Revert "MAINT: update DarkBench dataset (#844)"

[romanlutz]

This reverts commit 3da2bae.

Description

@Akash190104 suggested we update the reference to the DarkBench dataset. However, that causes the load_dataset function to fail. Backing out this change until it's fixed. FYI @paulinek13 I needed to back out since we had to do a release just now and this was a blocker. But once we make the required changes we definitely want that back in. I tried reproducing locally and it seemed to fail on the load_dataset step but a more detailed analysis might be needed.

For v0.8.1 this is already done.

Tests and Documentation

N/A

[paulinek13]

Oh, that's strange. I can't seem to reproduce it locally, regardless of the Python version I use. I tested with 3.10, 3.11, 3.12, and 3.13 (I also re-installed the deps with pip install -e .[dev]). For reference, I tested it at 2c9d7fc

🤔

Edit: Well, I couldn't reproduce it since it's using the latest cached version by default 😅, and I didn't disable it 😅. When I do, I get:

Using the latest cached version of the dataset since apart/darkbench couldn't be found on the Hugging Face Hub
Found the latest cached dataset configuration 'default' at ...

I'll certainly look into why apart/darkbench cannot be found

[jsong468]

Oh, that's strange. I can't seem to reproduce it locally, regardless of the Python version I use. I tested with 3.10, 3.11, 3.12, and 3.13 (I also re-installed the deps with pip install -e .[dev]). For reference, I tested it at 2c9d7fc

🤔

Edit: Well, I couldn't reproduce it since it's using the latest cached version by default 😅, and I didn't disable it 😅. When I do, I get:

Using the latest cached version of the dataset since apart/darkbench couldn't be found on the Hugging Face Hub
Found the latest cached dataset configuration 'default' at ...

I'll certainly look into why apart/darkbench cannot be found

Oh, that's strange. I can't seem to reproduce it locally, regardless of the Python version I use. I tested with 3.10, 3.11, 3.12, and 3.13 (I also re-installed the deps with pip install -e .[dev]). For reference, I tested it at 2c9d7fc

🤔

Edit: Well, I couldn't reproduce it since it's using the latest cached version by default 😅, and I didn't disable it 😅. When I do, I get:

Using the latest cached version of the dataset since apart/darkbench couldn't be found on the Hugging Face Hub
Found the latest cached dataset configuration 'default' at ...

I'll certainly look into why apart/darkbench cannot be found

Seems to work for me if I use data = load_dataset("apart/darkbench", "default", split="train", data_files="darkbench.tsv")!

[romanlutz]

@jsong468 is fixing this in #862 so I'll close this