I'm excited to share the MoD 150k subset, a selection from the broader Mixture of Data project I've been working on. This subset is crafted for those looking to fine-tune AI models for both Mixture of Experts (MoE) architectures and standard architectures, with a keen eye on accessibility for those with limited computational resources.
After diving deep into MoEs and conducting various experiments, I've found this 150k subset not only… See the full description on the dataset page:
https://huggingface.co/datasets/Crystalcareai/MoD-150k.
