Dataset metadata lost when loading evaluations with ModelCard.load()

[lewtun]

Describe the bug

When loading evaluation metadata with the ModelCard API, it seems some important information like dataset_config and dataset_split is lost.

For example, this model card has evaluations where both the dataset config and split are specified.

However, when loading the metadata with ModelCard.load(), these fields are absent since the metadata appears to be grouped under a single (task, dataset) and this optional information isn't collected. See below for a reproducible example.

Reproduction

Here's one way to see the problem:

from huggingface_hub import ModelCard

card = ModelCard.load("philschmid/bart-large-cnn-samsum")
# Expect to see dataset config/test fields here
card.data

Alternatively, we can run a check on the evaluation data itself:

from huggingface_hub import ModelCard

card = ModelCard.load("philschmid/bart-large-cnn-samsum")
metadata = card.data.to_dict()
eval_results = model_index_to_eval_results(metadata["model-index"])[1]
assert any(eval_result.dataset_config is not None for eval_result in eval_results) is True, "at least one config should exist"

Logs

No response

System info

- huggingface_hub version: 0.11.0
- Platform: Linux-4.19.0-22-cloud-amd64-x86_64-with-glibc2.10
- Python version: 3.8.13
- Running in iPython ?: Yes
- iPython shell: ZMQInteractiveShell
- Running in notebook ?: Yes
- Running in Google Colab ?: No
- Token path ?: /home/lewis_huggingface_co/.huggingface/token
- Has saved token ?: True
- Who am I ?: lewtun
- Configured git credential helpers: store
- FastAI: N/A
- Tensorflow: N/A
- Torch: 1.11.0
- Jinja2: 3.1.2
- Graphviz: N/A
- Pydot: N/A

[Wauplin]

Fixed in #1214 by @lewtun !