ashutoshsaboo
commented
2 years ago +1 on this. I faced the same issue when logging sklearn's classification report and confusion matrices with Aim. These are fairly common metrics used for classification, so would be helpful to have native official support in Aim for these.
@alberttorosyan @SGevorg
🚀 Feature
Add ability to track 1d, 2d arrays and display them as metrics, for example precision recall threshold curves, confusion matrix, rocauc
Motivation
In case of common DS metrics (precision/recall curve, RocAUC, etc) it make sense to save the entire array in a single command like this: run.track(aim.Vector(np.random.rand(10)),name='RandomMetric',context={'a':3}) Instead of this: [run.track(v,name='RandomMetric',context={'a':3}) for v in np.random.rand(10)]