ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

[NPU] Update C++ example to add DeepSeek-R1 (#12787)

Browse source
This commit is contained in:
Ruonan Wang 2025-02-07 11:23:34 +08:00 committed by GitHub
parent d0d9c9d636
commit b4f2be2b09
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 22 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
python/llm/example/NPU/HF-Transformers-AutoModels/LLM/CPP_Examples/README.md
View file

@ -11,6 +11,7 @@ In this directory, you will find a C++ example on how to run LLM models on Intel
| Qwen2 | [Qwen/Qwen2-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2-1.5B-Instruct), [Qwen/Qwen2-7B-Instruct](https://huggingface.co/Qwen/Qwen2-7B-Instruct) | | Qwen2 | [Qwen/Qwen2-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2-1.5B-Instruct), [Qwen/Qwen2-7B-Instruct](https://huggingface.co/Qwen/Qwen2-7B-Instruct) |
| Qwen2.5 | [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct), [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) | | Qwen2.5 | [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct), [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) |
| MiniCPM | [openbmb/MiniCPM-1B-sft-bf16](https://huggingface.co/openbmb/MiniCPM-1B-sft-bf16), [openbmb/MiniCPM-2B-sft-bf16](https://huggingface.co/openbmb/MiniCPM-2B-sft-bf16) | | MiniCPM | [openbmb/MiniCPM-1B-sft-bf16](https://huggingface.co/openbmb/MiniCPM-1B-sft-bf16), [openbmb/MiniCPM-2B-sft-bf16](https://huggingface.co/openbmb/MiniCPM-2B-sft-bf16) |
| DeepSeek-R1 | [DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B), [DeepSeek-R1-Distill-Qwen-7B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B) |
Please refer to [Quickstart](../../../../../../../docs/mddocs/Quickstart/npu_quickstart.md#c-api) for details about verified platforms. Please refer to [Quickstart](../../../../../../../docs/mddocs/Quickstart/npu_quickstart.md#c-api) for details about verified platforms.
@ -72,6 +73,12 @@ python convert.py --repo-id-or-model-path openbmb/MiniCPM-1B-sft-bf16 --save-di
:: to convert MiniCPM-2B-sft-bf16 :: to convert MiniCPM-2B-sft-bf16
python convert.py --repo-id-or-model-path openbmb/MiniCPM-2B-sft-bf16 --save-directory <converted_model_path> python convert.py --repo-id-or-model-path openbmb/MiniCPM-2B-sft-bf16 --save-directory <converted_model_path>
:: to convert DeepSeek-R1-Distill-Qwen-1.5B
python convert.py --repo-id-or-model-path deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B --save-directory <converted_model_path>
:: to convert DeepSeek-R1-Distill-Qwen-7B
python convert.py --repo-id-or-model-path deepseek-ai/DeepSeek-R1-Distill-Qwen-7B --save-directory <converted_model_path>
``` ```
Arguments info: Arguments info:

15
python/llm/example/NPU/HF-Transformers-AutoModels/LLM/CPP_Examples/llama-cli-npu.cpp
View file

@ -34,6 +34,7 @@ const std::string llama2_template = "<s>[INST] <<SYS>>\n\n<</SYS>>\n\n%s [/INST]
const std::string llama3_template = "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n%s<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"; const std::string llama3_template = "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\n\n%s<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n";
const std::string minicpm_template = "<用户>%s<AI>"; const std::string minicpm_template = "<用户>%s<AI>";
const std::string qwen2_template = "<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n%s<|im_end|>\n<|im_start|>assistant\n"; const std::string qwen2_template = "<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n%s<|im_end|>\n<|im_start|>assistant\n";
const std::string qwen2_deepseek_template = "<｜begin▁of▁sentence｜><｜begin▁of▁sentence｜>You are a helpful assistant.<｜User｜>%s<｜Assistant｜>";
std::string add_chat_history(npu_model_params model_params, std::string add_chat_history(npu_model_params model_params,
@ -66,6 +67,20 @@ std::string add_chat_history(npu_model_params model_params,
sprintf_s(prompt, res_template.c_str(), chat_history.c_str(), new_prompt.c_str()); sprintf_s(prompt, res_template.c_str(), chat_history.c_str(), new_prompt.c_str());
} }
} }
} else if (model_params.model_type == std::string("qwen2") && model_params.max_position_embeddings == 131072) {
// For DeepSeek-R1
if (chat_history == ""){
sprintf_s(prompt, qwen2_deepseek_template.c_str(), new_prompt.c_str());
}else{
if (is_input){
std::string input_template = "%s%s<｜Assistant｜>";
sprintf_s(prompt, input_template.c_str(), chat_history.c_str(), new_prompt.c_str());
}
else{
std::string res_template = "%s%s<｜User｜>";
sprintf_s(prompt, res_template.c_str(), chat_history.c_str(), new_prompt.c_str());
}
}
} else if (model_params.model_type == std::string("qwen2")) { } else if (model_params.model_type == std::string("qwen2")) {
if (chat_history == ""){ if (chat_history == ""){
sprintf_s(prompt, qwen2_template.c_str(), new_prompt.c_str()); sprintf_s(prompt, qwen2_template.c_str(), new_prompt.c_str());