fix(envs): Add source dataset and index to metadata (#388)

* add source dataset and index to metadata

* fix typo

* fix coach class and its test
This commit is contained in:
Zafir Stojanovski 2025-03-20 12:12:14 +01:00 committed by GitHub
parent 7475a20700
commit ce0a6c4878
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
104 changed files with 549 additions and 146 deletions

View file

@ -8,6 +8,8 @@ from typing import Any, Optional
from ..coaching import BaseCurriculum, RangeAttributeDefinition
from ..factory import ProceduralDataset, register_dataset
DATASET_NAME = "number_sorting"
@dataclass
class NumberSortingConfig:
@ -90,6 +92,8 @@ Please follow the instruction below:
"question": question,
"answer": str(answer),
"metadata": {
"source_dataset": DATASET_NAME,
"source_index": idx,
"original_numbers": number_strs,
"direction": direction,
"sorted_numbers": answer,
@ -198,4 +202,4 @@ class NumberSortingCurriculum(BaseCurriculum):
)
register_dataset("number_sorting", NumberSortingDataset, NumberSortingConfig, NumberSortingCurriculum)
register_dataset(DATASET_NAME, NumberSortingDataset, NumberSortingConfig, NumberSortingCurriculum)