Julien Cohen-Adad
Julien Cohen-Adad
> Have we done any benchmark comparisons between -jobs 1 and -jobs N? Yes, I have (and I know some students also did in the past), and there is _definitely_...
> Then, depending on the number of jobs requested (i.e. number of concurrent subjects to be processed), we change TF's behavior? 👍 although we need to keep in mind that...
SCT is no more relying on TF, so this issue is _possibly_ outdated. If it comes back, please feel free to reopen!
my 2 cents: - i would be in favour of doing a copy (errors caused by mutation are more annoying to keep track of-- if they happened in the past...
> The two "first attempts" are identical, and two "second attempts" are identical. So, it's not stochastic/random -- we're arriving at 2 distinct results in a reliable fashion. could there...
kudos for your remarkable investigations detective 👏🙏
Beautiful investigation @joshuacwnewton, thank you so much for looking into this. I've been aware of fishy code inside image.py for the past few years so I'm glad you are looking...
> I don't know what `sct_merge_images` does exactly, but it sounds pretty close to your proposal; could you say what the differences are? Currently, the usage of this function is...
> Another idea might be to include the sct_merge_images in the sct_label_utils? If label fusion is the primary use case, perhaps it would be convenient to have it there as...
for now in the README, and at some point we can integrate it in the new doc that @Drulex is working on https://github.com/neuropoly/spinalcordtoolbox/pull/2781