search

wasiur / dynamic_survival_analysis

This repository provides a Python implementation of the dynamical survival analysis method
6 stars 1 forks source link

Linux only? #1

Open Xaviermw opened 4 years ago

Xaviermw commented 4 years ago

Hi, I see that the Conda environment requires Linux dependencies, I just wanted to confirm that the notebooks are only runable with Linux. Thanks!

wasiur commented 4 years ago

Hi Xavier,

Thank you for your interest. The codes should run smoothly on all Unix-based platforms. I think if you do not use the MPI-based parallelization, it will run smoothly on Windows too. We have tested that. We use openmpi for parallelization. You might need to be careful with that if you want to run the MPI-based parallelization on Windows.

Sincerely, Wasiur

On Sun, Apr 19, 2020, 10:56 AM Xaviermw notifications@github.com wrote:

Hi, I see that the Conda environment requires Linux dependencies, I just wanted to confirm that the notebooks are only runable with Linux. Thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/wasiur/dynamic_survival_analysis/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEF5QZISGRXF5A2EI355HDRNMGJLANCNFSM4ML2FNIA .

Xaviermw commented 4 years ago

Great, thanks I've been able to get it running.

Do you have an estimate for how the run-time scales based on n observations (days) with and without the openmpi parallelization? I've been running it for around an hour and a half now with the example OSU data.

Additionally I've run into these negative sqrt warnings (not breaking) and was wondering if that should be expected.

RuntimeWarning: invalid value encountered in sqrt self.stda = 1/np.sqrt((aa1 - 2*aa3 + aa2)/d2) RuntimeWarning: invalid value encountered in sqrt self.stdb = 1/np.sqrt((bb1 - 2*bb3 + bb2)/d*2) RuntimeWarning: invalid value encountered in sqrt self.stdc = 1/np.sqrt((cc1 - 2cc3 + cc2)/d2)

Thanks for the help.

wasiur commented 4 years ago

Hi,

Sorry about the delay. If you do not use the parallelization, it would make sense to use a smaller n (say, 100 or 500 rather than 1000). I think it makes about 40-45 mins for n=100 without parallelization.

Regarding the RuntimeWarning, you can ignore them for now. We will push an updated version of the code that doesn't run into these issues.

Best regards, Wasiur

On Sun, Apr 19, 2020 at 4:23 PM Xaviermw notifications@github.com wrote:

Great, thanks I've been able to get it running.

Do you have an estimate for how the run-time scales based on n observations (days) with and without the openmpi parallelization? I've been running it for around an hour and a half now with the example OSU data.

Additionally I've run into these negative sqrt warnings (not breaking) and was wondering if that should be expected.

RuntimeWarning: invalid value encountered in sqrt self.stda = 1/np.sqrt((aa1 - 2

*aa3 + aa2)/d*2) RuntimeWarning: invalid value encountered in sqrt self.stdb = 1/np.sqrt((bb1 - 2bb3 + bb2)/d

2) RuntimeWarning: invalid value encountered in sqrt self.stdc = 1/np.sqrt((cc1 - 2cc3 + cc2)/d*2)

Thanks for the help.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/wasiur/dynamic_survival_analysis/issues/1#issuecomment-616219233, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEF5QZZ3E7H7X34VKOBJMDRNNMS7ANCNFSM4ML2FNIA .

Xaviermw commented 4 years ago

Great, thanks so much!