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

Zero Shot Bert Adapters

Zero Shot Bert Adapters is a Python library implementing Z-BERT-A, a two-stage multilingual intent-discovery pipeline that uses an adapter-fine-tuned Transformer trained for natural-language inference to classify unknown intents in a zero-shot setting.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Zero Shot Bert Adapters — 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 building or evaluating a Python NLP workflow that needs to surface intent categories not already represented in its label set.

Who it is for

  • NLP developers exploring unknown-intent detection in text datasets.
  • Teams investigating candidate intent labels for support or other message-classification workflows.

Documented uses

  • Find candidate intents in utterances without predefining every intent category.
  • Classify unknown intents in a documented Banking77 example.
  • Train and test a model using the README’s SNLI example.
Before you choose

Requirements & limits

The repository is public and MIT-licensed. Installation is documented through pip from GitHub.

  • The README demonstrates Banking77 for intent discovery and SNLI for natural-language-inference training.
  • The documented stack includes PyTorch, Adapter Transformers and spaCy with en_core_web_trf.
  • The README requires wget and specifies separate installation by Windows users.
  • The example selects CUDA when available and otherwise falls back to CPU.

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 the library from its GitHub repository

Install Zero Shot Bert Adapters using the repository’s documented pip command.

Before you start

  • Python environment with pip
  • wget; Windows users must install it separately
  • spaCy with en_core_web_trf; the package attempts to install it if absent

Steps

  1. Run the command below.

    Documented command
    pip install git+https://github.com/GT4SD/zero-shot-bert-adapters.git

  2. Check that spaCy’s en_core_web_trf package is available, following spaCy’s installation guidance if needed.

Expected outcome

The repository’s documented installation step is completed, with the stated spaCy and wget requirements addressed.

Guide limitations

  • The guide covers installation only, not model training or inference.
  • The README does not give platform-specific wget installation commands.
Documentation 1 ↗
Source-based practical guide

Run the documented unknown-intent example

Pass a sentence to the Z-BERT-A pipeline’s find_new_intents example.

Before you start

  • Installed Zero Shot Bert Adapters library
  • A model initialized as shown in the README
  • PyTorch; the example chooses CUDA when available and otherwise CPU

Steps

  1. Import PyTorch, ZBERTA and the repository’s model module.

  2. Set the model name to bert-base-uncased and select CUDA when available or CPU otherwise.

  3. Initialize the model with the README’s example parameters and create a ZBERTA instance.

  4. Call find_new_intents with the example sentence: “I want to buy a book but I lost all my money, where can I make a withdrawal?”

Expected outcome

The README’s example output is the intent label `make withdrawal`.

Guide limitations

  • The sample depends on the model setup shown in the README, including a model path of model.pt.
  • This is a documented example, not a verified execution or a performance guarantee.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
GT4SD/zero-shot-bert-adapters · Official homepage title · retained in its source language
How we review sourcesSuggest a correction