liuyongq commited on
Commit
51775c9
1 Parent(s): 5196976

Create README_en.md

Browse files
Files changed (1) hide show
  1. README_en.md +270 -0
README_en.md ADDED
@@ -0,0 +1,270 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <!-- markdownlint-disable first-line-h1 -->
2
+ <!-- markdownlint-disable html -->
3
+ ![](./pics/orion_start.PNG)
4
+
5
+ <div align="center">
6
+ <h1>
7
+ Orion-14B
8
+ </h1>
9
+ </div>
10
+
11
+ <p align="center">
12
+ 🤗 <a href="https://huggingface.co/OrionStarAI/Orion-14B" target="_blank">HuggingFace Model Download</a> | 🤖 <a href="https://modelscope.cn/models/OrionStarAI/Orion-14B/summary" target="_blank">ModelScope Download</a>
13
+ </p>
14
+
15
+ <div align="center">
16
+
17
+
18
+ <h4 align="center">
19
+ <p>
20
+ <b>🌐English</b> |
21
+ <a href="https://huggingface.co/OrionStarAI/Orion-14B-Base/new/main/README.md">🇨🇳中文</a>
22
+ <p>
23
+ </h4>
24
+
25
+ </div>
26
+
27
+ # Table of Contents
28
+
29
+ - [📖 Model Introduction](#model-introduction)
30
+ - [🔗 Model Download](#model-download)
31
+ - [🔖 Model Benchmark](#model-benchmark)
32
+ - [📊 Model Inference](#model-inference)
33
+ - [🥇 Company Introduction](#company-introduction)
34
+ - [📜 Declarations & License](#declarations-license)
35
+
36
+ # Model Introduction
37
+
38
+ - Orion-14B series models are open-source multilingual large language models trained from scratch by OrionStarAI. The base model is trained on 2.5T multilingual corpus, including Chinese, English, Japanese, Korean, etc, and it exhibits superior performance in these languages.
39
+
40
+ - In mainstream benchmark evaluations, the Orion-14B series models demonstrate outstanding competitiveness, significantly surpassing models of similar scales. Based on benchmark results, the Orion-14B series models are also the first to be evaluated across more than three languages in the domain of LLM. We hope that the contributions of all Orion Star colleagues establish a new benchmark for the research field of multilingual LLMs.
41
+
42
+
43
+ # Model Download
44
+
45
+ Model release and download links are provided in the table below:
46
+
47
+ | Model Name | HuggingFace Download Links | ModelScope Download Links |
48
+ |-------------------------|-----------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------|
49
+ | ⚾Orion-14B-Base | [Orion-14B-Base](https://huggingface.co/OrionStarAI/Orion-14B-Base) | [Orion-14B-Base](https://modelscope.cn/models/OrionStarAI/Orion-14B-Base/summary) |
50
+ | 😛Orion-14B-Chat | [Orion-14B-Chat](https://huggingface.co/OrionStarAI/Orion-14B-Chat) | [Orion-14B-Chat](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat/summary) |
51
+ | 📃Orion-14B-LongChat | [Orion-14B-LongChat](https://huggingface.co/OrionStarAI/Orion-14B-LongChat) | [Orion-14B-LongChat](https://modelscope.cn/models/OrionStarAI/Orion-14B-LongChat/summary) |
52
+ | 🔎Orion-14B-Chat-RAG | [Orion-14B-Chat-RAG](https://huggingface.co/OrionStarAI/Orion-14B-Chat-RAG) | [Orion-14B-Chat-RAG](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat-RAG/summary) |
53
