qfeval-functions

qfevalは、Preferred Networks 金融チームが開発している、金融時系列処理のためのフレームワークです。 データ形式の仕様定義、金融時系列データを効率的に扱うためのクラス/関数群、および金融時系列モデルの評価フレームワークが含まれます。

qfeval-functionsは、qfevalの中でも、金融時系列データを効率的に扱うための関数群を提供します。

qfeval is a framework developed by Preferred Networks' Financial Solutions team for processing financial time series data. It includes: data format specification definitions, a set of classes/functions for efficiently handling financial time series data, and a framework for evaluating financial time series models.

qfeval-functions specifically provides a collection of functions within qfeval that facilitate efficient processing of financial time series data.

Installation

pip install qfeval-functions

Usage

TBD

Pitfalls

Calling qfeval_functions.random.seed without fast=True may slow down qfeval_functions.functions.rand*

Calling qfeval_functions.random.seed without fast=True makes QF.rand* functions use reproducible random number generators implemented on CPU. Setting fast=True lets qfeval_functions.functions.rand* functions to use random number generators provided by CUDA via PyTorch. It is fast, but it is closed source. It cannot be reproducible on CPUs, and it may not be reproducible with other PyTorch/CUDA versions. For reproducibility, calling qfeval_functions.random.seed forces to use a reproducible way by default.