vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

vocab : fix double-eos check (#11273)

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2025-01-17 09:28:00 +02:00 • committed by GitHub
parent 960ec65273
commit a133566d34
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/llama-vocab.cpp
View file

@ -439,7 +439,7 @@ struct llm_tokenizer_bpe_session {
"also starts with a BOS token. So now the final prompt starts with 2 BOS tokens. " "also starts with a BOS token. So now the final prompt starts with 2 BOS tokens. "
"Are you sure this is what you want?\n", __FUNCTION__); "Are you sure this is what you want?\n", __FUNCTION__);
} }
if (vocab.get_add_bos() && output.size() >= 2 && *(output.end()-2) == vocab.token_eos()) { if (vocab.get_add_eos() && output.size() >= 2 && *(output.end()-2) == vocab.token_eos()) {
LLAMA_LOG_WARN( LLAMA_LOG_WARN(
"%s: Added a EOS token to the prompt as specified by the model but the prompt " "%s: Added a EOS token to the prompt as specified by the model but the prompt "
"also ends with a EOS token. So now the final prompt ends with 2 EOS tokens. " "also ends with a EOS token. So now the final prompt ends with 2 EOS tokens. "