Upload 7 files

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+assets/cli_demo.gif filter=lfs diff=lfs merge=lfs -text
+assets/web_demo.gif filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,7 +1,15 @@
-<p align="left">
-    中文</a>&nbsp ｜ &nbsp<a href="README_EN.md">English</a>&nbsp
-</p>
-<br>
+---
+license: apache-2.0
+language:
+- zh
+- en
+library_name: transformers
+tags:
+- qihoo360
+- 奇虎360
+- 360Zhinao
+- pretrain
+---
 
 <div align="center">
 <h1>
@@ -9,16 +17,18 @@
 </h1>
 </div>
 <div align="center">
-    🤗 <a href="">Hugging Face</a>&nbsp&nbsp | &nbsp&nbsp
-    🤖 <a href="">ModelScope</a>&nbsp&nbsp ｜ &nbsp&nbsp
-    💬 <a href="">WeChat (微信)</a>&nbsp&nbsp
+    🤖 <a href="https://www.modelscope.cn/profile/qihoo360">ModelScope</a>&nbsp&nbsp ｜ &nbsp&nbsp
+    🔥 <a href="https://github.com/Qihoo360/360zhinao/blob/main/assets/WeChat.png">GitHub</a>&nbsp&nbsp | &nbsp&nbsp
+    💬 <a href="https://github.com/Qihoo360/360zhinao/blob/main/assets/WeChat.png">WeChat (微信)</a>&nbsp&nbsp
 </div>
 <br>
 <p align="center">
  欢迎访问360智脑官网<a href="https://ai.360.com"> https://ai.360.com </a>体验更多更强大的功能。
 </p>
 
-# 项目介绍
+<br>
+
+# 模型介绍
  🎉🎉🎉我们开源了360智脑大模型的系列工作，本次开源了以下模型：
  - **360Zhinao-7B-Base**
  - **360Zhinao-7B-Chat-4K**
@@ -27,10 +37,14 @@
 
 360智脑大模型特点如下：
 - **基础模型**：采用 3.4 万亿 Tokens 的高质量语料库训练，以中文、英文、代码为主，在相关基准评测中，同尺寸有竞争力。
-- **对话模型**：具有强大的对话能力，开放4k、32k、360k三种不同窗口长度。据了解，360k（约50万字）在国内目前开源的长文本能力中最长。
+- **对话模型**：具有强大的对话能力，开放4K、32K、360K三种不同文本长度。据了解，360K（约50万字）是当前国产开源模型文本长度最长的。
+
+<br>
 
 # 更新信息
-- [2024.04.10] 我们发布了360Zhinao-7B 1.0版本，同时开放Base模型和4k、32k、360k三种文本长度的Chat模型。
+- [2024.04.10] 我们发布了360Zhinao-7B 1.0版本，同时开放Base模型和4K、32K、360K三种文本长度的Chat模型。
+
+<br>
 
 # 目录
 - [下载地址](#下载地址)
@@ -40,39 +54,109 @@
 - [模型微调](#模型微调)
 - [许可证](#许可证)
 
+<br>
+
 # 下载地址
 本次发布版本和下载链接见下表：
-| | Zhinao-Base | Zhinao-Chat | Zhinao-Chat(Int8) | Zhinao-Chat(Int4) |
-|-|-|-|-|-|
-| 1.8B  |  <a href="">🤖</a>  <a href="">🤗</a> |  <a href="">🤖</a>  <a href="">🤗</a>  |  <a href="">🤖</a>  <a href="">🤗</a> |  <a href="">🤖</a>  <a href="">🤗</a>  |
-| 7B  |  <a href="">🤖</a>  <a href="">🤗</a>  | <a href="">🤖</a>  <a href="">🤗</a>  |  <a href="">🤖</a>  <a href="">🤗</a> |  <a href="">🤖</a>  <a href="">🤗</a>  |
+| Size | Model | BF16 | Int4|
+|:-:|-|:-:|:-:|
+| 7B | 360Zhinao-7B-Base | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Base/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Base">🤗</a> |  |
+| 7B | 360Zhinao-7B-Chat-4K | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-4K/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-4K">🤗</a> | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-4K-Int4/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-4K-Int4">🤗</a> |
+| 7B | 360Zhinao-7B-Chat-32K | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-32K/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-32K">🤗</a> | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-32K-Int4/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-32K-Int4">🤗</a> |
+| 7B | 360Zhinao-7B-Chat-360K | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-360K/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-360K">🤗</a> | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-360K-Int4/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-360K-Int4">🤗</a> |
+
+<br>
 
 # 模型评估
-我们在OpenCompass的主流评测数据集上验证了我们的模型性能，包括C-Eval、AGIEval、MMLU、CMMLU、HellaSwag、MATH、GSM8K、HumanEval、MBPP、BBH、LAMBADA，考察的能力包括自然语言理解、知识、数学计算和推理、代码生成、逻辑推理等。
 
