Onnx example #13
Onnx example #13
Conversation
|
I don't have funding for this work, so removing myself from assignees. |
| import xarray as xr | ||
| from openeo.udf.debug import inspect | ||
|
|
||
| sys.path.insert(0, "onnx_deps") |
There was a problem hiding this comment.
Are we sure this path has to be prepended at the start of sys.path? Because doing this increases the risk on "frankenstein" imports (packages that are partly imported from different versions). If it works by appending at the end, I think it's better to promote that pattern.
|
|
||
| # This will be required to import onnxruntime in the UDF. | ||
| dependencies_url = "https://artifactory.vgt.vito.be:443/auxdata-public/openeo/onnx_dependencies.zip" | ||
| # You can upload your own model and paste the url here. |
There was a problem hiding this comment.
This a bit confusing, I don't think you mean that users can upload to artifactory themself.
| # You can upload your own model and paste the url here. | |
| # Optionally, put the URL to your own model here |
| # This will be required to import onnxruntime in the UDF. | ||
| dependencies_url = "https://artifactory.vgt.vito.be:443/auxdata-public/openeo/onnx_dependencies.zip" | ||
| # You can upload your own model and paste the url here. | ||
| model_url = "https://artifactory.vgt.vito.be:443/auxdata-public/openeo/test_onnx_model.zip" |
There was a problem hiding this comment.
This is apparently a zip file with just one .onnx file. Isn't it possible to just provide this .onnx file?
|
@JeroenVerstraelen it seems like these comments are easy enough to fix, can you have a look so we can merge this pr? |
Onnx allows users to create their models in e.g. tensorflow or pytorch and then save them to a common file format. They can then be executed in an openEO UDF without needing to provide the original libraries as dependencies.
Open-EO/openeo-geopyspark-driver#543
https://onnx.ai/