mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-06-28 15:20:20 +00:00
2e97c5f96f
* server: Never return 0.0 post-sampling probabilities * backend sampling: support returning post-sampling probs