By making OLMo and its training data fully accessible to the public, AI2 aims to collaboratively build the best open language model in the world. The framework features a suite of open AI development tools, including full pretraining data, training code and model weights, and an evaluation suite. In the coming months, AI2 plans to continue iterating on OLMo and will introduce different model sizes, modalities, datasets, and capabilities into the OLMo family.
Key takeaways:
- The Allen Institute for AI (AI2) has released OLMo 7B, an open, state-of-the-art large language model with the pre-training data and training code.
- OLMo and its framework are designed to aid researchers in training and experimenting with large language models, and are available for direct download on Hugging Face and GitHub.
- By making OLMo and its training data fully available to the public, AI2 aims to collaboratively build the best open language model in the world, and will continue to iterate on OLMo in the coming months.
- With OLMo, AI researchers and developers will have full insight into the training data behind the model, which will reduce developmental redundancies and enable scientific testing of the model's performance.