Added quickTunerGen.py, the main driver of the quick tuner scripts #1576
Conversation
which generates quick tuner perf configs given the dataframe/tsv file produced by quickTunerPreproc.py.
instead of 'TFlops'
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## develop #1576 +/- ##
===========================================
- Coverage 77.51% 77.40% -0.12%
===========================================
Files 90 92 +2
Lines 24358 24580 +222
Branches 3414 3451 +37
===========================================
+ Hits 18881 19026 +145
- Misses 4132 4189 +57
- Partials 1345 1365 +20
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
| as input. | ||
| Needs the input to be a combined normalized dataframe (default from quickTunerPreproc.py) | ||
|
|
||
| Usage: clusterConfigs.py [-h] --input-file INPUT_FILE [--method {default,topNSelect,topMode,takeNEach,fairSelect,hardcoded} [{default,topNSelect,topMode,takeNEach,fairSelect,hardcoded} ...]] [--save] [--debug] [--num NUM] [--perfconfig--format] |
| os.environ['OMP_NUM_THREADS'] = '1' | ||
| import sys | ||
|
|
||
| sys.path.append('../..') |
There was a problem hiding this comment.
... ???
You shouldn't need to do this
| import re | ||
| import glob | ||
|
|
||
| class quickTunerMethod(object): |
There was a problem hiding this comment.
Same case/(object) nits as on other PRs
|
|
||
| trans_cols = ['TransA', 'TransB'] | ||
|
|
||
| param_cols = [ 'G', 'M', 'N','K'] |
|
|
||
|
|
||
| self.default_f16 = pd.DataFrame({ | ||
| "M/block": [128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 32, 32, 32, 32, 32, 32, 16, 16, 16, 16], |
There was a problem hiding this comment.
.. what's this about? Why do we have this? Where are the notes to keep it up to date?
There was a problem hiding this comment.
Also, if you're doing to do this, build the dataframe from a list of lists and give the column names separately - this is unreadable.
|
|
||
| if __name__ == '__main__': | ||
| main(sys.argv[1:]) | ||
|
|
There was a problem hiding this comment.
Something feels really awkward here - I can't for the life of me follow what all this is supposed to be doing and there's a whole lot of for loops and manual bouncing around of columns
Generates quick tuner perf configs given the dataframe/tsv file produced by quickTunerPreproc.py.