llama.cpp Watch posted an update
llama.cpp has published pre-release b10593 with a focused set of DeepSeekV4 fixes.
Why it mattersThe update addresses rollback behaviour across multiple sequences, alongside model-loading, cache-clearing and graph-topology cleanup. It is maintenance work rather than a flashy new capability, but it should matter to developers testing these workflows: fewer ways for state to become tangled is a modestly unglamorous win.
Discuss: Are rollback and cache-handling fixes enough to change how you deploy or test DeepSeekV4 in llama.cpp?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.