Loading…
10
Kept blaming the model when the real issue was my 400 sloppy rows from a Google Sheets export in March
Spent two days convinced the fine tuning was busted before I opened the actual training file and saw half the labels were swapped in the CSV. Anyone built a habit that catches junk data before it ever hits the pipeline?
1 comments
Log in to join the discussion
Log In1 Comment
grayjohnson8d ago
Wait, two days? Man I feel that. But honestly the labels being swapped isn't really a data problem, that's a you problem, the file told you what it was and nobody checked. Habit that saves me: I never open the sheet, I just diff row counts and spot check like 20 rows before anything moves. Takes five minutes and catches most of this junk.
6