+ | 🔌Orion-14B-Chat-Plugin | [Orion-14B-Chat-Plugin](https://huggingface.co/OrionStarAI/Orion-14B-Chat-Plugin) | [Orion-14B-Chat-Plugin](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat-Plugin/summary) |
54
+ | 💼Orion-14B-Base-Int4 | [Orion-14B-Base-Int4](https://huggingface.co/OrionStarAI/Orion-14B-Base-Int4) | [Orion-14B-Base-Int4](https://modelscope.cn/models/OrionStarAI/Orion-14B-Base-Int4/summary) |
55
+ | 📦Orion-14B-Chat-Int4 | [Orion-14B-Chat-Int4](https://huggingface.co/OrionStarAI/Orion-14B-Chat-Int4) | [Orion-14B-Chat-Int4](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat-Int4/summary) |
56
+
57
+ # Model Benchmarks
58
+
59
+ ## LLM evaluation results on examination and professional knowledge
60
+ | Model | C-Eval | CMMLU | MMLU | AGIEval | Gaokao | BBH |
61
+ |--------------------|----------|----------|----------|----------|----------|----------|
62
+ | LLaMA2-13B | 41.4 | 38.4 | 55.0 | 30.9 | 18.2 | 45.6 |
63
+ | Skywork-13B | 59.1 | 61.4 | 62.7 | 43.6 | 56.1 | 48.3 |
64
+ | Baichuan2-13B | 59.0 | 61.3 | 59.5 | 37.4 | 45.6 | 49.0 |
65
+ | QWEN-14B | 71.7 | 70.2 | 67.9 | 51.9 | **62.5** | 53.7 |
66
+ | InternLM-20B | 58.8 | 59.0 | 62.1 | 44.6 | 45.5 | 52.5 |
67
+ | **Orion-14B** | **72.9** | **70.6** | **69.9** | **54.7** | 62.1 | **56.5** |
68
+
69
+ ## LLM evaluation results on language understanding and common knowledge
70
+ | Model |RACE-middle|RACE-high |HellaSwag | PIQA | Lambada | WSC |
71
+ |--------------------|----------|----------|----------|----------|----------|----------|
72
+ | LLaMA 2-13B | 63.0 | 58.9 | 77.5 | 79.8 | 76.5 | 66.3 |
73
+ | Skywork-13B | 87.6 | 84.1 | 73.7 | 78.3 | 71.8 | 66.3 |
74
+ | Baichuan 2-13B | 68.9 | 67.2 | 70.8 | 78.1 | 74.1 | 66.3 |
75
+ | QWEN-14B | 93.0 | 90.3 | **80.2** | 79.8 | 71.4 | 66.3 |
76
+ | InternLM-20B | 86.4 | 83.3 | 78.1 | **80.3** | 71.8 | 68.3 |
77
+ | **Orion-14B** | **93.3** | **91.3** | 78.5 | 79.5 | **78.9** | **70.2** |
78
+
79
+ ## LLM evaluation results of OpenCompass testsets
80
+ | Model | Average | Examination | Language | Knowledge | Understanding | Reasoning |
81
+ |-----------------|----------|----------|----------|----------|----------|----------|
82
+ | LLaMA 2-13B | 47.3 | 45.2 | 47.0 | 58.3 | 50.9 | 43.6 |
83
+ | Skywork-13B | 53.6 | 61.1 | 51.3 | 52.7 | 64.5 | 45.2 |
84
+ | Baichuan 2-13B | 49.4 | 51.8 | 47.5 | 48.9 | 58.1 | 44.2 |
85
+ | QWEN-14B | 62.4 | 71.3 | 52.67 | 56.1 | 68.8 | 60.1 |
86
+ | InternLM-20B | 59.4 | 62.5 | 55.0 | **60.1** | 67.3 | 54.9 |
87
+ | **Orion-14B** | **64.4** | **71.4** | **55.0** | 60.0 | **71.9** | **61.6** |
88
+
89
+ ## Comparison of LLM performances on Japanese testsets
90
+ | Model |**Average**| JCQA | JNLI | MARC | JSQD | JQK | XLS | XWN | MGSM |
91
+ |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------|
92
+ | PLaMo-13B | 52.3 | 56.7 | 42.8 | 95.8 | 70.6 | 71.0 | 8.70 | 70.5 | 2.40 |
93
+ | WebLab-10B | 50.7 | 66.6 | 53.7 | 82.1 | 62.9 | 56.2 | 10.0 | 72.0 | 2.40 |
94
+ | ELYZA-jp-7B | 48.8 | 71.7 | 25.3 | 86.6 | 70.8 | 64.1 | 2.50 | 62.1 | 7.20 |
95
