Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-27 06:39:25 +01:00
Code Issues Packages Projects Releases Wiki Activity

llama : print a log of the total cache size

Browse Source
This commit is contained in:
Georgi Gerganov 2024-05-29 21:44:55 +03:00
parent 1494a1841e
commit 8a8f8b953f
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 16 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
llama.cpp
View File

@ -4842,21 +4842,28 @@ static void llm_load_vocab(
} }
); );
LLAMA_LOG_INFO("%s: special tokens cache size = %u.\n", __func__, (uint32_t)vocab.cache_special_tokens.size()); LLAMA_LOG_INFO("%s: special tokens cache size = %u\n", __func__, (uint32_t)vocab.cache_special_tokens.size());
} }
// build token to piece caches // build token to piece caches
{ {
size_t size_cache = 0;
std::vector<llama_vocab::token> cache_token_to_piece (n_vocab); std::vector<llama_vocab::token> cache_token_to_piece (n_vocab);
std::vector<llama_vocab::token> cache_token_to_piece_special(n_vocab); std::vector<llama_vocab::token> cache_token_to_piece_special(n_vocab);
for (uint32_t id = 0; id < n_vocab; ++id) { for (uint32_t id = 0; id < n_vocab; ++id) {
cache_token_to_piece[id] = llama_token_to_piece(&model, id, false); cache_token_to_piece[id] = llama_token_to_piece(&model, id, false);
cache_token_to_piece_special[id] = llama_token_to_piece(&model, id, true); cache_token_to_piece_special[id] = llama_token_to_piece(&model, id, true);
size_cache += cache_token_to_piece[id].size();
size_cache += cache_token_to_piece_special[id].size();
} }
std::swap(vocab.cache_token_to_piece, cache_token_to_piece); std::swap(vocab.cache_token_to_piece, cache_token_to_piece);
std::swap(vocab.cache_token_to_piece_special, cache_token_to_piece_special); std::swap(vocab.cache_token_to_piece_special, cache_token_to_piece_special);
LLAMA_LOG_INFO("%s: token to piece cache size = %.4f MB\n", __func__, size_cache / 1024.0 / 1024.0);
} }
} }