Du Chen commited on
Commit
23cd9dc
1 Parent(s): cbd5ffb

update readme

Browse files
Files changed (2) hide show
  1. README.md +71 -63
  2. README_cn.md +68 -62
README.md CHANGED
@@ -11,7 +11,9 @@ pipeline_tag: text-generation
11
 
12
  <!-- markdownlint-disable first-line-h1 -->
13
  <!-- markdownlint-disable html -->
14
- ![](./assets/imgs/orion_start.PNG)
 
 
15
 
16
  <div align="center">
17
  <h1>
@@ -26,7 +28,7 @@ pipeline_tag: text-generation
26
  <p>
27
  <b>🌐English</b> |
28
  <a href="https://huggingface.co/OrionStarAI/Orion-14B-Chat-RAG/blob/main/README_cn.md">🇨🇳中文</a><br><br>
29
- 🤗 <a href="https://huggingface.co/OrionStarAI" target="_blank">HuggingFace Mainpage</a> | 🤖 <a href="https://modelscope.cn/organization/OrionStarAI" target="_blank">ModelScope Mainpage</a><br>🎬 <a href="https://huggingface.co/spaces/OrionStarAI/Orion-14B-App-Demo" target="_blank">HuggingFace Demo</a> | 🎫 <a href="https://modelscope.cn/studios/OrionStarAI/Orion-14B-App-Demo/summary" target="_blank">ModelScope Demo</a>
30
  <p>
31
  </h4>
32
 
@@ -40,14 +42,14 @@ pipeline_tag: text-generation
40
  - [🔗 Model Download](#model-download)
41
  - [🔖 Model Benchmark](#model-benchmark)
42
  - [📊 Model Inference](#model-inference)
43
- - [🥇 Company Introduction](#company-introduction)
44
  - [📜 Declarations & License](#declarations-license)
 
45
 
46
- # Model Introduction
47
 
48
  - **Orion-14B-Chat-RAG:** A chat-model fine-tuned on a custom retrieval augmented generation dataset, achieving superior performance in retrieval augmented generation tasks.
49
 
50
- - 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.
51
 
52
  - The Orion-14B series models exhibit the following features:
53
  - Among models with 20B-parameter scale level, Orion-14B-Base model shows outstanding performance in comprehensive evaluations.
@@ -55,7 +57,9 @@ pipeline_tag: text-generation
55
  - The fine-tuned models demonstrate strong adaptability, excelling in human-annotated blind tests.
56
  - The long-chat version supports extremely long texts, extending up to 200K tokens.
57
  - The quantized versions reduce model size by 70%, improve inference speed by 30%, with performance loss less than 1%.
58
- ![](./assets/imgs/model_cap_en.png)
 
 
59
 
60
  - Orion-14B series models including:
61
  - **Orion-14B-Base:** A multilingual large language foundational model with 14 billion parameters, pretrained on a diverse dataset of 2.5 trillion tokens.
@@ -66,7 +70,7 @@ pipeline_tag: text-generation
66
  - **Orion-14B-Base-Int4:** A quantized base model utilizing 4-bit integer weights. It significantly reduces the model size by 70% and increases the inference speed by 30% while incurring a minimal performance loss of only 1%.
67
  - **Orion-14B-Chat-Int4:** A quantized chat model utilizing 4-bit integer weights.
68
 
69
- # Model Download
70
 
71
  Model release and download links are provided in the table below:
72
 
@@ -80,10 +84,10 @@ Model release and download links are provided in the table below:
80
  | 💼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) |
81
  | 📦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) |
82
 
83
- # Model Benchmarks
84
 
85
- ## 1. Base Model Benchmarks
86
- ### LLM evaluation results on examination and professional knowledge
87
  | Model | C-Eval | CMMLU | MMLU | AGIEval | Gaokao | BBH |
88
  |--------------------|----------|----------|----------|----------|----------|----------|
89
  | LLaMA2-13B | 41.4 | 38.4 | 55.0 | 30.9 | 18.2 | 45.6 |
@@ -93,7 +97,7 @@ Model release and download links are provided in the table below:
93
  | InternLM-20B | 58.8 | 59.0 | 62.1 | 44.6 | 45.5 | 52.5 |
94
  | **Orion-14B-Base** | **72.9** | **70.6** | **69.9** | **54.7** | 62.1 | **56.5** |
95
 
96
- ### LLM evaluation results on language understanding and common knowledge
97
  | Model |RACE-middle|RACE-high |HellaSwag | PIQA | Lambada | WSC |
98
  |--------------------|----------|----------|----------|----------|----------|----------|
99
  | LLaMA 2-13B | 63.0 | 58.9 | 77.5 | 79.8 | 76.5 | 66.3 |
@@ -103,7 +107,7 @@ Model release and download links are provided in the table below:
103
  | InternLM-20B | 86.4 | 83.3 | 78.1 | **80.3** | 71.8 | 68.3 |
104
  | **Orion-14B-Base** | **93.3** | **91.3** | 78.5 | 79.5 | **78.9** | **70.2** |
105
 
