Find a Way to Handle Multiple Results for Benchmarks #5

New Issue

gballan · 2023-07-06T00:28:48-04:00

gballan commented

2023-07-06 00:28:48 -04:00

I require every benchmark be run at least four times and average the most consistent/best runs of the three to tease out any flukey runs. Or in extreme cases signify there's a problem with the test setup that the results are so wild. Thus there should be a way to group results together as a singular test for a card to create a comparison with.

This might require a separate model that sits between the "hardware"/"benchmark" models and the "result" model to group the raw results.

I require every benchmark be run at least four times and average the most consistent/best runs of the three to tease out any flukey runs. Or in extreme cases signify there's a problem with the test setup that the results are so wild. Thus there should be a way to group results together as a singular test for a card to create a comparison with. This might require a separate model that sits between the "hardware"/"benchmark" models and the "result" model to group the raw results.

gballan added the

enhancement

label 2023-07-06 00:28:48 -04:00

gballan referenced this issue

2023-07-06 00:29:58 -04:00

Point Out Anomalous Results #6

gballan referenced this issue from a commit

2023-09-21 22:41:29 -04:00

[Issue #5] - Reworked app to better organize results with their corresponding tests

Sign in to join this conversation.