 ## 基础模型
-| Model | C-Eval | AGIEval | MMLU | CMMLU | HellaSwag | MATH | GSM8K | HumanEval | MBPP | BBH | LAMBADA |
-| - | - | - | - | - | - | - | - | - | - | - | - |
-| Phi-1.5-1.3B | 27.8 | 23.4 | 44.3 | 26 | 57.1 | 2.6 | 32.5 | 25 | 33 | 29.6 | 54.6 |
-| Qwen-1.8B | 53.3 | 36.5 | 46.4 | 51.9 | 58.7 | 2.4 | 10.2 | 7.3 | 14 | 22.6 | 54.3 |
-| Qwen-1.5-1.8B | 59.48 | 38.76 | 47.14 | 57.08 | 56.02 | 9.66 | 34.87 | 23.17 | 17.6 | 27.02 | 56.49 |
-| Baichuan2-7B-Base | 56.3 | 34.6 | 54.7 | 57 | 67 | 5.4 | 24.6 | 17.7 | 24 | 41.8 | 73.3 |
-| ChatGLM3-6B-Base | 67 | 47.4 | 62.8 | 66.5 | 76.5 | 19.2 | 61 | 44.5 | 57.2 | 66.2 | 77.1 |
-| DeepSeek-7B-Base | 45 | 24 | 49.3 | 46.8 | 73.4 | 4.2 | 18.3 | 25 | 36.4 | 42.8 | 72.6 |
-| InternLM2-7B | 65.7 | 50.2 | 65.5 | 66.2 | 79.6 | 19.9 | 70.6 | 41.5 | 42.4 | 64.4 | 72.1 |
-| InternLM-7B | 53.4 | 36.9 | 51 | 51.8 | 70.6 | 6.3 | 31.2 | 13.4 | 14 | 37 | 67 |
-| LLaMA-2-7B | 32.5 | 21.8 | 46.8 | 31.8 | 74 | 3.3 | 16.7 | 12.8 | 14.8 | 38.2 | 73.3 |
-| LLaMA-7B | 27.3 | 20.6 | 35.6 | 26.8 | 74.3 | 2.9 | 10 | 12.8 | 16.8 | 33.5 | 73.3 |
-| Mistral-7B-v0.1 | 47.4 | 32.8 | 64.1 | 44.7 | 78.9 | 11.3 | 47.5 | 27.4 | 38.6 | 56.7 | 75 |
-| MPT-7B | 23.5 | 21.3 | 27.5 | 25.9 | 75 | 2.9 | 9.1 | 17.1 | 22.8 | 35.6 | 70 |
-| Qwen-7B | 63.4 | 45.3 | 59.7 | 62.5 | 75 | 13.3 | 54.1 | 27.4 | 31.4 | 45.2 | 67.5 |
-| XVERSE-7B | 61.1 | 39 | 58.4 | 60.8 | 73.7 | 2.2 | 11.7 | 4.9 | 10.2 | 31 | 24 |
-| Yi-6B | 73 | 44.3 | 64 | 73.5 | 73.1 | 6.3 | 39.9 | 15.2 | 23.6 | 44.9 | 68 |
-| Zhinao-1.8B-Base | 49.78 | 31.87 | 50.05 | 52.58 | 57.31 | 4.82 | 15.01 | 14.02 | 19.4 | 29.76 | 69.77 |
-| 360Zhinao-7B-Base | 74.11 | 49.49 | 67.44 | 72.38 | 83.05 | 16.38 | 53.83 | 35.98 | 42.4 | 43.95 | 78.59 |
+我们在OpenCompass的主流评测数据集上验证了我们的模型性能，包括C-Eval、AGIEval、MMLU、CMMLU、HellaSwag、MATH、GSM8K、HumanEval、MBPP、BBH、LAMBADA，考察的能力包括自然语言理解、知识、数学计算和推理、代码生成、逻辑推理等。
+
+
+| <div style="width: 100pt">Model</div> | AVG   | CEval | AGIEval | MMLU | CMMLU | HellaSwag | MATH | GSM8K | HumanEval | MBPP | BBH | LAMBADA |
+|:----------------------|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|
+| Baichuan2-7B          | 41.49     | 56.3      | 34.6      | 54.7      | 57        | 67        | 5.4       | 24.6      | 17.7      | 24        | 41.8      | 73.3      |
+| Baichuan-7B           | 31.94     | 44.7      | 24.6      | 41.5      | 44.6      | 68.4      | 2.5       | 9.6       | 9.1       | 6.4       | 32.8      | 67.1      |
+| ChatGLM3-6B           | **58.67** | 67        | 47.4      | 62.8      | 66.5      | 76.5      | 19.2      | 61        | 44.5      | **57.2**  | **66.2**  | 77.1      |
+| DeepSeek-7B           | 39.8      | 45        | 24        | 49.3      | 46.8      | 73.4      | 4.2       | 18.3      | 25        | 36.4      | 42.8      | 72.6      |
+| InternLM2-7B          | 58.01     | 65.7      | 50.2      | 65.5      | 66.2      | 79.6      | 19.9      | **70.6**  | 41.5      | 42.4      | 64.4      | 72.1      |
+| InternLM-7B           | 39.33     | 53.4      | 36.9      | 51        | 51.8      | 70.6      | 6.3       | 31.2      | 13.4      | 14        | 37        | 67        |
+| LLaMA-2-7B            | 33.27     | 32.5      | 21.8      | 46.8      | 31.8      | 74        | 3.3       | 16.7      | 12.8      | 14.8      | 38.2      | 73.3      |
+| LLaMA-7B              | 30.35     | 27.3      | 20.6      | 35.6      | 26.8      | 74.3      | 2.9       | 10        | 12.8      | 16.8      | 33.5      | 73.3      |
+| Mistral-7B-v0.1       | 47.67     | 47.4      | 32.8      | 64.1      | 44.7      | 78.9      | 11.3      | 47.5      | 27.4      | 38.6      | 56.7      | 75        |
+| MPT-7B                | 30.06     | 23.5      | 21.3      | 27.5      | 25.9      | 75        | 2.9       | 9.1       | 17.1      | 22.8      | 35.6      | 70        |
+| Qwen1.5-7B            | 55.12     | 73.57     | **50.8**  | 62.15     | 71.84     | 72.62     | **20.36** | 54.36     | **53.05** | 36.8      | 40.01     | 70.74     |
+| Qwen-7B               | 49.53     | 63.4      | 45.3      | 59.7      | 62.5      | 75        | 13.3      | 54.1      | 27.4      | 31.4      | 45.2      | 67.5      |
+| XVERSE-7B             | 34.27     | 61.1      | 39        | 58.4      | 60.8      | 73.7      | 2.2       | 11.7      | 4.9       | 10.2      | 31        | 24        |
+| Yi-6B                 | 47.8      | 73        | 44.3      | 64        | **73.5**  | 73.1      | 6.3       | 39.9      | 15.2      | 23.6      | 44.9      | 68        |
+| **360Zhinao-7B**      | 56.15     | **74.11** | 49.49     | **67.44** | 72.38     | **83.05** | 16.38     | 53.83     | 35.98     | 42.4      | 43.95     | **78.59** |
 
