til/python/use-post-init-for-dataclass-validations.md

Use __post_init__ For dataclass Validations

The dataclass construct is a handy stdlib way of modeling some data with many improvements over a dict such as named attributes and type visibility.

from dataclasses import dataclass
from typing import ClassVar

@dataclass
class BPEConfig:
    BASE_VOCAB_SIZE: ClassVar[int] = 256

    vocab_size: int
    special_tokens: list[str]

I want to enhance BPEConfig a little by validating the vocab_size which cannot be less than the BASE_VOCAB_SIZE. The __post_init__ method is a good place for this kind of validation.

from dataclasses import dataclass
from typing import ClassVar

@dataclass
class BPEConfig:
    BASE_VOCAB_SIZE: ClassVar[int] = 256

    vocab_size: int
    special_tokens: list[str]

    def __post_init__(self):
        if self.vocab_size < self.BASE_VOCAB_SIZE:
            msg = f"vocab_size ({self.vocab_size}) must be greater than or equal to BASE_VOCAB_SIZE ({self.BASE_VOCAB_SIZE})"
            raise ValueError(msg)

With this in place, my program will fail fast if I try to use an invalid vocab_size:

>>> BPEConfig(22, [])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<string>", line 5, in __init__
  File "/Users/lastword/dev/misc/build-an-llm/chapter_02/bpe_tokenizer.py", line 24, in __post_init__
    raise ValueError(msg)
ValueError: vocab_size (22) must be greater than or equal to BASE_VOCAB_SIZE (256)

This example is pulled directly from the BPETokenizer I'm building.