> To leverage the capabilities of MI250X, ROCm enables the use of GPU matrix cores through its rocBLAS and MIOpen library implementations that, in turn,are leve

> To leverage the capabilities of MI250X, ROCm enables the use of GPU matrix cores through its rocBLAS and MIOpen library implementations that, in turn,are leveraged by PyTorch. - https://aclanthology.org/2023.emnlp-main.164.pdf https://aclanthology.org/2023.emnlp-main.164.pdf https://github.com/TurkuNLP/ https://github.com/TurkuNLP/