Petals

Petals is a decentralized platform that enables users to run and fine-tune large language models (LLMs) on their personal hardware, leveraging a BitTorrent-style network. It provides an API-like interface for interacting with models such as Llama, Mixtral, and Falcon, facilitating tasks like text generation and custom model adaptation. The project emphasizes community contribution by allowing users to share their GPUs to host parts of these models.

Petals is a distributed system for running large language models, allowing users to run LLMs at home, BitTorrent-style. It explicitly mentions generating text with various LLMs (Llama, Mixtral, Falcon, BLOOM) and fine-tuning them, which aligns with 'conversational-ai' and 'fine-tuning-and-custom-models'. The project provides an API-like interface with PyTorch and Hugging Face Transformers, directly supporting 'api-access'. While not explicitly stated as a core feature, the ability to run LLMs and interact with them for various tasks, including those that could involve code, suggests 'code-generation' is possible through the platform's flexibility. The platform's focus on distributed inference and fine-tuning of large models aligns well with the concept of an AI SaaS platform for developers and researchers.