# machine-learning
- t
Toasty
05/17/2021, 12:49 AMIt's using up almost all of the GPU. - m
mega b
05/17/2021, 12:50 AMjeeez - t
Toasty
05/17/2021, 1:00 AMThank god I bought Colab Pro - j
Justin
05/17/2021, 1:01 AM+1 - j
Justin
05/17/2021, 1:02 AMV100 helps a lot! - t
Toasty
05/17/2021, 1:06 AMoh no no no no no no no - t
Toasty
05/17/2021, 1:06 AMoh no - m
mega b
05/17/2021, 1:07 AMuh oh - s
SidPlays_144p
05/17/2021, 1:07 AMJust let it bake, I'm sure it'll go well - p
pi
05/17/2021, 1:07 AMhow many epochs was this? - p
pi
05/17/2021, 1:08 AM...? - t
Toasty
05/17/2021, 1:08 AM7 - t
Toasty
05/17/2021, 1:09 AMGuess I gotta be more patient. - p
pi
05/17/2021, 1:09 AM30 epochs at least - t
Toasty
05/17/2021, 1:09 AMAlright. - t
Toasty
05/17/2021, 1:09 AMWhich is more important - t
Toasty
05/17/2021, 1:09 AMthe epoch or validation loss - j
Justin
05/17/2021, 1:10 AMAtm increase epochs - p
pi
05/17/2021, 1:10 AMboth is a proxy for actual results - p
pi
05/17/2021, 1:10 AMbut val loss is more important - j
Justin
05/17/2021, 1:10 AMV100 has like 16gigs right - p
pi
05/17/2021, 1:10 AMbut a low val loss alone doesn't mean it's good - j
Justin
05/17/2021, 1:11 AMYes look at the alignment during the training process too - j
Justin
05/17/2021, 1:11 AMLet it run for sometime - t
Toasty
05/17/2021, 1:24 AMFair. - p
pi
05/17/2021, 2:48 AM:/ - m
mega b
05/17/2021, 2:48 AMNice drum kit! - s
SidPlays_144p
05/17/2021, 2:48 AMhammer_strikes_nail.mp3 - s
SidPlays_144p
05/17/2021, 2:49 AMwell he can say Homestar