search

Henry1iu / TNT-Trajectory-Prediction

A Unofficial Pytorch Implementation of TNT: Target-driveN Trajectory Prediction
504 stars 95 forks source link

Hardware requirements for model training #13

Closed Thirteentj closed 2 years ago

Thirteentj commented 2 years ago

Hi！What are the hardware requirements for the training of this model? Thank you

Henry1iu commented 2 years ago

Hi,

I think all the computers installed with a GPU that is compatible with the CUDA 10.2 can be capable of the training of my implementation. The better your hardware is, the faster the training.

As for the large dataset issue, if you are using a computer with NVME SSD, you can increase the size of the swap. Also, you can use the "ArgoverseInDisk" dataloader I implemented for the small memory devices. You can take a look at the "argoverse_loader_v2.py".

Best Regards, Jianbang

Thirteentj commented 2 years ago

That's OK. Thank you very much! My device has small memory. I'll try it.