 以上结果，在官方[Opencompass](https://rank.opencompass.org.cn/leaderboard-llm)上可查询或可复现。
 
+## Chat模型
+
+  我们采用两阶段的方式训练长文本模型.
+  
+  **第一阶段**：我们增大RoPE base，将上下文长度扩展至32K训练：
+    - 首先，对基础模型进行了约5B tokens的32K窗口继续预训练。
+    - 接着，SFT阶段使用了多种形式和来源的长文本数据，包括高质量的人工标注32K长文本数据。
+
+  **第二阶段**：我们将上下文长度扩展至360K进行训练，使用数据如下：
+    - 少量高质量人工标注数据。
+    - 由于带有标注的超长文本数据的稀缺性，我们构造了多种形式的合成数据：
+      - 多文档问答：类似[Ziya-Reader](https://arxiv.org/abs/2311.09198)，我们基于360自有数据构造了多种类型的多���档问答数据，同时将问答改为多轮，显著提升长文本的训练效率。
+      - 单文档问答：类似[LLama2 Long](https://arxiv.org/abs/2309.16039)，我们构造了基于超长文本各个片段的多轮问答数据。
+
+我们在多种长度和多种任务的评测Benchmark上验证不同版本模型的性能。
+
+- ### 360Zhinao-7B-Chat-32K模型长文本能力评测
+
+
+  我们使用LongBench验证长文本效果。[LongBench](https://github.com/THUDM/LongBench)是第一个多任务、中英双语、针对大语言模型长文本理解能力的评测基准。LongBench由六大类、二十一个不同的任务组成，我们选择其中与中文长文本应用最密切相关的中文单文档问答、多文档问答、摘要、Few-shot等任务进行评测。
+
+    | Model                     | Avg       | 单文档QA  | 多文档QA   | 摘要       | Few-shot学习 | 代码补全    |
+    | :------------------------ |:---------:|:--------:|:---------:|:---------:|:------------:|:---------:|
+    | GPT-3.5-Turbo-16k         | 37.84     | 61.2     | 28.7      | 16        | 29.2         | 54.1      |
+    | ChatGLM2-6B-32k           | 37.16     | 51.6     | 37.6      | 16.2      | 27.7         | 52.7      |
+    | ChatGLM3-6B-32k           | 44.62     | **62.3** | 44.8      | 17.8      | 42           | 56.2      |
+    | InternLM2-Chat-7B         | 42.20     | 56.65    | 29.15     | **17.99** | 43.5         | **63.72** |
+    | Qwen1.5-Chat-7B           | 36.75     | 52.85    | 30.08     | 14.28     | 32           | 54.55     |
+    | Qwen1.5-Chat-14B          | 39.80     | 60.39    | 27.99     | 14.77     | 37           | 58.87     |
+    | 360Zhinao-7B-Chat-32K     | **45.18** | 57.18    | **48.06** | 15.03     | **44**       | 61.64     |
+
+- ### 360Zhinao-7B-Chat-360K“大海捞针”测试
+
+  大海捞针测试（[NeedleInAHaystack](https://github.com/gkamradt/LLMTest_NeedleInAHaystack)）是将关键信息插入一段长文本的不同位置，再对该关键信息提问，从而测试大模型的长文本能力的一种方法。
+
+  360Zhinao-7B-Chat-360K在中英文大海捞针中都能达到98%以上的准确率。
+
+  - 英文"大海捞针"（和[NeedleInAHaystack](https://github.com/gkamradt/LLMTest_NeedleInAHaystack)相同）
+  
+    <p align="center">
+        <img src="assets/360Zhinao-7B-Chat-360K.en_score.png" width="600" />
+    <p>
+
+    **针**：The best thing to do in San Francisco is eat a sandwich and sit in Dolores Park on a sunny day.
+
+    **提问**：What is the best thing to do in San Francisco?
+
+
+  - 中文“大海捞针”
+
+    <p align="center">
+        <img src="assets/360Zhinao-7B-Chat-360K.zh_score.png" width="600" />
+    <p>
+
+    我们仿照[SuperCLUE-200K测评基准](https://mp.weixin.qq.com/s/QgoRf2LB-7vc3vTFOHJkpw)构造了中文大海捞针：
+
+    **海**：长篇小说。
+    
+    **针**：王莽是一名勤奋的店员，他每天凌晨就起床，赶在第一缕阳光照亮大地之前到达店铺，为即将开始的一天做准备。他清扫店铺，整理货架，为顾客提供方便。他对五金的种类和用途了如指掌，无论顾客需要什么，他总能准确地找到。\n然而，他的老板刘秀却总是对他吹毛求疵。刘秀是个挑剔的人，他总能在王莽的工作中找出一些小错误，然后以此为由扣他的工资。他对王莽的工作要求非常严格，甚至有些过分。即使王莽做得再好，刘秀也总能找出一些小问题，让王莽感到非常沮丧。\n王莽虽然对此感到不满，但他并没有放弃。他知道，只有通过自己的努力，才能获得更好的生活。他坚持每天早起，尽管他知道那天可能会再次被刘秀扣工资。他始终保持微笑，尽管他知道刘秀可能会再次对他挑剔。
+
+    **提问**：王莽在谁的手下工作？
+
+<br>
+
 # 快速开始
 简单的示例来说明如何利用🤖 ModelScope和🤗 Transformers快速使用360Zhinao-7B-Base和360Zhinao-7B-Chat
 
