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 c6d01541aa
commit 4c47527130
104 changed files with 549 additions and 146 deletions

View file

@ -5,6 +5,8 @@ from typing import Optional
from ..dataset import ProceduralDataset
from ..factory import register_dataset
DATASET_NAME = "arc_1d"
@dataclass
class Arc1DConfig:
@ -100,6 +102,8 @@ class Arc1DDataset(ProceduralDataset):
"question": question,
"answer": " ".join(str(x) for x in test_example["output"]),
"metadata": {
"source_dataset": DATASET_NAME,
"source_index": idx,
"task_name": task_name,
"size": size,
"train_examples": train_examples,
@ -109,4 +113,4 @@ class Arc1DDataset(ProceduralDataset):
# Register the dataset
register_dataset("arc_1d", Arc1DDataset, Arc1DConfig)
register_dataset(DATASET_NAME, Arc1DDataset, Arc1DConfig)