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

Koelectra

KoELECTRA is a family of pretrained Korean-language ELECTRA models distributed as checkpoints for use with the Transformers library. It includes Base and Small generator and discriminator variants in v1, v2 and v3, plus TensorFlow v1 archives. The models use WordPiece tokenization and were trained with ELECTRA’s replaced-token detection method.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Koelectra — 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 KoELECTRA when a Korean NLP project needs downloadable pretrained ELECTRA weights and can work with the Transformers library.

Who it is for

  • Developers building Korean-language text-processing systems.
  • Researchers prototyping Korean NLP tasks with pretrained models.

Documented uses

  • Load a pretrained Korean model and tokenizer through Transformers.
  • Use a discriminator as the starting point for fine-tuning.
  • Explore Korean text classification and token-level prediction workflows.
Before you choose

Requirements & limits

The project provides public model checkpoints and a public GitHub repository. The repository lists an Apache-2.0 license.

  • For fine-tuning, the README says to use the discriminator.
  • The repository offers Base and Small variants, with versioned checkpoints and different vocabulary sizes.
  • Transformers is the documented usage path; the README gives a TensorFlow 2 loading example for the v3 Base checkpoint.

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

Load a KoELECTRA discriminator with Transformers

Load a pretrained KoELECTRA model using the Transformers library.

Before you start

  • Transformers with ElectraModel support; the README specifies support from version 2.8.0.
  • A KoELECTRA checkpoint identifier, such as monologg/koelectra-base-v3-discriminator.

Steps

  1. Import ElectraModel and ElectraTokenizer from transformers.

  2. Load the tokenizer and model using the same KoELECTRA discriminator checkpoint identifier.

Expected outcome

A pretrained KoELECTRA discriminator and its tokenizer are loaded through Transformers.

Guide limitations

  • The documented example loads a model; it does not define a complete classification or token-prediction application.
  • The README notes that ElectraModel does not return pooled_output.
Documentation 1 ↗
Source-based practical guide

Load the v3 Base checkpoint with TensorFlow 2

Load the v3 Base KoELECTRA model with the documented TensorFlow 2 Transformers interface.

Before you start

  • Transformers with TFElectraModel support.
  • The checkpoint identifier monologg/koelectra-base-v3-discriminator.

Steps

  1. Import TFElectraModel from transformers.

  2. Load the checkpoint with TFElectraModel.from_pretrained("monologg/koelectra-base-v3-discriminator", from_pt=True).

Expected outcome

The TensorFlow 2 model is loaded from the documented PyTorch checkpoint.

Guide limitations

  • The README provides this example for the v3 Base discriminator; it does not document a full downstream task workflow.
Documentation 1 ↗
Evidence and transparency

Sources & review

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