Closed hellwue closed 2 months ago
@hellwue What version of the Python API are you using? It could be a mismatch there if the environment has been upgraded with a newer API release, but I would only expect that to be the case in a custom environment (not the default arcgispro-py3
with DL installer overlaid)
@hellwue Haven't heard back so closing this, but if you continue to have issues with the package versions let us know the details and we can take a look.
ArcGIS Pro Version: 3.2.2. deep-learning-essentials installed like described here under "Developer install steps".
When trying to train a UNet model through the "Train Deep Learning Model" tool it results in an error.
The stack trace when doing this in a Notebook reveals that this issue stems from
fastai
when it checks if there are pretrained models intorchvision.models
, as there are none with the suffix "_weights" as shown in the screenshot below:This weight suffix only seems available for torchvision>=0.13, but version 0.9.2 is installed...
This error still occurs when installing the arcgis deep learning backbones for disconnected environments.