mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2025-04-20 21:46:07 +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/lookahead
Demonstration of lookahead decoding technique: