3 ms·

Has anyone tried training yet and are there any obvious pitfalls for multi GPU training like there were in mamba-1?