TGTGInsighttelegram intelligenceLIVE / telegram public index
← Python Заметки

TGINSIGHT SIMILAR POSTS

Најди сличен содржај

Изворен канал @pythonotes · Post #397 · 12 ное.

Использование Pydantic сегодня стало нормой, и это правильно. Но иногда на ревью вижу, что используют его не всегда корректно. Например, метод BaseModel.model_dump() по умолчанию не преобразует стандартные типы, такие как datetime, UUID или Decimal, в простой сериализуемый для JSON вид. Тогда пишут кастмоный сериализатор для этих типов чтобы функция json.dump() не падала с ошибкой. import uuid from datetime import datetime from decimal import Decimal from uuid import UUID from pydantic import BaseModel class MyModel(BaseModel): id: UUID date: datetime value: Decimal obj = MyModel( id=uuid.uuid4(), date=datetime.now(), value='1.23' ) print(obj.model_dump()) # не подходит для json.dump # { # 'id': UUID('4f8c1bc4-25fd-40cd-9dbe-2c73639b0dc1'), # 'date': datetime.datetime(2025, 12, 12, 12, 12, 12, 111111), # 'value': Decimal('1.23') # } # добавляем свой кастомный сериализатор json.dumps(obj.model_dump(), cls=MySerializer) # { # 'id': '4f8c1bc4-25fd-40cd-9dbe-2c73639b0dc1', # 'date': '2025-12-12T12:12:12.111111', # 'value': '1.23' # } В данном случае класс MySerializer обрабатывает datetime, UUID и Decimal. Например так: class MySerializer(json.JSONEncoder): def default(self, o): if isinstance(o, Decimal): return str(o) elif isinstance(o, datetime): return o.isoformat() elif isinstance(o, UUID): return str(o) return super().default(o) Специально для тех, кто всё еще так делает - в этом нет необходимости! Pydantic может это сделать сам, просто нужно добавить параметр mode="json". json.dumps(obj.model_dump(mode="json")) # { # 'id': '4f8c1bc4-25fd-40cd-9dbe-2c73639b0dc1', # 'date': '2012-12-12T12:12:12.111111', # 'value': '1.23' # } #pydantic#libs

Резултати

Пронајдени 1 слични објави

Пребарај: #copyrightdata

当前筛选 #copyrightdata清除筛选
AI & Law

@ai_and_law · Post #202 · 28.12.2023 г., 08:04

USA: New Bill Proposes Transparency Rules for AI Training Data Hello everybody! In US a new AI Foundation Model Transparency Act have been introduced. The aim is to ensure transparency in the usage of copyrighted data in training AI models. Companies developing foundation models must reveal the sources of their training data. This move is geared towards informing copyright holders if their data has been utilized. The bill tasks the Federal Trade Commission (FTC) to collaborate with the National Institute of Standards and Technology (NIST) in establishing rules for reporting training data transparency. Companies must disclose the origin of training data, how it's retained during inference, model limitations or risks, alignment with NIST’s AI Risk Management Framework, and adherence to federal standards. AI developers are required to report efforts to "red team" the model, especially in areas like medical queries, cybersecurity, elections, and services impacting vulnerable populations. This bill complements the Biden administration’s AI executive order, reinforcing the push for standardized reporting in AI models. The bill is yet to go through committee discussions, and its fate before the election campaign season remains uncertain. #AIlaw#CopyrightData#Legislation#TechNews#AIRegulation