mirror of
https://github.com/ikawrakow/ik_llama.cpp.git
synced 2026-01-27 09:39:53 +00:00
Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
llama.cpp/examples/speculative
Demonstration of speculative decoding and tree-based speculative decoding techniques
More info: