agenticup

6 ms·

1. 1.▲
   by agenticup 3mo ago

   i guess inference engineering, like dpsark or dflash specific speculative decoding technqiues
2. 2.▲
   by agenticup 3mo ago

   can the orniths self scaffolding could learn to scaffold the rlm loop?
3. 3.▲
   by agenticup 3mo ago

   qwen 3.6 27b and qen35b a3b work like magic, if we get dpark speculative decoding versions of these models it will further improve the throughput