Failure with RAC and more than one text in pipeline

[pmbaumgartner]

If you change the example in https://huggingface.co/knowledgator/gliclass-base-v2.0-rac-init so that it contains more than one text, the pipeline returns:

  File "/Users/pbaumgartner/Library/Caches/uv/environments-v2/gliclass-demo-9ee78d962bb2dcd2/lib/python3.12/site-packages/gliclass/utils.py", line 42, in <genexpr>
    label for example in examples for label in example.get("true_labels", [])
                                               ^^^^^^^^^^^
AttributeError: 'str' object has no attribute 'get'

Replicate by changing this line and putting text in a list:

results = pipeline([text], labels, threshold=0.1, rac_examples=[example_1, example_2])[0]

[BioMikeUkr]

Hello, when you predict for multiple texts, the number of examples should be equal to the number of texts (an example in such a case could be a list with just one dictionary or a list of up to 3 dictionaries).

results = pipeline([text], labels, threshold=0.1, rac_examples=[[example_1, example_2]])[0]

or

results = pipeline([text, text], labels, threshold=0.1, rac_examples=[[example_1, example_2], [example_1, example_2]])[0]

or

results = pipeline([text, text], labels, threshold=0.1, rac_examples=[[example_1], [example_1, example_2]])[0]