Quentin Anthony

Zamba is a hybrid language model combining state-space and transformer architectures.

Zyda, a dataset with 1.3 trillion tokens, enhances language model training.

Tree Attention improves efficiency in processing long sequences for machine learning models.

A study on enhancing data sharing in transformer model training.

New compression techniques speed up training for large language models while maintaining accuracy.

RedPajama datasets aim to enhance language model training through transparency and quality data.