Simon Kit Sang Chu

Or, the shorter questions are - What input shape does `ApproximateGP` take with `GridInterpolationVariationalStrategy`? - How to build a variational strategy with `k` dimensional input and a single scalar output?

Thanks. I will close the issue after heterogeneous state is also implemented for adaptive solver. Let me know when it is done.

@Shenggan Thanks for your prompt reply. I thought the conda install command would suffice. However, I receive another cuda error when installing through `setup.py`. ``` python setup.py install torch.__version__ =...

After update nvcc to match the cuda version, there is a new error on gnu version. ``` python setup.py install torch.__version__ = 1.11.0 /share/siegellab/software/kschu/anaconda3/envs/fastfold/lib/python3.8/site-packages/setuptools/dist.py:490: UserWarning: Normalizing '0.1.0-beta' to '0.1.0b0' warnings.warn(tmpl.format(**locals()))...

I am unable to access the uniclust page (page not found). Would you mind sharing another link?

Based on your suggestions, I have made a few little changes to the original script. In particular, I turned off `OMP_PROC_BIND` and changed the `--max-seq-len` and `--target-cov 0.95` flags. I...

I built mmseqs from the latest Git and the issue is solved. There is now another warning. ``` there must be an error: 37 deleted from 4 that now is...

Another issue I notice is that not all representative sequences in 30% are found in 50% or 70%, i.e. the clustering is not "hierarchical". I added the following lines to...

One more question. Which part of the script should be changed if I am interested in doing the same with foldseek?

Yes, I am still experiencing the same issue, i.e. some sequences in clust30 do not show up in clust50 / clust70.