Add link to the paper, Github repository and project page

README.md

@@ -16,6 +16,8 @@ tags:
 
 # MiniCPM-2B-128k
 
+This repository contains the model based on CPM-2B, trained as described in the paper [MiniCPM4: Ultra-Efficient LLMs on End Devices](https://huggingface.co/papers/2506.07900)
+
 <!-- Provide a quick summary of what the model is/does. -->
 
 [OpenBMB Technical Blog Series](https://openbmb.vercel.app/)
@@ -26,7 +28,10 @@ To our best knowledge, MiniCPM-2B-128k is the first long context(>=128k) SLM sma
 In comparison with the previous released [MiniCPM-2B](https://huggingface.co/openbmb/MiniCPM-2B-sft-bf16), the improvements include:
 
 - Supports 128k context, achieving the best score under 7B on the comprehensive long-text evaluation InfiniteBench, but performance drops within 4k context
-- To facilitate community developers, the model has updated the <user>{}<AI> directive template to chatml format (user\n{}\nassistant\n) during alignment, which also aids users in deploying and using the vllm openai compatible server mode.
+- To facilitate community developers, the model has updated the <user>{}<AI> directive template to chatml format (user
+{}
+assistant
+) during alignment, which also aids users in deploying and using the vllm openai compatible server mode.
 - Due to the parallel mechanism requirement, removed tie_embedding and expanded the vocabulary to 127660.
 
 For more details, please refer to the [GitHub repo](https://github.com/OpenBMB/MiniCPM) and [Blog](https://openbmb.vercel.app/minicpm-2b-128k-en).
@@ -36,7 +41,10 @@ MiniCPM 是面壁与清华大学自然语言处理实验室共同开源的系列
 MiniCPM-2B-128k 是一次基于 [MiniCPM-2B](https://huggingface.co/openbmb/MiniCPM-2B-sft-bf16) 的长度扩展尝试，也是第一个 3B 以下的长文本模型。相对于之前发布的版本，改进如下：
 
 - 支持 128k 上下文，在综合长文本评测 [InfiniteBench](https://github.com/OpenBMB/InfiniteBench) 上取得 7B 以下最佳成绩，但在 4k 以内性能有下降
-- 为方便社区开发者使用，该模型在对齐时将 <用户>{}<AI> 指令模板更新为了 chatml 格式（<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n），这也有助于用户使用 vllm openai compatible server 模式部署和使用。
+- 为方便社区开发者使用，该模型在对齐时将 <用户>{}<AI> 指令模板更新为了 chatml 格式（<|im_start|>user
+{}<|im_end|>
+<|im_start|>assistant
+），这也有助于用户使用 vllm openai compatible server 模式部署和使用。
 - 由于并行机制需要,去除了 tie_embedding，并扩展词表到 127660。
 
 更多细节请参考 [GitHub repo](https://github.com/OpenBMB/MiniCPM) 和 [Blog](https://openbmb.vercel.app/minicpm-2b-128k)
@@ -90,4 +98,4 @@ model = AutoModelForCausalLM.from_pretrained(path, torch_dtype=torch.bfloat16, d
 
 responds, history = model.chat(tokenizer, "山东省最高的山是哪座山, 它比黄山高还是矮？差距多少？", temperature=0.8, top_p=0.8)
 print(responds)
-```
+```