mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2024-11-11 21:39:52 +00:00
1b67731e18
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 |
||
---|---|---|
.. | ||
CMakeLists.txt | ||
lookahead.cpp | ||
README.md |
llama.cpp/examples/lookahead
Demonstration of lookahead decoding technique: