CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-04 09:32:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : add infill sampler (#9896)

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2024-10-15 16:35:33 +03:00
committed by GitHub
parent 223c25a72f
commit 755a9b2bf0
9 changed files with 300 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
src/llama-vocab.cpp
View File

@@ -1858,6 +1858,23 @@ int32_t llama_token_to_piece_impl(const struct llama_vocab & vocab, llama_token
return 0;
}
bool llama_token_is_prefix_impl(
const struct llama_vocab & vocab,
llama_token token0,
llama_token token1) {
char text_buf_0[128];
char text_buf_1[128];
const int32_t len0 = llama_token_to_piece_impl(vocab, token0, text_buf_0, sizeof(text_buf_0) - 1, 0, false);
const int32_t len1 = llama_token_to_piece_impl(vocab, token1, text_buf_1, sizeof(text_buf_1) - 1, 0, false);
if (len0 <= 0 || len1 <= 0) {
return false;
}
return len0 <= len1 && memcmp(text_buf_0, text_buf_1, len0) == 0;
}
int32_t llama_detokenize_impl(
const struct llama_vocab & vocab,
const llama_token * tokens,