Chapter 4 --- Supervised Classification

[shravanidev]

When using SentenceTransformer.encode with a Column object from a Hugging Face datasets dataset (e.g., data["train"]["text"]), a TypeError occurs because the method attempts to index the Column object with numpy.int64 indices, which are not supported by datasets. This happens during the sentence sorting step in encode. Converting the Column to a Python list (e.g., list(data["train"]["text"])) resolves the issue, but native support for datasets Column objects would improve compatibility.

[MaartenGr]

Thank you for sharing this! I can take a look at this and fix where appropriate (especially wrt versioning of datasets. Do you have a preferred solution in mind?