set joblib compress to 3

[vuillaut]

We don't use any compression for the RF dump on disk.

I picked compress=3 from a quick study (see below) that seems to give a good compromise: 1.2s --> 12.1s to dump, 1.5-->3.8s to read, 788MB-->179MB on disk

Study:

In [13]: for i in range(9):
    ...:     print(i)
    ...:     %time joblib.dump(disp_sign,f'disp_sign_{i}.sav', compress=i)
    ...: 
0
CPU times: user 339 ms, sys: 890 ms, total: 1.23 s
Wall time: 1.41 s
1
CPU times: user 8.62 s, sys: 967 ms, total: 9.58 s
Wall time: 9.59 s
2
CPU times: user 9.11 s, sys: 599 ms, total: 9.71 s
Wall time: 9.72 s
3
CPU times: user 11.6 s, sys: 530 ms, total: 12.1 s
Wall time: 12.1 s
4
CPU times: user 12.7 s, sys: 504 ms, total: 13.2 s
Wall time: 13.2 s
5
CPU times: user 16.8 s, sys: 469 ms, total: 17.3 s
Wall time: 17.3 s
6
CPU times: user 28 s, sys: 517 ms, total: 28.5 s
Wall time: 28.5 s
7
CPU times: user 41.3 s, sys: 604 ms, total: 41.9 s
Wall time: 41.9 s
8
CPU times: user 2min 12s, sys: 839 ms, total: 2min 13s
Wall time: 2min 13s

In [15]: for i in range(9):
    ...:     print(i)
    ...:     %time joblib.load(f'disp_sign_{i}.sav')
    ...: 
0
CPU times: user 639 ms, sys: 894 ms, total: 1.53 s
Wall time: 1.53 s
1
CPU times: user 3.62 s, sys: 631 ms, total: 4.25 s
Wall time: 4.26 s
2
CPU times: user 3.39 s, sys: 475 ms, total: 3.87 s
Wall time: 3.87 s
3
CPU times: user 3.28 s, sys: 488 ms, total: 3.77 s
Wall time: 3.77 s
4
CPU times: user 3.35 s, sys: 496 ms, total: 3.85 s
Wall time: 3.85 s
5
CPU times: user 3.24 s, sys: 461 ms, total: 3.7 s
Wall time: 3.7 s
6
CPU times: user 3.12 s, sys: 507 ms, total: 3.62 s
Wall time: 3.62 s
7
CPU times: user 3.14 s, sys: 474 ms, total: 3.62 s
Wall time: 3.62 s
8
CPU times: user 3.09 s, sys: 452 ms, total: 3.55 s
Wall time: 3.55 s

In [17]: ls -sh disp_sign*.sav
788M disp_sign_0.sav  
198M disp_sign_1.sav
187M disp_sign_2.sav 
179M disp_sign_3.sav 
181M disp_sign_4.sav
172M disp_sign_5.sav
167M disp_sign_6.sav 
166M disp_sign_7.sav
162M disp_sign_8.sav

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

Merging #947 (b856209) into master (2271df3) will decrease coverage by 0.03%.
The diff coverage is 44.44%.

@@            Coverage Diff             @@
##           master     #947      +/-   ##
==========================================
- Coverage   85.53%   85.50%   -0.04%     
==========================================
  Files          78       78              
  Lines        6464     6464              
==========================================
- Hits         5529     5527       -2     
- Misses        935      937       +2     

Impacted Files	Coverage Δ
lstchain/reco/dl1_to_dl2.py	74.46% <0.00%> (ø)
lstchain/tests/test_lstchain.py	96.87% <100.00%> (ø)
lstchain/reco/r0_to_dl1.py	93.08% <0.00%> (-0.63%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2271df3...b856209. Read the comment docs.