Spycsh / xtalker

Faster Talking Face Animation on Xeon CPU
MIT License
120 stars 9 forks source link

( Trying to understand the README) "Install related dependencies of SadTalker" => What do you mean? Do we have to install torch==2.0.0 or torch==1.12.1+cu113,? #13

Open AIhasArrived opened 11 months ago

AIhasArrived commented 11 months ago

Hello I am having a hard time understanding the Readme It says "Install related dependencies of SadTalke", but it the readme from sadtalker mentions a different version of Torch, then just before that, the xtalker readme mention a newer version of Torch. Same for Python I am not sure which one I will need.

I really want to make this xtalker work, even ready to borrow a shappire cpu to see how it goes.

Spycsh commented 11 months ago

Hello @AIhasArrived , you can skip the torch version in sadtalker and just follow the torch installation in xtalker, although I think both can work. Meanwhile, you should install the ffmpeg and pip install the requirements.txt under xtalker.

For resources, I am sorry that I am not the one who sells the machines, but maybe you can use Intel Developer cloud to check whether they offer a shappire rapids xeon cpu.

Spycsh commented 11 months ago

@AIhasArrived , you can also refer to AWS or GCP. Here are the mapping info:

AWS SPR ==> m7i.48xlarge. GCP SPR ==> c3-highcpu-176

AIhasArrived commented 11 months ago

@AIhasArrived , you can also refer to AWS or GCP. Here are the mapping info:

AWS SPR ==> m7i.48xlarge. GCP SPR ==> c3-highcpu-176 Thanks, is there a page where we can see these mapping infos? Also, can you do a pip freeze? to obtain the requirements.txt you have in your xtalker instalaltion? That would be more clear to be honest:).