CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

vocab : mark EOT token for Granite models (#16499)

Browse Source 
* vocab : mark EOT token for Granite models

* sampling : fallback to EOS when EOT is not found
This commit is contained in:
Georgi Gerganov
2025-10-10 17:17:31 +03:00
committed by GitHub
parent 68ee98ae18
commit 81086cd6a3
2 changed files with 6 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/llama-sampling.cpp
View File

@@ -2541,8 +2541,13 @@ static void llama_sampler_infill_apply(struct llama_sampler * smpl, llama_token_
if (n_non_eog == 0) { if (n_non_eog == 0) {
cur_p->size = 1; cur_p->size = 1;
cur_p->data[0].id = ctx->vocab->token_eot(); cur_p->data[0].id = ctx->vocab->token_eot();
if (cur_p->data[0].id == LLAMA_TOKEN_NULL) {
cur_p->data[0].id = ctx->vocab->token_eos();
}
cur_p->data[0].logit = 1.0f; cur_p->data[0].logit = 1.0f;
GGML_ASSERT(cur_p->data[0].id != LLAMA_TOKEN_NULL);
return; return;
} }

1
src/llama-vocab.cpp
View File

@@ -2171,6 +2171,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
|| t.first == "<|end|>" || t.first == "<|end|>"
|| t.first == "<end_of_turn>" || t.first == "<end_of_turn>"
|| t.first == "<|endoftext|>" || t.first == "<|endoftext|>"
|| t.first == "<|end_of_text|>" // granite
|| t.first == "<EOT>" || t.first == "<EOT>"
|| t.first == "_<EOT>" || t.first == "_<EOT>"
|| t.first == "<｜end▁of▁sentence｜>" // DeepSeek || t.first == "<｜end▁of▁sentence｜>" // DeepSeek