Write the model card
Why it matters A model card (Mitchell et al. 2019) is how a model is handed to someone else: what it is, how it was trained, how well it does, and where it was not tested. Most cards are written from memory. Yours is checked against your run log.
This lesson reruns your baseline and final runs (seed 0) into a fresh log, scores the final model on the hidden set, and gives you the numbers. Write the card as a dictionary with these fields:
track,final settings(the FINAL dict),weights(the final model's size),training FLOPs(its cost)baseline validation score,validation score,hidden score,hidden interval[low, high]finding(one sentence with your two probe numbers),ablations(one sentence per change with its verdict),limitations(what it was not tested on: at least one full sentence)
The check compares every number with the log and the hidden scoring, to 4 decimal places. Then answer the self-checklist honestly: True or False for each line.
Words and symbols in this lesson
- baseline
- the plain run every change is measured against, over the same seeds
- ablation
- one change to the baseline, run over several seeds, to see what it does
- model card
- a short record of what a model is, how it was trained, how well it does and where it was not tested
Fill card and checklist.
Spotted a problem in this lesson? Report it · Ask about this lesson
This exercise opens soon
Finish the lesson before it and this one unlocks. The whole course is free, no account needed, and you can read this lesson right now.
This browser is blocking cookies for learn.floati.life, so lessons you finished here cannot open in order. Allow cookies for this site and reload.