@@ -122,6 +206,7 @@ inputs = inputs.to(model.device)
 pred = model.generate(input_ids=inputs["input_ids"], generation_config=generation_config)
 print("outputs:\n", tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))
 ```
+
 ### Chat模型推理
 
 此代码演示使用transformers快速使用360Zhinao-7B-Chat-4K模型进行推理
@@ -270,17 +355,11 @@ curl --location --request POST 'http://localhost:8360/v1/chat/completions' \
 }'
 ```
 
+<br>
+
 # 模型推理
 ## 模型量化
-我们提供了基于AutoGPTQ的量化方案，并开源了Int4量化模型。模型的效果损失很小，但能显著降低显存占用并提升推理速度。
-
-对BF16，Int8和Int4模型在基准评测上做了测试，结果如下所示：
-
-| Quantization | MMLU | CEval (val) | GSM8K | Humaneval |
-|-|-|-|-|-|
-| 360Zhinao-7B-Chat-4K (BF16) |-|-|-|-|
-| 360Zhinao-7B-Chat-4K (Int8) |-|-|-|-|
-| 360Zhinao-7B-Chat-4K (Int4) |-|-|-|-|
+我们提供了基于AutoGPTQ的量化方案，并开源了Int4量化模型。
 
 ## 模型部署
 ### vLLM安装环境
@@ -295,7 +374,8 @@ pip install vllm==0.3.3
 
 >安装完成后，还需要以下操作~
 1. 把vllm/zhinao.py文件复制到env环境对应的vllm/model_executor/models目录下。
-2. 然后在vllm/model_executor/models/\_\_init\_\_.py文件增加一行代码
+2. 把vllm/serving_chat.py文件复制到env环境对应的vllm/entrypoints/openai目录下。
+3. 然后在vllm/model_executor/models/\_\_init\_\_.py文件增加一行代码
 
