Splitting Model on Different GPUs for Training

zhyever / PatchFusion

[CVPR 2024] An End-to-End Tile-Based Framework for High-Resolution Monocular Metric Depth Estimation

https://zhyever.github.io/patchfusion/

MIT License

958 stars 64 forks source link

Splitting Model on Different GPUs for Training #29

Closed andrewivan123 closed 6 months ago

andrewivan123 commented 6 months ago

Is it possible to train with each of the coarse, fine and fusion network split between different GPUs as currently the VRAM requirement to train is too high for RTX4090?

zhyever commented 6 months ago

Hi, the current version splits the training to three steps, which is aligned with our paper