ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-12 23:10:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

When tokenizer info is missing in the model, use llama3 by default

Browse Source
This commit is contained in:
Kawrakow
2024-07-19 12:29:01 +03:00
parent 6a94ca46ad
commit 8bf126c1d6
1 changed files with 10 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
llama.cpp
View File

@@ -4882,14 +4882,21 @@ static void llm_load_vocab(
// for now, only BPE models have pre-tokenizers
if (vocab.type == LLAMA_VOCAB_TYPE_BPE) {
if (tokenizer_pre.empty()) {
LLAMA_LOG_WARN("%s: missing pre-tokenizer type, using: 'default'\n", __func__);
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// OK - I don't feel like recreati8ng the LLaMA-v3 models. Considering that, at least for now,
// LLaMA-v3 is the only model wehere we end up here, let's just force the pre-tokanizer to be
// llama3.
tokenizer_pre = "llama3";
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_LLAMA3;
LLAMA_LOG_WARN("%s: missing pre-tokenizer type, using: 'llama3'\n", __func__);
LLAMA_LOG_WARN("%s: \n", __func__);
LLAMA_LOG_WARN("%s: ************************************ \n", __func__);
LLAMA_LOG_WARN("%s: GENERATION QUALITY WILL BE DEGRADED! \n", __func__);
LLAMA_LOG_WARN("%s: GENERATION QUALITY MAY BE DEGRADED! \n", __func__);
LLAMA_LOG_WARN("%s: CONSIDER REGENERATING THE MODEL \n", __func__);
LLAMA_LOG_WARN("%s: ************************************ \n", __func__);
LLAMA_LOG_WARN("%s: \n", __func__);
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
//vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
} else if (tokenizer_pre == "default") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
} else if (