+ | StableLM-jp-7B | 51.1 | 33.4 | 43.3 | **96.7** | 70.6 | 78.1 | 10.7 | 72.8 | 2.80 |
96
+ | LLaMA 2-13B | 46.3 | 75.0 | 47.6 | 38.8 | 76.1 | 67.7 | 18.1 | 63.2 | 10.4 |
97
+ | Baichuan 2-13B | 57.1 | 73.7 | 31.3 | 91.6 | 80.5 | 63.3 | 18.6 | 72.2 | 25.2 |
98
+ | QWEN-14B | 65.8 | 85.9 | 60.7 | 97.0 | 83.3 | 71.8 | 18.8 | 70.6 | 38.0 |
99
+ | Yi-34B | 67.1 | 83.8 | 61.2 | 95.2 | **86.1** | 78.5 | **27.2** | 69.2 | 35.2 |
100
+ | **Orion-14B** | **69.1** | **88.2** | **75.8** | 94.1 | 75.7 | **85.1** | 17.3 | **78.8** | **38.0** |
101
+
102
+ ## Comparison of LLM performances on Korean testsets. n = 0 and n = 5 stand for n-shot prompts used in the evaluation
103
+ |Model | **Average**<br>n=0&nbsp;&nbsp;n=5 | HellaSwag<br>n=0&nbsp;&nbsp;n=5 | COPA<br> n=0&nbsp;&nbsp;n=5 | BooIQ<br>n=0&nbsp;&nbsp;n=5 | SentiNeg<br>n=0&nbsp;&nbsp;n=5|
104
+ |-----------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
105
+ | KoGPT | 53.0 &nbsp;&nbsp; 70.1 | 55.9 &nbsp;&nbsp; 58.3 | 73.5 &nbsp;&nbsp; 72.9 | 45.1 &nbsp;&nbsp; 59.8 | 37.5 &nbsp;&nbsp; 89.4 |
106
+ | Polyglot-ko-13B | 69.6 &nbsp;&nbsp; 73.7 |**59.5** &nbsp;&nbsp; **63.1**|**79.4** &nbsp;&nbsp; **81.1**| 48.2 &nbsp;&nbsp; 60.4 | 91.2 &nbsp;&nbsp; 90.2 |
107
+ | LLaMA 2-13B | 46.7 &nbsp;&nbsp; 63.7 | 41.3 &nbsp;&nbsp; 44.0 | 59.3 &nbsp;&nbsp; 63.8 | 34.9 &nbsp;&nbsp; 73.8 | 51.5 &nbsp;&nbsp; 73.4 |
108
+ | Baichuan 2-13B | 52.1 &nbsp;&nbsp; 58.7 | 39.2 &nbsp;&nbsp; 39.6 | 60.6 &nbsp;&nbsp; 60.6 | 58.4 &nbsp;&nbsp; 61.5 | 50.3 &nbsp;&nbsp; 72.9 |
109
+ | QWEN-14B | 53.8 &nbsp;&nbsp; 73.7 | 45.3 &nbsp;&nbsp; 46.8 | 64.9 &nbsp;&nbsp; 68.9 | 33.4 &nbsp;&nbsp; 83.5 | 71.5 &nbsp;&nbsp; 95.7 |
110
+ | Yi-34B | 54.2 &nbsp;&nbsp; 72.1 | 44.6 &nbsp;&nbsp; 44.7 | 58.0 &nbsp;&nbsp; 60.6 | 65.9 &nbsp;&nbsp; 90.2 | 48.3 &nbsp;&nbsp; 92.9 |
111
+ | **Orion-14B** |**74.5** &nbsp;&nbsp; **79.6**| 47.0 &nbsp;&nbsp; 49.6 | 77.7 &nbsp;&nbsp; 79.4 |**81.6** &nbsp;&nbsp; **90.7**|**92.4** &nbsp;&nbsp; **98.7**|
112
+
113
+ ## Multilingual evaluation
114
+ | Model | Train Lang | Japanese | Korean | Chinese | English |
115
+ |--------------------|------------|----------|----------|----------|----------|
116
+ | PLaMo-13B | En,Jp | 52.3 | * | * | * |
117
+ | Weblab-10B | En,Jp | 50.7 | * | * | * |
118
+ | ELYZA-jp-7B | En,Jp | 48.8 | * | * | * |
119
+ | StableLM-jp-7B | En,Jp | 51.1 | * | * | * |
120
+ | KoGPT-6B | En,Ko | * | 70.1 | * | * |
121
+ | Polyglot-ko-13B | En,Ko | * | 70.7 | * | * |
122
+ | Baichuan2-13B | Multi | 57.1 | 58.7 | 50.8 | 57.1 |
123
+ | Qwen-14B | Multi | 65.8 | 73.7 | 64.5 | 65.4 |
124
+ | Llama2-13B | Multi | 46.3 | 63.7 | 41.4 | 55.3 |
125
+ | Yi-34B | Multi | 67.1 | 72.2 | 58.7 | **68.8** |
126
+ | **Orion-14B** | Multi | **69.1** | **79.5** | **67.9** | 67.3 |
127
+
128
+ ## Evaluation for data contamination
129
+ | Model | C-Eval | CMMLU | MMLU |Lambada |HellaSwag |
130
+ |------------------------|----------|----------|----------|----------|----------|
131
+ | GPT-4 | 69.9 | 71.0 | 83.0 | 65.5 | **91.4** |
132
+ | Qwen-72B | 83.3 | 61.8 | 77.3 | 76.1 | 85.4 |
133
