sampling: a reasoning budget of 0 bounds the think; a prefilled token cannot consume the forced end - #95

sampling: a reasoning budget of 0 bounds the think; a prefilled token cannot consume the forced end#95

joelteply merged 3 commits into

masterCambrianTech/llama.cpp:masterfrom

8b479782/reasoning-budget-zeroCambrianTech/llama.cpp:8b479782/reasoning-budget-zeroCopy head branch name to clipboard