     ```shell
     "ZhinaoForCausalLM": ("zhinao", "ZhinaoForCausalLM"),
@@ -309,8 +389,8 @@ python -m vllm.entrypoints.openai.api_server \
     --served-model-name 360Zhinao-7B-Chat-4K \
     --model qihoo360/360Zhinao-7B-Chat-4K \
     --trust-remote-code \
-    --tensor-parallel-size 1
-    --max-model-len 18000 \
+    --tensor-parallel-size 1 \
+    --max-model-len 4096 \
     --host 0.0.0.0 \
     --port 8360
 ```
@@ -322,7 +402,7 @@ curl http://localhost:8360/v1/chat/completions \
 -d '{
     "model": "360Zhinao-7B-Chat-4K",
     "max_tokens": 200,
-    "top_k": 0,
+    "top_k": -1,
     "top_p": 0.8,
     "temperature": 1.0,
     "presence_penalty": 0.0,
@@ -343,7 +423,7 @@ curl http://localhost:8360/v1/chat/completions \
 from openai import OpenAI
 # Set OpenAI's API key and API base to use vLLM's API server.
 openai_api_key = "EMPTY"
-openai_api_base = "http://localhost:8000/v1"
+openai_api_base = "http://localhost:8360/v1"
 
 client = OpenAI(
     api_key=openai_api_key,
@@ -369,6 +449,8 @@ print("Chat response:", chat_response)
 
 > 注意：如需要开启重复惩罚，建议使用 *presence_penalty* 和 *frequency_penalty* 参数。
 
+<br>
+
 # 模型微调
 ## 训练数据
 
@@ -396,6 +478,7 @@ print("Chat response:", chat_response)
   }
 ]
 ```
+
 ## 微调训练
 训练脚本如下：
 ```shell
@@ -458,8 +541,10 @@ bash finetune/ds_finetune.sh
 - 可通过配置fp16、bf16实现混合精度训练，建议使用bf16，与预训练模型保持一致。
 - 可通过配置is_concat参数，控制训练数据是否拼接，当训练数据量级较大时，可通过拼接提升训练效率。
 
+<br>
+
 # 许可证
 
 本仓库源码遵循开源许可证Apache 2.0。
 
-360智脑开源模型支持商用，若需将本模型及衍生模型用于商业用途，请通过邮箱（g-zhinao-opensource@360.cn）联系进行申请， 具体许可协议请见[《360智脑开源模型许可证》](./360智脑开源模型许可证.txt)。
+360智脑开源模型支持商用，若需将本模型及衍生模型用于商业用途，请通过邮箱(g-zhinao-opensource@360.cn)联系进行申请， 具体许可协议请见[《360智脑开源模型许可证》](./360智脑开源模型许可证.txt)。

README_EN.md

@@ -1,36 +1,38 @@
-<p align="left">
-    中文</a>&nbsp ｜ &nbsp<a href="README_EN.md">English</a>&nbsp
-</p>
-<br>
-
 <div align="center">
 <h1>
-  360智脑
+  360Zhinao (360智脑)
 </h1>
 </div>
 <div align="center">
-    🤗 <a href="">Hugging Face</a>&nbsp&nbsp | &nbsp&nbsp
-    🤖 <a href="">ModelScope</a>&nbsp&nbsp ｜ &nbsp&nbsp
-    💬 <a href="">WeChat (微信)</a>&nbsp&nbsp
+    🤖 <a href="https://www.modelscope.cn/profile/qihoo360">ModelScope</a>&nbsp&nbsp ｜ &nbsp&nbsp
+    🔥 <a href="https://github.com/Qihoo360/360zhinao/blob/main/assets/WeChat.png">GitHub</a>&nbsp&nbsp | &nbsp&nbsp
+    💬 <a href="https://github.com/Qihoo360/360zhinao/tree/main/assets/WeChat.png">WeChat (微信)</a>&nbsp&nbsp
 </div>
 <br>
 <p align="center">
- Welcome to visit 360Zhinao official website<a href="https://ai.360.com"> https://ai.360.com</a> to experience more powerful functions.
+ Feel free to visit 360Zhinao's official website<a href="https://ai.360.com"> https://ai.360.com</a> for more experience.
 </p>
 
+<br>
+
 # Models Introduction
- 🎉🎉🎉We opensource our 360Zhinao series，The following models are open sourced：
+ 🎉🎉🎉We open-source the 360Zhinao model series:
  - **360Zhinao-7B-Base**
  - **360Zhinao-7B-Chat-4K**
  - **360Zhinao-7B-Chat-32K**
  - **360Zhinao-7B-Chat-360K**
 
-The characteristics of the 360Zhinao open-source project are：
-- **Base Model:** Leveraging a high-quality corpus of 3.4 trillion Tokens, primarily in Chinese, English, and code, we achieved competitive performance in relevant benchmark evaluations of the same scale.
-- **Chat Model:** Powerful chat capabilities and three different sequence lengths of 4k, 32k, and 360k. It is understood that 360k (about 500,000 words) is the longest sequcence length among the current chinese open source Large language model. 
+
+The characteristics of the 360Zhinao open-source models are：
+- **Base Model:** Leveraging a high-quality corpus of 3.4 trillion Tokens which mainly consist of Chinese, English and code, we achieved competitive performance on relevant benchmark evaluations of the same model scale.
+- **Chat Model:** Powerful chat capabilities and three different sequence lengths of 4K, 32K and 360K. 360K (about 500k Chinese characters) is the longest sequcence length among open-sourced Chinese models until now.
+
+<br>
 
 # News and Updates
-- 2024.04.10 We release **360Zhinao-7B** 1.0 version,  include the base model and three chat model with sequence length of 4k, 32k, 360k. 
+- 2024.04.11 We release **360Zhinao-7B** 1.0 version,  include the base model and three chat model with sequence lengths of 4K, 32K adn 360K. 
+
+<br>
 
 # Table of contents
 - [Download URL](#Download-URL)
@@ -40,39 +42,106 @@ The characteristics of the 360Zhinao open-source project are：
 - [Model Finetune](#Model-Finetune)
 - [License](#License)
 
+<br>
 
 # Download URL
 See the following table for this release and download links:
-| | Zhinao-Base | Zhinao-Chat | Zhinao-Chat(Int8) | Zhinao-Chat(Int4) |
-|-|-|-|-|-|
-| 1.8B  |  <a href="">🤖</a>  <a href="">🤗</a> |  <a href="">🤖</a>  <a href="">🤗</a>  |  <a href="">🤖</a>  <a href="">🤗</a> |  <a href="">🤖</a>  <a href="">🤗</a>  |
-| 7B  |  <a href="">🤖</a>  <a href="">🤗</a>  | <a href="">🤖</a>  <a href="">🤗</a>  |  <a href="">🤖</a>  <a href="">🤗</a> |  <a href="">🤖</a>  <a href="">🤗</a>  |
+| Size | Model | BF16 | Int4|
+|-|-|-|-|
+| 7B | 360Zhinao-7B-Base | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Base/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Base">🤗</a> |  |