+ | Yi-34B | 81.8 | 82.6 | 76.3 | 73.1 | 82.0 |
134
+ | Orion-14B | 72.8 | 70.6 | 69.9 | 78.8 | 78.5 |
135
+ | Orion-14B(contaminated)| **92.7** | **82.9** | **85.4** | **78.5** | 85.8 |
136
+
137
+ ## Chat model standard evaluation
138
+ | Model | CMMLU | MMLU | BBH |HellaSwag | PIQA | WSC |
139
+ |----------------------|----------|----------|----------|----------|----------|----------|
140
+ | Baichuan2-13B-Chat | 58.4 | 57.0 | 49.9 | 66.9 | 77.6 | **71.2** |
141
+ | Qwen-14B-Chat | **70.0** | **66.4** | **58.0** | 65.2 | 74.0 | 66.3 |
142
+ | Llama2-13B-Chat | 38.7 | 54.6 | 40.2 | **78.2** | **78.8** | 68.3 |
143
+ | InternLM-20B-Chat | 52.2 | 52.5 | 35.3 | 69.2 | 76.7 | 61.5 |
144
+ | **Orion-14B-Chat** | 63.7 | 61.71 | 49.05 | 76.7 | 78.4 | 71.15 |
145
+
146
+ ## Chat model subjective evaluation of MTBench
147
+ | Model | First-Turn | Second-Turn | **Average** |
148
+ |----------------------|----------|----------|----------|
149
+ | Baichuan2-13B-Chat | 7.05 | 6.47 | 6.76 |
150
+ | Qwen-14B-Chat | 7.30 | 6.62 | 6.96 |
151
+ | Llama2-13B-Chat | 7.10 | 6.20 | 6.65 |
152
+ | InternLM-20B-Chat | 7.03 | 5.93 | 6.48 |
153
+ | **Orion-14B-Chat** | **7.68** | **7.07** | **7.37** |
154
+
155
+ ## Chat model subjective evaluation of AlignBench
156
+ | Model | Math. | Logi. | Basic. | Chi. | Comp. | Writ. | Role. | Prof. |**Avg.**|
157
+ |--------------------|--------|--------|--------|--------|--------|--------|--------|--------|--------|
158
+ | Baichuan2-13B-Chat | 3.76 | 4.07 | 6.22 | 6.05 | 7.11 | 6.97 | 6.75 | 6.43 | 5.25 |
159
+ | Qwen-14B-Chat |**4.91**|**4.71**|**6.90**| 6.36 | 6.74 | 6.64 | 6.59 | 6.56 |**5.72**|
160
+ | Llama2-13B-Chat | 3.05 | 3.79 | 5.43 | 4.40 | 6.76 | 6.63 | 6.99 | 5.65 | 4.70 |
161
+ | InternLM-20B-Chat | 3.39 | 3.92 | 5.96 | 5.50 |**7.18**| 6.19 | 6.49 | 6.22 | 4.96 |
162
+ | Orion-14B-Chat | 4.00 | 4.24 | 6.18 |**6.57**| 7.16 |**7.36**|**7.16**|**6.99**| 5.51 |
163
+
164
+
165
+ # Model Inference
166
+
167
+ Model weights, source code, and configuration needed for inference are published on Hugging Face, and the download link
168
+ is available in the table at the beginning of this document. We demonstrate various inference methods here, and the
169
+ program will automatically download the necessary resources from Hugging Face.
170
+
171
+ ## Python Code
172
+
173
+ ```python
174
+ import torch
175
+ from transformers import AutoModelForCausalLM, AutoTokenizer
176
+ from transformers.generation.utils import GenerationConfig
177
+
178
+ tokenizer = AutoTokenizer.from_pretrained("OrionStarAI/Orion-14B", use_fast=False, trust_remote_code=True)
179
+ model = AutoModelForCausalLM.from_pretrained("OrionStarAI/Orion-14B", device_map="auto",
180
+ torch_dtype=torch.bfloat16, trust_remote_code=True)
181
+
182
+ model.generation_config = GenerationConfig.from_pretrained("OrionStarAI/Orion-14B")
183
+ messages = [{"role": "user", "content": "Hello, what is your name? "}]
184
+ response = model.chat(tokenizer, messages, streaming=False)
185
+ print(response)
186
+
187
+ # Hello, my name is ChatMax. Nice to meet you.