106
- ### LLM evaluation results of OpenCompass testsets
107
  | Model | Average | Examination | Language | Knowledge | Understanding | Reasoning |
108
  |------------------|----------|----------|----------|----------|----------|----------|
109
  | LLaMA 2-13B | 47.3 | 45.2 | 47.0 | 58.3 | 50.9 | 43.6 |
@@ -113,7 +117,7 @@ Model release and download links are provided in the table below:
113
  | InternLM-20B | 59.4 | 62.5 | 55.0 | **60.1** | 67.3 | 54.9 |
114
  |**Orion-14B-Base**| **64.4** | **71.4** | **55.0** | 60.0 | **71.9** | **61.6** |
115
 
116
- ### Comparison of LLM performances on Japanese testsets
117
  | Model |**Average**| JCQA | JNLI | MARC | JSQD | JQK | XLS | XWN | MGSM |
118
  |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------|
119
  | PLaMo-13B | 52.3 | 56.7 | 42.8 | 95.8 | 70.6 | 71.0 | 8.70 | 70.5 | 2.40 |
@@ -126,7 +130,7 @@ Model release and download links are provided in the table below:
126
  | Yi-34B | 67.1 | 83.8 | 61.2 | 95.2 | **86.1** | 78.5 | **27.2** | 69.2 | 35.2 |
127
  | **Orion-14B-Base** | **69.1** | **88.2** | **75.8** | 94.1 | 75.7 | **85.1** | 17.3 | **78.8** | **38.0** |
128
 
129
- ### Comparison of LLM performances on Korean testsets. n = 0 and n = 5 stand for n-shot prompts used in the evaluation
130
  |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|
131
  |------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
132
  | 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 |
@@ -137,7 +141,7 @@ Model release and download links are provided in the table below:
137
  | 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 |
138
  |**Orion-14B-Chat**|**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**|
139
 
140
- ### Multilingual evaluation
141
  | Model | Train Lang | Japanese | Korean | Chinese | English |
142
  |--------------------|------------|----------|----------|----------|----------|
143
  | PLaMo-13B | En,Jp | 52.3 | * | * | * |
@@ -153,8 +157,8 @@ Model release and download links are provided in the table below:
153
  | **Orion-14B-Chat** | Multi | **69.1** | **79.5** | **67.9** | 67.3 |
154
 
155
 
156
- ## 2. Chat Model Benchmarks
157
- ### Chat model subjective evaluation of MTBench
158
  | Model | First-Turn | Second-Turn | **Average** |
159
  |----------------------|----------|----------|----------|
160
  | Baichuan2-13B-Chat | 7.05 | 6.47 | 6.76 |
@@ -164,7 +168,7 @@ Model release and download links are provided in the table below:
164
  | **Orion-14B-Chat** | **7.68** | **7.07** | **7.37** |
165
  \* use vllm for inference
166
 
167
- ### Chat model subjective evaluation of AlignBench
168
  | Model | Math. | Logi. | Basic. | Chi. | Comp. | Writ. | Role. | Prof. |**Avg.**|
169
  |--------------------|--------|--------|--------|--------|--------|--------|--------|--------|--------|
170
  | Baichuan2-13B-Chat | 3.76 | 4.07 | 6.22 | 6.05 | 7.11 | 6.97 | 6.75 | 6.43 | 5.25 |
@@ -175,8 +179,8 @@ Model release and download links are provided in the table below:
175
 
176
  \* use vllm for inference
177
 
178
- ## 3. LongChat Model Benchmarks
179
- ### LongChat evaluation of LongBench
180
  | Model | NarrativeQA|MultiFieldQA-en|MultiFieldQA-zh| DuReader | QMSum | VCSUM | TREC | TriviaQA | LSHT |RepoBench-P|
181
  |--------------------------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|
182
  | GPT-3.5-Turbo-16k | **23.60** | **52.30** | **61.20** | 28.70 | 23.40 | **16.00** | 68.00 | **91.40** | 29.20 | 53.60 |
@@ -186,8 +190,8 @@ Model release and download links are provided in the table below:
186
  | Orion-14B-LongChat | 19.47 | 48.11 | 55.84 | **37.02** | **24.87** | 15.44 | **77.00** | 89.12 | **45.50** | 54.31 |
187
 
188
 
189
- ## 4. Chat RAG Model Benchmarks
190
- ### LLM evaluation results of self-built RAG testsets
191
  |Model|Effectiveness of Response(Keyword)|*Effectiveness of Response(subjective evaluation)|Quoting Ability|Fallback Ability|*AutoQA|*Data Extraction|
192
  |---------------------|------|------|------|------|------|------|
193
  | Baichuan2-13B-Chat | 85 | 76 | 1 | 0 | 69 | 51 |
@@ -197,8 +201,8 @@ Model release and download links are provided in the table below:
197
  | Orion-14B-Chat-RAG | 86 | 87 | 91 | 97 | 73 | 71 |
198
  \* means manual assessment
199
 
