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

Contrastors

Contrastors is an Apache-2.0 Python toolkit for training and evaluating contrastive machine-learning models, including text, image and multimodal embedding models.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Contrastors — 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 Contrastors when you need a Python training toolkit and can manage its CUDA, PyTorch and custom-kernel requirements, rather than a hosted embedding service.

Who it is for

  • Researchers developing contrastive learning methods or embedding models.
  • Engineers building dense-retrieval or multimodal retrieval systems.
  • Teams with access to a CUDA-capable training environment.

Documented uses

  • Pretrain or fine-tune text embedding models using contrastive objectives.
  • Train image embedding models with CLIP- or LiT-style learning.
  • Evaluate contrastive models and develop embeddings for dense retrieval or multimodal RAG.
Before you choose

Requirements & limits

The repository is public and licensed under Apache-2.0. Accessing the provided nomic-embed-text-v1 dataset requires an Atlas account and login through the Nomic Python client.

  • The documented setup requires CUDA 11.8 or later and custom Flash Attention kernels.
  • PyTorch and Hugging Face support are central parts of the documented model workflow.
  • The toolkit includes MLM pretraining, contrastive pretraining and fine-tuning, and flexible embedding sizes through Matryoshka Representation Learning.
  • Using the provided dataset involves separate account and storage-credential setup.

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

Prepare a CUDA environment and install Contrastors

Set up the documented Python environment and install the package and its Flash Attention kernels.

Before you start

  • CUDA 11.8 or later
  • Python 3 and a CUDA-appropriate PyTorch installation
  • wheel, packaging, ninja and setuptools

Steps

  1. Create and activate a Python virtual environment.

  2. Install PyTorch, torchvision and torchaudio using the PyTorch installation guidance for the system.

  3. Install wheel, packaging, ninja and setuptools.

  4. Install Flash Attention and the listed custom kernels, then install Contrastors in editable mode with pip.

Expected outcome

The documented package and its required kernels are installed in the virtual environment.

Guide limitations

  • The setup relies on custom Flash Attention kernels and requires CUDA 11.8 or later.
  • PyTorch installation details depend on the system.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
nomic-ai/contrastors · Official homepage title · retained in its source language
How we review sourcesSuggest a correction