+| 7B | 360Zhinao-7B-Chat-4K | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-4K/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-4K">🤗</a> | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-4K-Int4/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-4K-Int4">🤗</a> |
+| 7B | 360Zhinao-7B-Chat-32K | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-32K/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-32K">🤗</a> | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-32K-Int4/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-32K-Int4">🤗</a> |
+| 7B | 360Zhinao-7B-Chat-360K | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-360K/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-360K">🤗</a> | <a href="https://www.modelscope.cn/models/qihoo360/360Zhinao-7B-Chat-360K-Int4/summary">🤖</a>  <a href="https://huggingface.co/qihoo360/360Zhinao-7B-Chat-360K-Int4">🤗</a> |
+
+<br>
 
 # Model Evaluation
-We validate the performance of our model on the mainstream OpenCompass evaluation datasets, including C-Eval, AGIEval, MMLU, CMMLU, HellaSwag, MATH, GSM8K, HumanEval, MBPP, BBH, LAMBADA. The competencies examined include natural language understanding, knowledge, mathematical computation and reasoning, code generation, logical reasoning, etc.
-
-## Base Models
-| Model | C-Eval | AGIEval | MMLU | CMMLU | HellaSwag | MATH | GSM8K | HumanEval | MBPP | BBH | LAMBADA |
-| - | - | - | - | - | - | - | - | - | - | - | - |
-| Phi-1.5-1.3B | 27.8 | 23.4 | 44.3 | 26 | 57.1 | 2.6 | 32.5 | 25 | 33 | 29.6 | 54.6 |
-| Qwen-1.8B | 53.3 | 36.5 | 46.4 | 51.9 | 58.7 | 2.4 | 10.2 | 7.3 | 14 | 22.6 | 54.3 |
-| Qwen-1.5-1.8B | 59.48 | 38.76 | 47.14 | 57.08 | 56.02 | 9.66 | 34.87 | 23.17 | 17.6 | 27.02 | 56.49 |
-| Baichuan2-7B-Base | 56.3 | 34.6 | 54.7 | 57 | 67 | 5.4 | 24.6 | 17.7 | 24 | 41.8 | 73.3 |
-| ChatGLM3-6B-Base | 67 | 47.4 | 62.8 | 66.5 | 76.5 | 19.2 | 61 | 44.5 | 57.2 | 66.2 | 77.1 |
-| DeepSeek-7B-Base | 45 | 24 | 49.3 | 46.8 | 73.4 | 4.2 | 18.3 | 25 | 36.4 | 42.8 | 72.6 |
-| InternLM2-7B | 65.7 | 50.2 | 65.5 | 66.2 | 79.6 | 19.9 | 70.6 | 41.5 | 42.4 | 64.4 | 72.1 |
-| InternLM-7B | 53.4 | 36.9 | 51 | 51.8 | 70.6 | 6.3 | 31.2 | 13.4 | 14 | 37 | 67 |
-| LLaMA-2-7B | 32.5 | 21.8 | 46.8 | 31.8 | 74 | 3.3 | 16.7 | 12.8 | 14.8 | 38.2 | 73.3 |
-| LLaMA-7B | 27.3 | 20.6 | 35.6 | 26.8 | 74.3 | 2.9 | 10 | 12.8 | 16.8 | 33.5 | 73.3 |
-| Mistral-7B-v0.1 | 47.4 | 32.8 | 64.1 | 44.7 | 78.9 | 11.3 | 47.5 | 27.4 | 38.6 | 56.7 | 75 |
-| MPT-7B | 23.5 | 21.3 | 27.5 | 25.9 | 75 | 2.9 | 9.1 | 17.1 | 22.8 | 35.6 | 70 |
-| Qwen-7B | 63.4 | 45.3 | 59.7 | 62.5 | 75 | 13.3 | 54.1 | 27.4 | 31.4 | 45.2 | 67.5 |
-| XVERSE-7B | 61.1 | 39 | 58.4 | 60.8 | 73.7 | 2.2 | 11.7 | 4.9 | 10.2 | 31 | 24 |
-| Yi-6B | 73 | 44.3 | 64 | 73.5 | 73.1 | 6.3 | 39.9 | 15.2 | 23.6 | 44.9 | 68 |
-| Zhinao-1.8B-Base | 49.78 | 31.87 | 50.05 | 52.58 | 57.31 | 4.82 | 15.01 | 14.02 | 19.4 | 29.76 | 69.77 |
-| 360Zhinao-7B-Base | 74.11 | 49.49 | 67.44 | 72.38 | 83.05 | 16.38 | 53.83 | 35.98 | 42.4 | 43.95 | 78.59 |
-
-The above results, the official [Opencompass](https://rank.opencompass.org.cn/leaderboard-llm) can query or can emersion.
+## Base Model
+We evaluate the performance of our model on the OpenCompass evaluation datasets, including C-Eval, AGIEval, MMLU, CMMLU, HellaSwag, MATH, GSM8K, HumanEval, MBPP, BBH, LAMBADA. The ablity evaluated of model include natural language understanding, knowledge, mathematical computation and reasoning, code generation, logical reasoning, etc.
+
+| <div style="width: 100pt">Model</div> | AVG   | CEval | AGIEval | MMLU | CMMLU | HellaSwag | MATH | GSM8K | HumanEval | MBPP | BBH | LAMBADA |
+|:----------------------|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:|
+| Baichuan2-7B          | 41.49     | 56.3      | 34.6      | 54.7      | 57        | 67        | 5.4       | 24.6      | 17.7      | 24        | 41.8      | 73.3      |
+| Baichuan-7B           | 31.94     | 44.7      | 24.6      | 41.5      | 44.6      | 68.4      | 2.5       | 9.6       | 9.1       | 6.4       | 32.8      | 67.1      |
+| ChatGLM3-6B           | **58.67** | 67        | 47.4      | 62.8      | 66.5      | 76.5      | 19.2      | 61        | 44.5      | **57.2**  | **66.2**  | 77.1      |
+| DeepSeek-7B           | 39.8      | 45        | 24        | 49.3      | 46.8      | 73.4      | 4.2       | 18.3      | 25        | 36.4      | 42.8      | 72.6      |
+| InternLM2-7B          | 58.01     | 65.7      | 50.2      | 65.5      | 66.2      | 79.6      | 19.9      | **70.6**  | 41.5      | 42.4      | 64.4      | 72.1      |
+| InternLM-7B           | 39.33     | 53.4      | 36.9      | 51        | 51.8      | 70.6      | 6.3       | 31.2      | 13.4      | 14        | 37        | 67        |
+| LLaMA-2-7B            | 33.27     | 32.5      | 21.8      | 46.8      | 31.8      | 74        | 3.3       | 16.7      | 12.8      | 14.8      | 38.2      | 73.3      |
+| LLaMA-7B              | 30.35     | 27.3      | 20.6      | 35.6      | 26.8      | 74.3      | 2.9       | 10        | 12.8      | 16.8      | 33.5      | 73.3      |
+| Mistral-7B-v0.1       | 47.67     | 47.4      | 32.8      | 64.1      | 44.7      | 78.9      | 11.3      | 47.5      | 27.4      | 38.6      | 56.7      | 75        |
+| MPT-7B                | 30.06     | 23.5      | 21.3      | 27.5      | 25.9      | 75        | 2.9       | 9.1       | 17.1      | 22.8      | 35.6      | 70        |
