seamless_communication
seamless_communication copied to clipboard
RAM Usage
can let me know minimum RAM requirements about each model (M4T_Large, M4T_Medium)?
You should be able to extrapolate the amount of VRAM required based on the size of the model.
I ran it successfully on a Linux VM x64, with 16 Gb RAM. (host machine s a M1 Mac). Can't make it work on M1 natively tho.
@billylo1 M1 build of fairseq2 is work in progress right now. Please follow here.