echodict/llama.cpp

1# llama.cpp/examples/lookup2 3Demonstration of Prompt Lookup Decoding4 5https://github.com/apoorvumang/prompt-lookup-decoding6 7The key parameters for lookup decoding are `ngram_min`, `ngram_max` and `n_draft`. The first two determine the size of the ngrams to search for in the prompt for a match. The latter specifies how many subsequent tokens to draft if a match is found.8 9More info:10 11https://github.com/ggml-org/llama.cpp/pull/448412https://github.com/ggml-org/llama.cpp/issues/422613