Python implementation of the paper "Limits of Transformer Language Models on Learning to Compose Algorithms", NeurIPS 2024.

Dolomite Engine is a library for pretraining/finetuning LLMs

Chat Complex PDF with Tables Using IBM WatsonX, Langchain and LlamaParser.

Open source project for data preparation of LLM application builders

Storing code used in Generative AI Developer Guides on the IBM Developer Website

ModuleFormer is a MoE-based architecture that includes two different types of experts: stick-breaking attention heads and feedforward experts. We released a collection of ModuleFormer-based Language Models (MoLM) ranging in scale from 4 billion to 8 billion parameters.

🦄 Unitxt: a python library for getting data fired up and set for training and evaluation

Dromedary: towards helpful, ethical and reliable LLMs.

Ansible for LMT

Boolean Question Answering with multi-task learning and uses large LM embeddings like BERT, RoBERTa

Cloud Pak for Network Automation (CP4NA) (previously ALM/TNCO) command line tool