Fabio De Sousa Ribeiro

> This is because we have a target sequence, `trg`, of something like `[, A, B, C, ]`. We want our decoder to predict what the next item in the...

Hi, You could start by downloading the MorphoMNIST dataset as described in the readme, then launching: [https://github.com/biomedia-mira/causal-gen/blob/main/src/run_local.sh](url) Best, Fabio

Hi! I believe the minimum Python version that should work is 3.8. As for pandas, I think this came from auto-generating the requirements file, I'd try the 2.0.1 version and...

No worries, happy to help! To answer your question, the "x" key in the batch dictionary refers to an input image variable - it is returned by the respective dataloader....

My apologies for the delay. Note that `train_cf.py` is for the optional counterfactual training/fine-tuning step described in the paper, which you may not need depending on how well your base...

Hi, thanks :) As we focused on the variational perspective rather than SDEs we chose not to include that derivation. What you're looking for can be found in Appendix E...

No worries! To clarify, that paper translates the notation from the original SDE paper into VDM's, the actual derivation is in the former as mentioned. Best, Fabio

Hi, You can find some checkpoints on huggingface here: [https://huggingface.co/spaces/mira-causality/counterfactuals/tree/main/checkpoints/m_b_v_s](url) Best, Fabio

Hi, They are HVAE checkpoints; simple VAEs don't tend to work well for larger images. It depends what you're trying to evaluate. If you want to measure counterfactual _effectiveness_ and...