Jamie Morton

Got it. Yes we generate those types of alignment visualizations as depicted in the tensorboard readouts. https://github.com/flatironinstitute/deepblast/blob/master/deepblast/trainer.py#L248-L251 We are currently in the middle of doing a overhaul of the tutorials,...

Hi, it looks like you are using the same methods that I used for evaluation. The original notebooks can be found on the zenodo linked in the paper : https://doi.org/10.5281/zenodo.7731163

Hi, the edges are the match coordinates between the two sequences. Regarding f1 score, if there is an off-by-1 error, the f1 score can be zero, even if the structural...

Hi, we cannot currently do that from the CLI -- you'll need to batch it into smaller chunks, encode and create the database from the encodings (otherwise you'll cram too...

Hi, no that is not feasible. Youd need a much larger GPU cluster to encode that many proteins. We have considered using ESM2 instead of Protrans, then that could take...

Hi I havent tested this on windows, so I cant comment here. Id also double check if your model / data are on the gpu. On Sat, Sep 16, 2023...

Hi - not sure why this is the case. We can look into this, although it may take a while to allocate time. But if you find a solution, feel...

Hi @konstin , sorry for the delayed responses (currently in the midst of revisions again ...) This is an awesome finding! Thank you for posting this. Regarding scores that are...

Wow. Yes that is a typo … On Thu, Jun 23, 2022 at 9:01 AM Stefan Janssen ***@***.***> wrote: > Hi @mortonjt , > shouldn't it be A[i - j,...

Yes, you are right -- this looks like it was inherited from an older version of pytorch where max returned both the value and indices. This will need to updated.