Skip to content

[v2] Can't compute stats for reranking datasets #2972

@KennethEnevoldsen

Description

@KennethEnevoldsen

Describe the bug

@Samoed seems like I can't compute descriptive stats for reranking datasets:

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
File /Users/au56[1](https://file+.vscode-resource.vscode-cdn.net/Users/au561649/Github/mteb/tmp.py:1)649/Github/mteb/tmp.py:1
----> 1 mteb.get_task("JQaRAReranking").calculate_metadata_metrics()

File ~/Github/mteb/mteb/abstasks/AbsTask.py:313, in AbsTask.calculate_metadata_metrics(self, overwrite_results)
    311             descriptive_stats[split][hf_subset_stat][hf_subset] = split_details
    312     else:
--> [313](https://file+.vscode-resource.vscode-cdn.net/Users/au561649/Github/mteb/~/Github/mteb/mteb/abstasks/AbsTask.py:313)         split_details = self._calculate_metrics_from_split(split)
    314         descriptive_stats[split] = split_details
    316 with self.metadata.descriptive_stat_path.open("w") as f:

File ~/Github/mteb/mteb/abstasks/AbsTaskRetrieval.py:430, in AbsTaskRetrieval._calculate_metrics_from_split(self, split, hf_subset, compute_overall)
    426 if "default" in self.dataset and split != "default":
    427     return self._calculate_metrics_from_split(
    428         split=split, hf_subset="default"
    429     )
--> [430](https://file+.vscode-resource.vscode-cdn.net/Users/au561649/Github/mteb/~/Github/mteb/mteb/abstasks/AbsTaskRetrieval.py:430) split_data = self.dataset["default"][split]
    431 queries = split_data["queries"]
    432 corpus = split_data["corpus"]

KeyError: 'default'

The task in questions are:

XGlueWPRReranking # has a unique error 
JQaRAReranking
JaCWIRReranking

To reproduce

mteb.get_task("JaCWIRReranking").calculate_metadata_metrics() 
mteb.get_task("JQaRAReranking").calculate_metadata_metrics()
mteb.get_task("XGlueWPRReranking").calculate_metadata_metrics()

Additional information

No response

Are you interested to contribute a fix for this bug?

Yes

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions