Catalog

The AI4EOSC catalog gathers the different types of AI assets that the platform currently offers.

AI Modules

The AI Modules catalog is a set of AI models designed to perform given AI tasks (eg. image classification). The catalog shows both models developed specifically for the platform, as well as external model catalogs that have been integrated to make them deployable in the platform.

You can use filters to quickly find the module you want. We provide filtering by:

• Libraries: this is the Deep Learning the module uses (eg. Pytorch, Tensorflow)

• Tasks: this is the broad task the module addresses (eg. Computer Vision, Natural Language Processing)

• Platform Categories: platform-specific tags. Current options include:

  • AI4 pretrained: modules that already come with a trained AI model

  • AI4 inference: modules that can be used for inference (usually overlaps with AI4 pretrained)

  • AI4 trainable: modules that can be trained on a new dataset

  For example the image classifier is both pretrained (because is comes with a model trained on ImageNet), inference-ready (because it allows to use that model to do predictions) and trainable (because it allows to retrain that model on a different dataset, to create for example a plant classifier )

• Data type: type of data the module processes (eg. Image, Audio)

• Tags: any module-specific tag

Tools

The Tools catalog is a collection of tools that come handy in assisting users during the whole Machine Learning workflow cycle.

We currently offer for example:

• The Development Environment to develop new modules (Develop a model from scratch)

• Several Federated Learning Servers to make federated privacy-friendly trainings (Flower, NVFLARE)

• The Computer Vision Annotation Tool (CVAT) to label images (Labeling images with CVAT)

LLMs

The LLM catalog is a collection of LLM models that users can leverage for generative tasks.

There are currently two kinds of LLMs:

• Platform-wise LLMs: those are LLMs that are permanently deployed for all platform users. For more information on how to use them, please refer to AI4EOSC LLM documentation.

• Self-deployable LLM: those are LLMs that are deployable per user. The catalog has multiple open-source models from a wide range of AI-relevant actors (Qwen, DeepSeek, Meta). For more information on how to use them, please refer to the tutorial: Deploy your own LLM chatbot.