We've open-sourced FlashKDA, our high-performance CUTLASS-based implementation of Kimi Delta Attention kernels.
It delivers 1.72×–2.22× prefill speedup over the flash-linear-attention baseline on H20, and works as a drop-in backend for flash-linear-attention.
Explore on GitHub:
https://t.co/QScjWsJqSy
hey guys, build a smol fun side project. gives you a fun personalized rpg build for your cinema taste.
you can add manually and even better, easily connect to your lbxd account with just your username
try it out here: https://t.co/0W1mvd2ygs
hey guys, build a smol fun side project. gives you a fun personalized rpg build for your cinema taste.
you can add manually and even better, easily connect to your lbxd account with just your username
try it out here: https://t.co/0W1mvd2ygs
hey guys whats up, i am back.
a lil update on where i have been, got an internship, didnt like that one, left, got another one in a reputable unicorn as an AI engineer, converted it.
kinda ridiculous how much of modern life now runs through one pocket computer, which now has intelligence built right into it.
comms. relationships. work. money. maps. photos. rent. rides. food. dating. identity. all of it mediated through one small slab of glass.
if you lose your phone for even few hours you basically become a medieval peasant at this point.
the hunger is still there, studying, making out time to work, balance life, these past few months have been thrilling and a lot has happened. me and my friend are working on things side by side, propelling it. we will release one this week for all the bengaluru folks!!