OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Ratransformers

RATransformers is a Python library built on Hugging Face Transformers that modifies transformer self-attention to incorporate relation-aware input features during training or fine-tuning. Its documented example encodes table relationships for tabular question answering.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Ratransformers — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a transformer training workflow needs explicit relations in its input, such as table-cell-to-column and same-row relationships for tabular question answering.

Who it is for

  • Developers training or fine-tuning transformer models with structured relationships in their inputs.
  • Researchers prototyping table question-answering or comparing plain-text and relation-enriched inputs.

Documented uses

  • Represent relationships between table cells, columns and rows in model inputs.
  • Prepare supported transformer models for training with selected relation kinds.
  • Add features associated with individual input words or tokens.
Before you choose

Requirements & limits

  • It is a developer library, not a hosted model-training service; the README provides a Python installation and initialization workflow.
  • The supported model list is limited to six named architectures.
  • The example requires developers to specify relation kinds and a model class; a tokenizer name or path can be supplied separately when it differs from the model.
  • The repository identifies the project as open source under the MIT license.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install RATransformers and prepare a model for training

Install the package and initialize a supported transformer with selected relation kinds, model class and tokenizer.

Before you start

  • Python environment with the package and its Hugging Face Transformers dependencies available.
  • A supported model and a model class; the README example uses AutoModelForSequenceClassification.
  • Relation kinds to encode, and a tokenizer name or path if it differs from the model.

Steps

  1. Install the package with the command below.

    Documented command
    pip install ratransformers

  2. Import RATransformer and the chosen model class.

  3. Initialize RATransformer with a model name, relation_kinds, model_cls and, if needed, pretrained_tokenizer_name_or_path.

  4. Retrieve the initialized model and tokenizer from ratransformer.model and ratransformer.tokenizer.

Expected outcome

A RATransformer model and tokenizer prepared for a training workflow.

Guide limitations

  • The README lists BART, BERT, GPT-2, RoBERTa, T5 and LongT5 as supported; it does not establish support for other architectures.
  • The setup example does not document a complete training run or its results.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
JoaoLages/RATransformers · Official homepage title · retained in its source language
How we review sourcesSuggest a correction