LLM-And-More is an open-source, all-in-one toolkit from IceBearAI for taking a large language model from data to deployed service. It packages the complete workflow — data preparation, training, monitoring, evaluation, deployment and interaction — into a plug-and-play solution with production best practices baked in. The GitHub repository has around 380 stars and was last updated in 2024.
Building an LLM product involves far more than fine-tuning: data pipelines, distributed training configuration, evaluation harnesses and serving all need separate tooling. LLM-And-More organises that journey into six modules — data, training, monitoring, evaluation, deployment and interaction — so professional developers and business teams can follow one coherent path from idea to service. It integrates a high-performance model-parallel training framework to cut compute costs during training and inference, supports a range of mainstream open models, and lets users customise models and services to their own scenarios. The documentation (largely in Chinese) reflects its origin in the Chinese LLM community, and the codebase encodes scenario-specific know-how gathered from real deployments rather than presenting a bare training script.
The six modules can also be adopted independently — a team might start with the training module for fine-tuning and add evaluation and deployment later — and the interaction module turns a finished model into a callable service without a separate serving stack. Sample configurations for popular open models are included to shorten the path from cloning the repository to a first successful training run.
LLM-And-More is free and open source, with no paid tier or hosted edition. You supply your own compute for training and serving.
Developers and teams — especially in the Chinese LLM ecosystem — who want a single guided toolkit for fine-tuning open models and shipping them as services.