+| Qwen1.5-7B            | 55.12     | 73.57     | **50.8**  | 62.15     | 71.84     | 72.62     | **20.36** | 54.36     | **53.05** | 36.8      | 40.01     | 70.74     |
+| Qwen-7B               | 49.53     | 63.4      | 45.3      | 59.7      | 62.5      | 75        | 13.3      | 54.1      | 27.4      | 31.4      | 45.2      | 67.5      |
+| XVERSE-7B             | 34.27     | 61.1      | 39        | 58.4      | 60.8      | 73.7      | 2.2       | 11.7      | 4.9       | 10.2      | 31        | 24        |
+| Yi-6B                 | 47.8      | 73        | 44.3      | 64        | **73.5**  | 73.1      | 6.3       | 39.9      | 15.2      | 23.6      | 44.9      | 68        |
+| **360Zhinao-7B**      | 56.15     | **74.11** | 49.49     | **67.44** | 72.38     | **83.05** | 16.38     | 53.83     | 35.98     | 42.4      | 43.95     | **78.59** |
+
+The above results could be viewed or reproduced on [Opencompass](https://rank.opencompass.org.cn/leaderboard-llm).
+
+## Chat Models
+
+  We adopted a two-stage approach to train the long context models.
+
+  **First stage**: We increased RoPE base and extended the context length to 32K.
+    - Firstly, we performed Continual Pretraining on approximately 5B tokens with a 32K context window.
+    - Then during the SFT stage, we fine-tuned the model using long data from various sources, including high-quality human-labeled 32K data.
+
+  **Second stage**: We extended the context length to 360K, training with the following data:
+    - A small amount of high-quality human-labeled super-long data.
+    - Due to the scarcity of annotated super-long data, we constructed various forms of synthetic data.
+        - Multi-Doc QA: Similar to [Ziya-Reader](https://arxiv.org/abs/2311.09198), we generated multi-document QA pairs based on 360's database. Multiple QA pairs are constructed for one row of Multi-Doc QA data input, resulting in a multi-turn format and significantly improving the training efficiency.
+        - Single-Doc QA: Similar to [LLama2 Long](https://arxiv.org/abs/2309.16039), we constructed multi-turn QA data based on different segments within one row of long-text input.
+
+We evaluated our models across various lengths and benchmarks.
+
+- ### Long Context Benchmarks
+
+
+  We evaluated our 32K and 360K models on [LongBench](https://github.com/THUDM/LongBench), a multi-task bilingual benchmark for long contexts. We report results on Chinese tasks that are the most relevant to downstream applications: Single/Multi-Doc QA, Summarization, Few-Shot Learning and Code Completion.
+
+    | Model                     | Avg       | 单文档QA  | 多文档QA   | 摘要       | Few-shot学习 | 代码补全    |
+    | :------------------------ |:---------:|:--------:|:---------:|:---------:|:------------:|:---------:|
+    | GPT-3.5-Turbo-16k         | 37.84     | 61.2     | 28.7      | 16        | 29.2         | 54.1      |
+    | ChatGLM2-6B-32k           | 37.16     | 51.6     | 37.6      | 16.2      | 27.7         | 52.7      |
+    | ChatGLM3-6B-32k           | 44.62     | **62.3** | 44.8      | 17.8      | 42           | 56.2      |
+    | InternLM2-Chat-7B         | 42.20     | 56.65    | 29.15     | **17.99** | 43.5         | **63.72** |
+    | Qwen1.5-Chat-7B           | 36.75     | 52.85    | 30.08     | 14.28     | 32           | 54.55     |
+    | Qwen1.5-Chat-14B          | 39.80     | 60.39    | 27.99     | 14.77     | 37           | 58.87     |
+    | 360Zhinao-7B-Chat-32K     | **45.18** | 57.18    | **48.06** | 15.03     | **44**       | 61.64     |
+
+- ### 360Zhinao-7B-Chat-360K on "NeedleInAHaystack"
+
+  [NeedleInAHaystack](https://github.com/gkamradt/LLMTest_NeedleInAHaystack) places one small piece of information in different positions of long text and queries this information as a test of LLM's long-context capabilities.
+
+  360Zhinao-7B-Chat-360K could achieve over 98% accuracy on both English and Chinese NeedleInAHaystack tasks.
+
+  - English version（same as [NeedleInAHaystack](https://github.com/gkamradt/LLMTest_NeedleInAHaystack)）
+  
+    <p align="center">
+        <img src="assets/360Zhinao-7B-Chat-360K.en_score.png" width="600" />
+    <p>
+
+    **needle**：The best thing to do in San Francisco is eat a sandwich and sit in Dolores Park on a sunny day.
+
+    **query**：What is the best thing to do in San Francisco?
+
+
+  - Chinese version
+
+    <p align="center">
+        <img src="assets/360Zhinao-7B-Chat-360K.zh_score.png" width="600" />
+    <p>
+
+    We constructed the Chinese version following the [SuperCLUE-200K benchmark](https://mp.weixin.qq.com/s/QgoRf2LB-7vc3vTFOHJkpw):
+
+    **haystack**：Chinese novels.
+  
+    **needle**：(in Chinese) 王莽是一名勤奋的店员，他每天凌晨就起床，赶在第一缕阳光照亮大地之前到达店铺，为即将开始的一天做准备。他清扫店铺，整理货架，为顾客提供方便。他对五金的种类和用途了如指掌，无论顾客需要什么，他总能准确地找到。\n然而，他的老板刘秀却总是对他吹毛求疵。刘秀是个挑剔的人，他总能在王莽的工作中找出一些小错误，然后以此为由扣他的工资。他对王莽的工作要求非常严格，甚至有些过分。即使王莽做得再好，刘秀也总能找出一些小问题，让王莽感到非常沮丧。\n王莽虽然对此感到不满，但他并没有放弃。他知道，只有通过自己的努力，才能获得更好的生活。他坚持每天早起，尽管他知道那天可能会再次被刘秀扣工资。他始终保持微笑，尽管他知道刘秀可能会再次对他挑剔。
+
+    **query**：(in Chinese) 王莽在谁的手下工作？
+
+<br>
 
 # Quickstart
 Simple examples to illustrate how to use 360Zhinao-7B-Base and 360Zhinao-7B-Chat quickly using 🤖 ModelScope and 🤗 Transformers
@@ -189,7 +258,7 @@ pred = model.generate(input_ids=inputs["input_ids"], generation_config=generatio
 print("outputs:\n", tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))
 ```
 
