Update README.md

README.md

@@ -23,7 +23,7 @@ Hello World! This is CodeFuse!
 
 In this release, we are open sourcing 
 1. [**The MFT (Multi-Task Fine-Tuning) framework, known as MFTCoder**](https://github.com/codefuse-ai/MFTCoder);
-2. **Two datasets for enhancing the coding capabilities of LLMs**, that is, [Code Exercise](https://huggingface.co/datasets/codefuse/CodeExercise-Python-27k) and [Evol-Instruction](https://huggingface.co/datasets/codefuse/Evol-instruction-66k);
+2. **Two datasets for enhancing the coding capabilities of LLMs, that is, [Code Exercise](https://huggingface.co/datasets/codefuse/CodeExercise-Python-27k) and [Evol-Instruction](https://huggingface.co/datasets/codefuse/Evol-instruction-66k);**
 3. [**A faster and more reliable deployment framework based on FasterTransformer**](https://github.com/codefuse-ai/FasterTransformer4CodeFuse);
 
 The resulting model ensemble, which includes [CodeFuse-13B](https://huggingface.co/codefuse/CodeFuse-13B) and [CodeFuse-CodeLlama-34B](https://huggingface.co/codefuse/CodeFuse-CodeLlama-34B), supports various code-related tasks such as code completion, text-to-code conversion, and unit test generation. In particular, [CodeFuse-CodeLlama-34B](https://huggingface.co/codefuse/CodeFuse-CodeLlama-34B), built upon CodeLlama as the base model and fine-tuned using the proposed MFT framework, achieves an impressive score of **74.4% (greedy decoding)** in the HumanEval Python pass@1 evaluation, **even surpassing the performance of GPT-4 (67%)**. We have plans to incorporate additional base LLMs into the ensemble in the near future.
@@ -38,7 +38,7 @@ CodeFuse的使命是开发专门设计用于支持整个软件开发生命周期
 
 在本次发布中，我们开源了以下内容：
 1. [**MFT（多任务微调）框架，也称为MFTCoder**](https://github.com/codefuse-ai/MFTCoder)；
-2. **两个用于增强LLMs编码能力的数据集**，包括[Code Exercise](https://huggingface.co/datasets/codefuse/CodeExercise-Python-27k)和[Evol-Instruction](https://huggingface.co/datasets/codefuse/Evol-instruction-66k)；
+2. **两个用于增强LLMs编码能力的数据集，包括[Code Exercise](https://huggingface.co/datasets/codefuse/CodeExercise-Python-27k)和[Evol-Instruction](https://huggingface.co/datasets/codefuse/Evol-instruction-66k)；**
 3. [**基于FasterTransformer的更快速、更可靠的部署框架**](https://github.com/codefuse-ai/FasterTransformer4CodeFuse);。
 
 由此产生的模型集合包括[CodeFuse-13B](https://huggingface.co/codefuse/CodeFuse-13B)和[CodeFuse-CodeLlama-34B](https://huggingface.co/codefuse/CodeFuse-CodeLlama-34B)，支持多种与代码相关的任务，如代码补全、文本转代码、单元测试生成等。值得一提的是，[CodeFuse-CodeLlama-34B](https://huggingface.co/codefuse/CodeFuse-CodeLlama-34B)基于CodeLlama作为基础模型，并利用我们提出的MFT框架进行微调，在HumanEval Python pass@1评估中取得高达的**74.4%（贪婪解码）**的好成绩，甚至**超过了GPT-4（67%）的表现**。我们计划在不久的将来将更多的基础LLMs纳入到我们的模型集合中。