues with pytorch lightning 1.4.x over 1.3.x. Please see the discussion here https://github.com/PyTorchLightning/pytorch-lightning/issues/9304 Not sure if we should...
This PR is a draft and not completed, but it shows a slightly different approach which is compatible with DDP. Explanation: PyTorch Lightning injects a DistributedSampler within the DataLoader. To...
Right now, scANVI returns the prediction for both the cells with unobserved labels and observed labels. We should add a parameter (and default True) to return the observed labels for...
I find myself frequently doing pairwise differential expression for all clusters in the data, and sometimes I end up running it as an overnight job. It would be very handy...
Would be great to have validation set ELBO loss for pyro models. As discussed with Adam here https://github.com/YosefLab/scvi-tools/pull/1059#discussion_r633935024 @adamgayoso I can implement this, but could you point me where this...
We should add an `is_sparse` param to our distributions (at least [NB](https://github.com/YosefLab/scvi-tools/blob/0ef7ff0e9a06d65cd458b59c516971099bb55756/scvi/distributions/_negative_binomial.py#L236-L280), with LL computation [here](https://github.com/YosefLab/scvi-tools/blob/0ef7ff0e9a06d65cd458b59c516971099bb55756/scvi/distributions/_negative_binomial.py#L67-L102)) As an example in Pyro for Poisson, see [here](https://github.com/pyro-ppl/pyro/pull/2802/files) If X is known to...