OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Commonsense Rc

Commonsense Rc is a public Python research repository containing an attention-based LSTM model for commonsense machine comprehension, developed for SemEval-2018 Task 11. It includes instructions for preparing task data and training the model, plus a documented RACE-pretraining and multi-seed ensembling approach for reproducing the competition submission.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Commonsense Rc — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when investigating or reproducing this specific research approach and when you can work with its legacy software requirements and task data.

Who it is for

  • Researchers studying commonsense machine comprehension and attention-based LSTM models.
  • Developers who want to train or examine a SemEval-2018 Task 11 question-answering system.

Documented uses

  • Train the model on SemEval-2018 Task 11 data.
  • Prepare task data with GloVe embeddings and ConceptNet using the repository’s scripts.
  • Explore the documented RACE pretraining and multi-seed ensemble workflow.
Before you choose

Requirements & limits

The repository is public and specifies an MIT license.

  • The README specifies PyTorch 0.2–0.4 and spaCy 2.0 or later, and warns of incompatibility with Python 3.7 and later.
  • Training data can be downloaded preprocessed, or prepared using the repository’s scripts with GloVe and ConceptNet.
  • A GPU is preferred; CPU training is described as much slower.
  • The RACE pretraining and nine-model ensemble are separate steps in the documented competition-submission workflow.
  • Replacing the legacy PyTorch stack with a modern version requires modification; no modern-version port or test is supplied.

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

Train the model using preprocessed task data

Set up the documented data folder and start model training.

Before you start

  • Python 3 below 3.7, as the README warns that Python 3.7 and later will not work.
  • PyTorch 0.2, 0.3 or 0.4 and spaCy 2.0 or later.
  • Preprocessed task data from one of the README’s linked sources.
  • A GPU is preferred; CPU training is documented as much slower.

Steps

  1. Download the preprocessed data from Google Drive or Baidu Cloud Disk, unzip it, and place it under the repository’s data/ folder.

  2. Start training with the command below.

    Documented command
    python3 src/main.py --gpu 0

  3. The README says development-set accuracy will be approximately 83% after 50 epochs.

Expected outcome

The documented training process runs the model on the task data and reports development-set accuracy.

Guide limitations

  • The README flags a Python 3.7 compatibility issue and specifies legacy PyTorch versions.
  • Reported accuracy is an approximate README figure, not a guaranteed result.
  • The README notes that CPU training will be much slower.
Documentation 1 ↗
Evidence and transparency

Sources & review

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