CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : add support for Deepseek-R1-Qwen distill model (#11310)

Browse Source 
* llama : add support for Deepseek-R1-Qwen distill model

* coding style
This commit is contained in:
Xuan Son Nguyen
2025-01-20 14:35:07 +01:00
committed by GitHub
parent ef6dada60c
commit ec7f3ac9ab
6 changed files with 208 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/llama-chat.cpp
View File

@@ -152,7 +152,7 @@ llm_chat_template llm_chat_detect_template(const std::string & tmpl) {
return LLM_CHAT_TEMPLATE_MINICPM;
} else if (tmpl_contains("'Assistant: ' + message['content'] + eos_token")) {
return LLM_CHAT_TEMPLATE_DEEPSEEK_2;
} else if (tmpl_contains(LU8("'<｜Assistant｜>' + message['content'] + '<｜end▁of▁sentence｜>'"))) {
} else if (tmpl_contains(LU8("<｜Assistant｜>")) && tmpl_contains(LU8("<｜User｜>")) && tmpl_contains(LU8("<｜end▁of▁sentence｜>"))) {
return LLM_CHAT_TEMPLATE_DEEPSEEK_3;
} else if (tmpl_contains("[|system|]") && tmpl_contains("[|assistant|]") && tmpl_contains("[|endofturn|]")) {
// ref: https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct/discussions/8#66bae61b1893d14ee8ed85bb

3
src/llama-vocab.cpp
View File

@@ -1523,7 +1523,8 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
pre_type = LLAMA_VOCAB_PRE_TYPE_COMMAND_R;
clean_spaces = false;
} else if (
tokenizer_pre == "qwen2") {
tokenizer_pre == "qwen2" ||
tokenizer_pre == "deepseek-r1-qwen") {
pre_type = LLAMA_VOCAB_PRE_TYPE_QWEN2;
clean_spaces = false;
} else if (