10-14-2026, 04:11 AM
Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly.
small models can still be good
|
speculative decoding in llama.cpp
|
|
« Next Oldest | Next Newest »
|