Papers

To ease the reproduction of papers, and the upload of full models on HuggingFace, the xpmir.papers package can be used.

Configuration

Papers experimental parameters are defined by data classes. The main one is xpmir.papers.helpers.PaperExperiment that defines an id (for experimaestro), a title and a description. These informations can be used e.g. when uploading the trained models on HuggingFace.

class xpmir.papers.helpers.PaperExperiment(*, id: str = '???', title: str = '', description: str = '')[source]

Bases: object

description: str

A description of the model

This description is used to generate reports

id: str

The experiment ID

The ID should be unique, and might be used when exporting the model e.g. to HuggingFace

title: str

The model title

The title is used to generate report

class xpmir.papers.helpers.NeuralIRExperiment(*, id: str = '???', title: str = '', description: str = '', gpu: bool = True, seed: int = 0)[source]

Bases: PaperExperiment

Settings most neural IR experiments

description: str

A description of the model

This description is used to generate reports

gpu: bool: Use GPU for computation

id: str

The experiment ID

The ID should be unique, and might be used when exporting the model e.g. to HuggingFace

seed: int: The seed used for experiments

title: str

The model title

The title is used to generate report

Helpers

Pipelines factorize the code necessary to run some experiments: for instance, training re-rankers on MS-Marco is usually performed with similar training data, evaluation datasets.

Implemented papers

This page give some information about reproduction of papers based on XPMIR.

xpmir/cross-encoders:
- monoBERT