Correction of various issues after first large compute

PaulBautin commented 4 years ago

DONE

Display the time after each command in iteration loop #52
Only generate the QC for the vertebral labeling, and the first segmentation (untransformed, un-rescaled) #51

FIX #51 FIX #52

jcohenadad commented 4 years ago

@PaulBautin could you pls display an example of timing you got? so we can compare with compute canada outputs

PaulBautin commented 4 years ago

Test on 2 subjects, 1 rescaling, 2 transfo Total Duration: 00h03m40s

Hardware specs: OS: linux (Linux-5.4.0-42-generic-x86_64-with-debian-bullseye-sid) Processor: Intel® Core™ i5-9600KF CPU @ 3.70GHz × 6 CPU cores: Available: 6, Used by SCT: 1

results for sub-amu01 (results were almost identical):

TIME: Duration of segmentation on non cropped image: 0hrs 1min 3sec TIME: Duration of labelling: 0hrs 1min 19sec

RESCALING = 1 TIME: Duration of image transfo t1: 0hrs 0min 1sec TIME: Duration of labelling transfo t1: 0hrs 0min 4sec TIME: Duration of segmentation t1: 0hrs 0min 6sec TIME: Duration of sct_process_segmentation t1: 0hrs 0min 6sec

TIME: Duration of of image transfo t2: 0hrs 0min 1sec TIME: Duration of labelling transfo t2: 0hrs 0min 3sec TIME: Duration of segmentation t2: 0hrs 0min 6sec TIME: Duration of sct_process_segmentation t2: 0hrs 0min 5sec

RESCALING = 0.95 TIME: Duration of image transfo t1: 0hrs 0min 2sec TIME: Duration of labelling transfo t1: 0hrs 0min 3sec TIME: Duration of segmentation t1: 0hrs 0min 6sec TIME: Duration of sct_process_segmentation t1: 0hrs 0min 6sec

TIME: Duration of image transfo t2: 0hrs 0min 1sec TIME: Duration of labelling transfo t2: 0hrs 0min 3sec TIME: Duration of segmentation t2: 0hrs 0min 6sec TIME: Duration of sct_process_segmentation t2: 0hrs 0min 6sec

jcohenadad commented 4 years ago

@PaulBautin can you please indicate hardware specs in your post above (RAM, number of CPU cores, clock)

jcohenadad commented 4 years ago

@PaulBautin here is the output from compute canada (one subject): can you please clean up the other verbose: https://pastebin.com/ihUhcvBc

PaulBautin commented 4 years ago

@jcohenadad, I do not understant what you mean by other verbose? Do you want to remove other verbose than https://github.com/sct-pipeline/csa-atrophy/pull/53#issuecomment-679212632 in pastebin?

jcohenadad commented 4 years ago

@jcohenadad, I do not understant what you mean by other verbose? Do you want to remove other verbose than #53 (comment) in pastebin?

just curate the pastbin i made to only show the TIME items for easy comparison with https://github.com/sct-pipeline/csa-atrophy/pull/53#issuecomment-679212632

PaulBautin commented 4 years ago

results from compute canada on sub-unf06 SCT info:

version: git-master-253936cfc07712c587d9f4cfbbcb760f151172f0
path: /home/jcohen/sct OS: linux (Linux-3.10.0-1127.8.2.el7.x86_64-x86_64-with-centos-7.8.2003-Core) CPU cores: Available: 32, Used by SCT: 1 RAM: MemTotal: 527986272 kB total used free shared buff/cache available Mem: 515611 121291 342871 3724 51448 384671 Swap: 10239 648 9591

TIME: Duration of reorienting and resampling: 0hrs 0min 17sec TIME: Duration of segmentation on non cropped image: 0hrs 2min 13sec TIME: Duration of labelling: 0hrs 2min 21sec

RESCALING = 1 TIME: Duration of of image transfo t1: 0hrs 0min 3sec TIME: Duration of labelling transfo t1: 0hrs 0min 4sec TIME: Duration of segmentation t1: 0hrs 0min 18sec TIME: Duration of sct_process_segmentation t1: 0hrs 0min 11sec

TIME: Duration of of image transfo t2: 0hrs 0min 2sec TIME: Duration of labelling transfo t2: 0hrs 0min 3sec TIME: Duration of segmentation t2: 0hrs 0min 19sec TIME: Duration of sct_process_segmentation t2: 0hrs 0min 10sec

TIME: Duration of of image transfo t3: 0hrs 0min 2sec TIME: Duration of labelling transfo t3: 0hrs 0min 3sec TIME: Duration of segmentation t3: 0hrs 0min 19sec TIME: Duration of sct_process_segmentation t3: 0hrs 0min 10sec

RESCALING = 0.99 TIME: Duration of of image transfo t1: 0hrs 0min 2sec TIME: Duration of labelling transfo t1: 0hrs 0min 4sec TIME: Duration of segmentation t1: 0hrs 0min 19sec TIME: Duration of sct_process_segmentation t1: 0hrs 0min 11sec

TIME: Duration of of image transfo t2: 0hrs 0min 2sec TIME: Duration of labelling transfo t2: 0hrs 0min 4sec TIME: Duration of segmentation t2: 0hrs 0min 19sec TIME: Duration of sct_process_segmentation t2: 0hrs 0min 11sec

TIME: Duration of of image transfo t3: 0hrs 0min 2sec TIME: Duration of labelling transfo t3: 0hrs 0min 4sec TIME: Duration of segmentation t3: 0hrs 0min 19sec TIME: Duration of sct_process_segmentation t3: 0hrs 0min 11sec

jcohenadad commented 4 years ago

ok, so about 30s/iteration. still far from the 4-9min per iteration reported in https://github.com/sct-pipeline/csa-atrophy/issues/52#issuecomment-679104617.

Maybe this is due to the number of subjects. Let's approve this PR, wait for a fix to https://github.com/neuropoly/spinalcordtoolbox/issues/2859 and then i will re-run a processing (now with the profiling).

PaulBautin commented 4 years ago

Let's approve this PR, wait for a fix to neuropoly/spinalcordtoolbox#2859 and then i will re-run a processing (now with the profiling).

Sounds good. I've implemented the quick fix from https://github.com/neuropoly/spinalcordtoolbox/issues/2859#issuecomment-679207356, could you describe the fix you are waiting for?

jcohenadad commented 4 years ago

I've implemented the quick fix from neuropoly/spinalcordtoolbox#2859 (comment), could you describe the fix you are waiting for?

where did you implement it? i don't see a PR (should be specific PR)

jcohenadad commented 4 years ago

ah! i see you've implemented in this PR

again, again, again: PR should be specific to an issue, not a dump for many corrections that are unrelated to each other.

PaulBautin commented 4 years ago

I have re-arranged the PR. @jcohenadad, could you review?

sct-pipeline / csa-atrophy

Correction of various issues after first large compute #53