-### Demonstration of Base Model Inference
+### Demonstration of Chat Model Inference
 
 This code demonstrates using ModelScope to quickly use the 360Zhinao-7B-Chat-4K model for inference.
 
@@ -270,17 +339,11 @@ curl --location --request POST 'http://localhost:8360/v1/chat/completions' \
 }'
 ```
 
+<br>
+
 # Model Inference
 ## Quantization
-We provide quantization schemes based on AutoGPTQ and open source the Int4 quantization models. The quantization model has little effect loss, but it can significantly reduce the video memory occupation and improve the inference speed.
-
-The BF16, Int8, and Int4 models are tested on the benchmarks, and the results are as follows:
-
-| Quantization | MMLU | CEval (val) | GSM8K | Humaneval |
-|-|-|-|-|-|
-| 360Zhinao-7B-Chat-4K (BF16) |-|-|-|-|
-| 360Zhinao-7B-Chat-4K (Int8) |-|-|-|-|
-| 360Zhinao-7B-Chat-4K (Int4) |-|-|-|-|
+We provide quantization schemes based on AutoGPTQ and open source the Int4 quantization models. 
 
 ## Deployment
 ### vLLM Installation
@@ -295,7 +358,8 @@ Otherwise, please refer to the official vLLM [Installation Instructions](https:/
 
 >Once the installation is complete, you will need to do the following
 1. Copy the vllm/zhinao.py file to the vllm/model_executor/models directory corresponding to your env environment.
-2. Then add a line to vllm/model_executor/models/\_\_init\_\_.py
+2. Copy the vllm/serving_chat.py file to the vllm/entrypoints/openai corresponding to your env environment.
+3. Then add a line to vllm/model_executor/models/\_\_init\_\_.py
 
     ```shell
     "ZhinaoForCausalLM": ("zhinao", "ZhinaoForCausalLM"),
@@ -307,10 +371,10 @@ Starting the service
 ```shell
 python -m vllm.entrypoints.openai.api_server \
     --served-model-name 360Zhinao-7B-Chat-4K \
-    --model qihoo360/360Zhinao-7B-Chat-4k \
+    --model qihoo360/360Zhinao-7B-Chat-4K \
     --trust-remote-code \
-    --tensor-parallel-size 1
-    --max-model-len 18000 \
+    --tensor-parallel-size 1 \
+    --max-model-len 4096 \
     --host 0.0.0.0 \
     --port 8360
 ```
@@ -322,7 +386,7 @@ curl http://localhost:8360/v1/chat/completions \
 -d '{
     "model": "360Zhinao-7B-Chat-4K",
     "max_tokens": 200,
-    "top_k": 0,
+    "top_k": -1,
     "top_p": 0.8,
     "temperature": 1.0,
     "presence_penalty": 0.0,
@@ -342,7 +406,7 @@ Use python to request the service
 ```python
 from openai import OpenAI
 openai_api_key = "EMPTY"
-openai_api_base = "http://localhost:8000/v1"
+openai_api_base = "http://localhost:8360/v1"
 
 client = OpenAI(
     api_key=openai_api_key,
@@ -370,6 +434,8 @@ print("Chat response:", chat_response)
 
 > 
 
+<br>
+
 # Model Finetune
 ## Training data
 
@@ -458,6 +524,8 @@ bash finetune/ds_finetune.sh
 - By configuring the **fp16**、**bf16** realize mixed precision training, bf16 is recommended to be consistent with the pre-trained model.
 - By configuring **is_concat**, Whether the training data is concatenated or not is controlled. When the magnitude of the training data is large, the training efficiency can be improved by concatenation.
 
+<br>
+
 # License
 
 The source code of this warehouse follows the open source license Apache 2.0.