Resolve OOM issue + better examples for classification#847
Conversation
czaloom
changed the title
czaloom
requested review from
ekorman,
jyono and
rsbowman-striveworks
as code owners
| if idx < n_matched: | ||
| glabel = index_to_label[unique_matches[idx, 1]] | ||
| plabel = index_to_label[unique_matches[idx, 2]] | ||
| confusion_matrix[glabel][plabel]["count"] += 1 |
There was a problem hiding this comment.
now that we're giving all examples is it always true that the count here is just the length of the examples entry? I worry about redundant stuff like this because if these get out of sync it's confusing... Can we just return the examples list and let callers take the len of that? Or is that a breaking interface change?
There was a problem hiding this comment.
I will let Charles answer but i think we are going to try to separate "here is the matrix" from "here are the examples" to allow the UI to be more svelt showing the first, and more precise on the second
There was a problem hiding this comment.
It is a bit redundant now but the intent is to eventually split the examples from the counts entirely. We are still implementing a number of examples cap on the chariot-valor side.
3 participants
Changes
(n_score_thresholds * n_labels * n_datums * n_examples)to(n_score_thresholds * n_labels * n_datums)number_of_examplesoption from classificationExceptionstoErrors