Indeed :P Honestly I'm not sure how context "sharding" works on multiple GPUs atm. Decent, really long context OSS models like Yi 200K and YARN finetunes are v

Indeed :P Honestly I'm not sure how context "sharding" works on multiple GPUs atm. Decent, really long context OSS models like Yi 200K and YARN finetunes are very new.