multi-GPU environment - Githubissues

hieutrungle / XPINNs-Tensorflow2.0

Unofficial Extended Physics-Informed Neural Networks (XPINNs) Implementation Tensorflow 2.0

GNU General Public License v3.0

2 stars 0 forks source link

Open WeCcRy opened 2 weeks ago

WeCcRy commented 2 weeks ago

i wonder how to put it on a machine with multi-GPU to accelerate its training?

hieutrungle commented 2 weeks ago

Could you try distributed training tutorial from Tensorflow? https://www.tensorflow.org/guide/keras/distributed_training

WeCcRy commented 1 week ago

I tried but failed. XPINNs seems different from other DL projects.

hieutrungle commented 1 week ago

I have not used Tensorflow with multiple GPUs on a machine.

However, I did use multiprocessing with Tensorflow on Linux by adding following:

`import multiprocessing"
multiprocessing.set_start_method("spawn", force=True) I do not if that would help to solve your problem.

Note: I have migrated to Pytorch for many reasons. I also recommend to try Jax for PINN or work that requires differentiation.