Train UNET model using arcgis.learn on one machine multiple GPUs

[JadedEarth]

I'm using latest version 3.2 of Deep Learning and I'm trying to find sample codes for training deep learning model on a single machine with multiple GPUs.  I was told that multiple GPUs only works with arcgis.learn Python API.  But I can't find a sample code to do this on a single machine setup.  Most setup I found are for distributed parallel setup (multiple machines with GPUs).

Can anyone give me a link to sample codes or guides for this new feature in version 3.2 (multiple GPUs on single machine)?

Also, can I train the model using Jupyter Notebook in ArcGIS Pro?  New to arcgis.learn so not sure where to go from here.

Appreciate any help.

[JadedEarth · Accepted solution]

I found the following links helpful:

1.  To install ArcGIS API for Python:  Install and set up | ArcGIS API for Python

2.  To install deep learning libraries and backbones:  arcgis_dl_backbones 2.0.1 - Overview

3.  To train deep learning using multiple GPUs:  Train arcgis.learn models on multiple GPUs | ArcGIS API for Python

I'm still experimenting but I think I'm on the right path.

[adeel_nbs]

Hi @JadedEarth ,

Could you please point me to the resources you mentioned for distributed parallel setup (multiple machines with GPUs)

Thanks