srvk / DiViMe

ACLEW Diarization Virtual Machine
Apache License 2.0
32 stars 9 forks source link

minimum storage and RAM for VM #91

Open alecristia opened 5 years ago

alecristia commented 5 years ago

1.6 gHz intel core (dual) 4GB 1600 MHz DDR3

installation freezes and computer may reboot with the usual VM specs, namely:

vbox.cpus = 2
vbox.memory = 3072

Installation freezes but computer is intolerably slow with:

vbox.cpus = 2
vbox.memory = 2048

All seems okay with:

vbox.cpus = 1
vbox.memory = 1024
riebling commented 5 years ago

Very interesting; I didn't think some of the tools would work with less than 2GB of virtualized memory. Maybe that was tools we've since eliminated (LDC)?

alecristia commented 5 years ago

Need to check what happens in my computer when:

vbox.cpus = 1 vbox.memory = 2048

alecristia commented 5 years ago

With 1024, the VM is not functional:

$ vagrant ssh -c "launcher/test.sh" Starting tests Checking for HTK... HTK is installed. Testing noisemes... Noisemes failed - dependencies Noisemes failed - no RTTM output Testing OpenSmile SAD... OpenSmile SAD passed the test. Testing ToCombo SAD... TOCOMBO SAD passed the test. Testing DIARTK... DiarTK passed the test. Testing Yunitator... Yunitator failed - dependencies TEST DIR /vagrant/data/VanDam-Daylong/BN32/test/yunitator-test.log Yunitator failed - no output RTTM Testing the evaluation pipeline... launcher/test.sh: line 163: activate: No such file or directory rm: cannot remove ‘/vagrant/data/VanDam-Daylong/BN32/test/accuracy_noisemesSad_report.csv’: No such file or directory /home/vagrant/launcher/eval.sh: line 82: activate: No such file or directory Traceback (most recent call last): File "/home/vagrant/utils/compute_metrics.py", line 1, in from pyannote.core import Segment, Timeline, Annotation ImportError: No module named pyannote.core /home/vagrant/launcher/eval.sh: line 94: conda: command not found The evaluation pipeline failed - dependencies The evaluation pipeline failed the test - output does not match expected launcher/test.sh: line 172: conda: command not found Testing VCM... VCM failed - no output RTTM Testing WCE... Error using LSTMseg (line 132) LSTM segmentation Python call failed. Aborting.

Error in getSyllables (line 57)

Error in WCEtrain (line 151)

WCE failed - dependencies WCE failed - no successful adaptation Some tools did not pass the test, but you can still use others ###################################################################################### To wrap up, we will print out the results of the analyses that we ran during the test. Compare the following results, corresponding to your system, against the reference results printed out below. If the numbers are similar, then your system is working similarly to the original one. If you see bigger changes, then please paste this output onto an issue on https://github.com/srvk/DiViMe/issues/. RESULTS: LINES: 10 DURATION SUM: 296.23 FILE: /vagrant/data/VanDam-Daylong/BN32/test/diartk_goldSad_BN32_010007_test.rttm LINES: 88 DURATION SUM: 212.22 FILE: /vagrant/data/VanDam-Daylong/BN32/test/opensmileSad_BN32_010007_test.rttm LINES: 56 DURATION SUM: 63.66 FILE: /vagrant/data/VanDam-Daylong/BN32/test/tocomboSad_BN32_010007test.rttm LINES: 0 DURATION SUM: FILE: /vagrant/data/VanDam-Daylong/BN32/test/tocomboSad*.ToCombo.txt.rttm ** REFERENCE RESULTS BEGINS **. LINES: 101 DURATION SUM: 298.637 FILE: /vagrant/data/VanDam-Daylong/BN32/test/BN32_010007_test.rttm LINES: 101 DURATION SUM: 298.637 FILE: /vagrant/data/VanDam-Daylong/BN32/test/diartk_goldSad_BN32_010007_test.rttm LINES: 37 DURATION SUM: 31.9 FILE: /vagrant/data/VanDam-Daylong/BN32/test/noisemesSad_BN32_010007_test.rttm LINES: 88 DURATION SUM: 212.22 FILE: /vagrant/data/VanDam-Daylong/BN32/test/opensmileSad_BN32_010007_test.rttm LINES: 56 DURATION SUM: 63.66 FILE: /vagrant/data/VanDam-Daylong/BN32/test/tocomboSad_BN32_010007_test.rttm LINES: 31 DURATION SUM: 24.7 FILE: /vagrant/data/VanDam-Daylong/BN32/test/vcm_BN32_010007_test.rttm LINES: 105 DURATION SUM: 302 FILE: /vagrant/data/VanDam-Daylong/BN32/test/yunitator_BN32_010007_test.rttm ** REFERENCE RESULTS ENDS **. Evaluation pipeline:

Evaluating noisemesSad on /vagrant/data/VanDam-Daylong/BN32/test with respect to : accuracy

accuracy report detection accuracy true positive true negative false positive false %
item
BN32_010007_test.rttm 11.24 30.80 3.11 0.20 267.57 TOTAL 11.24 30.80 3.11 0.20 267.57 Connection to 127.0.0.1 closed.

alecristia commented 5 years ago

macbook 11" 1.6 gHz intel core (dual) 4GB 1600 MHz DDR3 crashes when vbox.cpus = 1 vbox.memory = 2048

If the hypothesis that errors in the log above are due to space constraints, then this means we need to specify a much bigger minimum RAM in the instructions.

alecristia commented 5 years ago

@fmetze, do you think we should update the installation instructions, to up the minimal requirements as described in this issue?

fmetze commented 5 years ago

Again, I have updated the documentation itself (see https://github.com/srvk/DiViMe/blob/master/docs/source/install.md), but we need to understand how this information gets propagated to https://divime.readthedocs.io/en/latest/install.html