sdg_op component fails with RHELAI 1.3

[MichaelClifford]

sdg_op with RHEL 1.3 fails at line 76 due to a change in how read_taxonomy now works. It now expects an additional parameter document_output_dir to be present.

ilab-on-ocp/sdg/components.py

Line 76 in e57915f

print(read_taxonomy(taxonomy_path, taxonomy_base))

This is the error:

TypeError: expected str, bytes or os.PathLike object, not NoneType

That said, it is unclear to me that line 76 here is doing much. I recommend simply removing it as read_taxonomy is called inside of the generate_data function anyways.

[tumido]

This error is interesting... Since the source code in the open source ILab says this param is fully optional..

https://github.com/instructlab/sdg/blob/eef8baedf36f21da73bda39270ed7378558541ad/src/instructlab/sdg/utils/taxonomy.py#L314-L319

[MichaelClifford]

Update:

Looks like /usr/share/instructlab/sdg/default_data_recipes/skills.yaml is no longer editable by the user. This causes the set_precomputed_skills_data_ratio() function that limits the size of the skills data used to raise an error: PermissionError: [Errno 13] Permission denied:

I will investigate if there is another work around for this.

[leseb]

Update:

Looks like /usr/share/instructlab/sdg/default_data_recipes/skills.yaml is no longer editable by the user. This causes the set_precomputed_skills_data_ratio() function that limits the size of the skills data used to raise an error: PermissionError: [Errno 13] Permission denied:

I will investigate if there is another work around for this.

#230 (comment)

[MichaelClifford]

Closed by #230