Source code for s3prl.dataio.encoder.category

"""
Simple categorical encoder

Authors:
  Leo 2022
"""

from typing import List


[docs]class CategoryEncoder: def __init__(self, category: List[str]) -> None: self.category = list(sorted(set(category))) def __len__(self) -> int: return len(self.category)
[docs] def encode(self, label: str) -> int: return self.category.index(label)
[docs] def decode(self, index: int) -> str: return self.category[index]
[docs]class CategoryEncoders: def __init__(self, categories: List[List[str]]) -> None: self.categories = [CategoryEncoder(c) for c in categories] def __len__(self) -> int: return sum([len(c) for c in self.categories]) def __iter__(self): for category in self.categories: yield category
[docs] def encode(self, labels: List[str]) -> List[int]: assert len(labels) == len(self.categories) return [ encoder.encode(label) for label, encoder in zip(labels, self.categories) ]
[docs] def decode(self, indices: List[int]) -> List[str]: return [ encoder.decode(index) for index, encoder in zip(indices, self.categories) ]