188
+ ```
189
+
190
+ In the above Python code, the model is loaded with `device_map='auto'` to utilize all available GPUs. To specify the
191
+ device, you can use something like `export CUDA_VISIBLE_DEVICES=0,1` (using GPUs 0 and 1).
192
+
193
+ ## Command Line Tool
194
+
195
+ ```shell
196
+ python cli_demo.py
197
+ ```
198
+
199
+ This command-line tool is designed for chat scenarios, and thus, it does not support calling the base model.
200
+
201
+ ## Direct Script Inference
202
+
203
+ ```shell
204
+ python text_generation.py --model OrionStarAI/Orion-14B --tokenizer OrionStarAI/Orion-14B --prompt 你好,你叫什么名字
205
+ ```
206
+
207
+ # Example Output
208
+
209
+ ## Casual Chat
210
+
211
+ `````
212
+ User: Hello, what is your name?
213
+ Orion-14B: Hello, my name is ChatMax. Nice to meet you.
214
+ User: What functions do you have?
215
+ Orion-14B: I am a versatile AI assistant. I can answer your questions, chat with you, help you look up information, provide life advice, and more.
216
+ User: Tell me a joke.
217
+ Orion-14B: Why is sausage always unhappy? Because it has no friends and can only be sandwiched between bread.
218
+ `````
219
+
220
+ ## Mathematical Problem
221
+
222
+ `````
223
+ User: In a cage, there are chickens and rabbits with a total of 26 heads and 68 feet. So, how many chickens and how many rabbits are there?
224
+ Orion-14B: Let the number of chickens be x, and the number of rabbits be y.
225
+ According to the problem, we can derive the following two equations:
226
+ x + y = 26 (Total number of heads)
227
+ 2x + 4y = 68 (Total number of feet)
228
+ Rewriting the first equation, we get y = 26 - x.
229
+ Substituting the value of y into the second equation, we get:
230
+ 2x + 4(26 - x) = 68
231
+ Solving this equation, we find x = 18.
232
+ Therefore, there are 18 chickens and 26 - 18 = 8 rabbits.
233
+ `````
234
+
235
+ # Company Introduction
236
+
237
+ OrionStar is a leading global service robot solutions company, founded in September 2016. OrionStar is dedicated to
238
+ using artificial intelligence technology to create the next generation of revolutionary robots, allowing people to break
239
+ free from repetitive physical labor and making human work and life more intelligent and enjoyable. Through technology,
240
+ OrionStar aims to make society and the world a better place.
241
+
242
+ OrionStar possesses fully self-developed end-to-end artificial intelligence technologies, such as voice interaction and
243
+ visual navigation. It integrates product development capabilities and technological application capabilities. Based on
244
+ the Orion robotic arm platform, it has launched products such as OrionStar AI Robot Greeting, AI Robot Greeting Mini,
245
+ Lucki, Coffee Master, and established the open platform OrionOS for Orion robots. Following the philosophy of "Born for
246
+ Truly Useful Robots", OrionStar empowers more people through AI technology.
247
+
248
+ # Declarations, License
249
+
250
+ ## Declarations
251
+
252
+ We strongly urge all users not to use the Orion-14B model for any activities that may harm national or social security or violate the law.
253
+ Additionally, we request users not to use the Orion-14B model for internet services without proper security review and filing.
254
+ We hope all users abide by this principle to ensure that technological development takes place in a regulated and legal environment.
255
+ We have done our best to ensure the compliance of the data used in the model training process. However, despite our
256
+ significant efforts, unforeseen issues may still arise due to the complexity of the model and data. Therefore, if any
257
+ problems arise due to the use of the Orion-14B open-source model, including but not limited to data security
258
+ issues, public opinion risks, or any risks and issues arising from the model being misled, abused, disseminated, or
259
+ improperly utilized, we will not assume any responsibility.
260
+
261
+ ## License
262
+
263
+ Community use of the Orion-14B model must comply with
264
+ the [Apache 2.0](https://github.com/OrionStarAI/Orion-14B/blob/main/LICENSE).
265
+
266
+ # Contact Us
267
+
268
+ Email: ai@orionstar.com
269
+
270
+ WhatsApp Group: https://chat.whatsapp.com/J30ig8Dx4ja5jc0cfx2nVs