Shypd.aiThe largest open index of AI tools
OpenELM is an open-source language model family that provides an efficient and transparent framework for training and inference. It uses a layer-wise scaling strategy for enhanced accuracy and includes a complete framework for training and evaluation on publicly available datasets.
Trending
About
OpenELM, developed by Apple Machine Learning Research, is a state-of-the-art open language model family designed to advance open research and ensure reproducibility and transparency in large language models. It employs a unique layer-wise scaling strategy to efficiently allocate parameters within each transformer layer, resulting in improved accuracy. For instance, OpenELM achieves a 2.36% accuracy improvement over OLMo with approximately one billion parameters, while requiring 2 times fewer pre-training tokens. Unlike other models that only provide weights and inference code, OpenELM offers a comprehensive framework for training and evaluation on publicly available datasets, including training logs, multiple checkpoints, and pre-training configurations. It also provides code to convert models for inference and fine-tuning on Apple devices using the MLX library, empowering the open research community.
Capabilities
Pricing & Plans
Open Source
Open Source
FAQs
Trending
