diff --git a/common/ngram-map.cpp b/common/ngram-map.cpp index 7e31ba054a..8eac44189e 100644 --- a/common/ngram-map.cpp +++ b/common/ngram-map.cpp @@ -23,7 +23,7 @@ llama_tokens common_ngram_simple_draft( common_ngram_simple_state & state, const llama_tokens & tokens, llama_token sampled) { - // Simple implementation of self-speculative decoding without draft model, without ngram-map. + // Simple implementation of self-speculative decoding without a draft model. // const size_t cur_len = tokens.size(); // Only check every check_rate tokens to save compute diff --git a/common/speculative.cpp b/common/speculative.cpp index a13f9c7f96..c4cf38537f 100644 --- a/common/speculative.cpp +++ b/common/speculative.cpp @@ -119,7 +119,7 @@ struct common_speculative_state { // TODO: track performance of most recent calls const bool gen_perf = true; // whether to generate performance stats. - int64_t gen_duration_us = 0; // total time spent in this implementation in milliseconds. + int64_t gen_duration_us = 0; // total time spent in this implementation in microseconds. virtual ~common_speculative_state() = default;