Refer to file on mountpoint in Python script

KNIME Server
1

Hi everyone,

we would like to refer to a file (a Pytorch model) within the Knime Mountpoint in a Python script. We tried this by using knime://LOCAL like this:

model_path = ‘knime://LOCAL/data/models/model.model’
model.load_state_dict(torch.load(model_path, map_location=device))

This did not work. How can we refer to a file in the Knime workflow space in a Python script? Relative or absolute file paths did not work as well.

Thanks,
Johannes

2

Hi Johannes,
currently KNIME URLs cannot be resolved in Python, but there is a workaround that should help you. Can you use the List Files node with “knime://LOCAL/data/models/” as folder and “model.model” in the filter? The result is a table with the resolved path and you can turn that into a flow variable using Table Row to Variable. The variable can then be passed into the Python Snippet.
Kind regards,
Alexander

3 Likes
3

Thanks for your answers. Because we are working on the Knime Server, the LOCAL method is not working for us. I did not mention this before, sorry.

Currently, we are trying to load some pre-trained models which are saved on the server within the Knime mountpoint using Python.

We tried the following

  • absolute file system path (srv/knime_server/workflows/…)
  • relative paths starting from the workflow (/…/models/mode.model)

Is there anything else we could try? When I run the same script outside of Knime on the server it works without any problems. I think there is an issue when I run Python within Knime.

Thanks
Johannes

4

Hi,
instead of local you can also just write knime.mountpoint to reference the current mountpoint.
Kind regards,
Alexander

5

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.