# llama.cpp ngram-mod and ngram-map-k4v drafter internals > ngram-mod stores one `int32` per slot and has no key check. A slot is chosen by `res = res*6364136223846793005 + token` over the n tokens, then `% size`. A lookup returns whatever token was last written to that slot, so two different n-grams that hash to one slot overwrite each other and can retu... Parent: [Mac local LLMs: Speculative decoding and MTP](https://llms-explorer.com/tree/mac-local-llms-speculative-decoding-and-mtp/) · 1 facets · 21 facts · page: https://llms-explorer.com/tree/llama-cpp-ngram-mod-and-ngram-map-k4v-drafter-in/ ## Facts - ngram-mod stores one `int32` per slot and has no key check. A slot is chosen by `res = res*6364136223846793005 + token` over the n tokens, then `% size`. A lookup returns whatever token was last written to that slot, so two different n-grams that hash to one slot overwrite each other and can return a wrong token. The draft is then only a guess that the target verifies. — source: `asserted` - In the source, the draft tokens are copied from `inp[match_pos + n + i]` (the most recent matching position), not from the winning slot's stored `value_idx`. When the latest occurrence differs from the dominant value, the draft is the latest continuation, not the dominant one. — source: `asserted` - ngram-map structures come from PR 18471 and `ngram-mod` from PR 19164 (header comments). — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.h) - `common_ngram_mod` hashes n tokens with the multiplier 6364136223846793005, takes the value modulo the table size, and stores a single `int32` next token per slot with EMPTY = -1. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-mod.cpp) - `common_ngram_mod::add` increments `used` only when the slot was EMPTY and overwrites the slot unconditionally, and `get` returns the slot content without checking the key. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-mod.cpp) - The ngram-mod drafter allocates `4*1024*1024` entries and asserts `sizeof(llama_token) == sizeof(entry_t)`. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/speculative.cpp) - ngram-mod indexes new prompt tokens only when `i_last + 32 < cur_len`. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/speculative.cpp) - ngram-mod returns no draft when the first EMPTY slot appears before `n_min` tokens were found, and otherwise returns the tokens found so far. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/speculative.cpp) - `begin()` for ngram-mod resets the shared table when occupancy exceeds 0.25 after re-indexing the new prompt. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/speculative.cpp) - A code comment in `seq_info` says "< 0.5" for the low-acceptance fraction while the code compares against 0.25. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/speculative.cpp) - The ngram-mod `accept()` returns early when `is_other` is true. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/speculative.cpp) - `COMMON_NGRAM_MAX_VALUES` is 4 and `COMMON_NGRAM_HASH_MAP_SIZE` is 262144, and `key_map` holds `uint32` history indices. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.h) - Each ngram-map value slot stores `value_idx`, `value_num` and `n_accepted` (initial -1, set to m when a slot is created). — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.h) - The ngram-map key hash is a 32-bit LCG with factor 2654435761, and a `key_map` hit is verified token by token before use. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.cpp) - A k4v draft is made only when `key_num >= min_hits` and either no other value was seen or the top count is at least twice the sum of the others. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.cpp) - The k4v draft length is `min(m, n_accepted)` of the chosen value slot, and `common_ngram_map_accept` stores the new accepted count. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.cpp) - The k4v draft tokens are copied from `inp[match_pos + n + i]`, not from the dominant slot's `value_idx`. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.cpp) - Value counts and key counts saturate at 16380 (`COMMON_NGRAM_MAX_VALUE_COUNT`). — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.cpp) - `common_ngram_map_begin` deletes stale keys, values and hash entries when the new prompt is shorter than the previous one, for reasoning chats whose earlier thinking is dropped. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/common/ngram-map.cpp) - The docs state `--spec-ngram-map-k-min-hits` defaults to 1, and the k4v example uses size-n 8, size-m 8, min-hits 2, draft-n-max 64. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/docs/speculative.md) - When a draft model is combined with a draftless type, the draftless type has higher precedence. — [source](https://raw.githubusercontent.com/ggml-org/llama.cpp/master/docs/speculative.md)