# machine-learning
- m
mega b
08/21/2021, 6:48 PMhave you tried CPU offloading @User - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:48 PMso no - m
mega b
08/21/2021, 6:48 PMjust means to load the model on cpu first, then convert the model to gpu - m
mega b
08/21/2021, 6:48 PMit helps with lower memory issues - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:49 PMmm ok well how do i do that - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:49 PMi have jupyter with anaconda - m
mega b
08/21/2021, 6:49 PMcant say right now, the talknet synthesis colab is pretty hidden with the code - m
mega b
08/21/2021, 6:50 PMits very clean for sure, but its all behind a server.py - c
Couch
08/21/2021, 6:51 PMwell you could always go over to nvidia nemo github and Figure IT Out - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:52 PMbut like i said - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:52 PMthere must be workaround - m
mega b
08/21/2021, 6:57 PMokay so - m
mega b
08/21/2021, 6:57 PMi just got talknet running on cpu - m
mega b
08/21/2021, 6:57 PMsuper easy - c
Couch
08/21/2021, 6:57 PMWOW! wahtd you change - m
mega b
08/21/2021, 6:57 PMyou have to make the denoiser run without cuda and change the device on controllable_talknet - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:58 PMhow would that do? - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:58 PMor like, do i need to change something/download something? - c
Couch
08/21/2021, 6:58 PMwell i had to do something like this but i had to edit plot_utils.py and train.py so that it Displays Information properly on tt2 - m
mega b
08/21/2021, 6:58 PMnear the top - m
mega b
08/21/2021, 6:59 PMhifi gan - m
mega b
08/21/2021, 6:59 PMremove those cudas that will appear - m
mega b
08/21/2021, 6:59 PMCTRL + S to save - a
Awesome Face (GTX 1660 Ti)
08/21/2021, 6:59 PMalright - m
mega b
08/21/2021, 6:59 PMand yer dune - c
Couch
08/21/2021, 6:59 PMcool