search

google-research / tuning_playbook

A playbook for systematically maximizing the performance of deep learning models.
Other
26.29k stars 2.18k forks source link

Request to cover "2-3 weeks upper bound with data-freeze" #73

Open 21kc-caracol opened 1 month ago

21kc-caracol commented 1 month ago

Hi,

I would greatly appreciate if you could cover your insights for a short time frame with "data freeze", before new training/validation/test data is added.

I've read your playbook, and I understand your dev' times are pretty longer than mine.

In my case, every couple of weeks, even every 3~ weeks, new data is added.

I wonder what are the most significant actions you would do with such a time frame

🙏