200
- ## 5. Chat Plugin Model Benchmarks
201
- ### LLM evaluation results of self-built plugin testsets
202
  |Model |Intent Recognition with Full Params |Intent Recognition with Missing Params |Non-Plugin Invocation Recognition |
203
  |-----------------------|--------|-----------|--------|
204
  | Baichuan2-13B-Chat | 25 | 0 | 0 |
@@ -206,20 +210,20 @@ Model release and download links are provided in the table below:
206
  | GPT-4 | **95** | 52.38 | 70 |
207
  | Orion-14B-Chat-Plugin | 92.5 | **60.32** | **90** |
208
 
209
- ## 6. Quantized Model Benchmarks
210
- ### Comparison of before and after quantization
211
  |Model |Size(GB)|Inference Speed(tokens/s)|C-Eval|CMMLU|MMLU|RACE|HellaSwag|
212
  |-------------------------|-------|-----|------|------|------|------|------|
213
  | OrionStar-14B-Base | 28.0 | 135 | 72.8 | 70.6 | 70.0 | 93.3 | 78.5 |
214
  | OrionStar-14B-Base-Int4 | 8.3 | 178 | 71.8 | 69.8 | 69.2 | 93.1 | 78.0 |
215
 
216
- # Model Inference
217
 
218
  Model weights, source code, and configuration needed for inference are published on Hugging Face, and the download link
219
  is available in the table at the beginning of this document. We demonstrate various inference methods here, and the
220
  program will automatically download the necessary resources from Hugging Face.
221
 
222
- ## Python Code
223
 
224
  ```python
225
  import torch
@@ -240,7 +244,7 @@ print(response)
240
  In the above Python code, the model is loaded with `device_map='auto'` to utilize all available GPUs. To specify the
241
  device, you can use something like `export CUDA_VISIBLE_DEVICES=0,1` (using GPUs 0 and 1).
242
 
243
- ## Command Line Tool
244
 
245
  ```shell
246
  CUDA_VISIBLE_DEVICES=0 python cli_demo.py
@@ -248,7 +252,7 @@ CUDA_VISIBLE_DEVICES=0 python cli_demo.py
248
 
249
  This command-line tool is designed for chat scenarios, and thus, it does not support calling the base model.
250
 
251
- ## Direct Script Inference
252
 
253
  ```shell
254
 
@@ -260,32 +264,31 @@ CUDA_VISIBLE_DEVICES=0 python demo/text_generation.py --model OrionStarAI/Orion-
260
 
261
  ```
262
 
263
- # Example Output
264
 
265
- ## Casual Chat
266
 
267
  `````
268
  User: Hello
269
  Orion-14B: Hello! How can I help you today? If you have any questions or need assistance, feel free to ask.
270
  User: What functions do you have?
271
  Orion-14B: As an AI language model, I have a wide range of functions and capabilities, including but not limited to:
272
-
273
- Answering general knowledge questions: I can provide information on various topics such as history, science, geography, and more.
274
- Providing definitions and explanations: I can explain complex concepts and provide definitions for words and phrases.
275
- Assisting with tasks: I can help with simple tasks like setting reminders, converting units of measurement, and even ordering food from restaurants.
276
- Generating text: I can write short stories, essays, articles, and other types of content based on prompts given to me.
277
- Translating languages: I can translate text from one language to another.
278
- Chatting: I can engage in conversation with users and respond to their queries in a conversational manner.
279
- Summarizing long texts: I can summarize longer texts like articles, books, and research papers.
280
- Generating random numbers and names: I can generate random numbers within a specified range and also suggest names based on certain criteria.
281
- Playing games: I can play simple games like Hangman, Trivia, and more.
282
- Providing recommendations: I can recommend movies, music, books, and other forms of entertainment based on user preferences.
283
- These are just some of the many functions that I am capable of performing. If you have a specific task in mind, feel free to ask me about it.
284
  User: Tell me a joke.
285
  Orion-14B: Sure, here's a classic one-liner: Why don't scientists trust atoms? Because they make up everything.
286
  `````
287
 
288
- ## Japanese & Korean Chat
289
 
290
  `````
291
  User:自己を紹介してください
@@ -294,22 +297,10 @@ User:자기소개를 해주세요.
294
  Orion-14B:안녕하세요, 제 이름은 ChatMax입니다. OrionStar에서 개발한 AI 어시스턴트입니다. 저는 사용자들에게 정확하고 객관적이며 유용한 답변을 제공하여 문제를 해결하고 정보를 얻는 데 도움을 줄 수 있습니다.
295
  `````
296
 
297
- # Company Introduction
298
-
299
- OrionStar is a leading global service robot solutions company, founded in September 2016. OrionStar is dedicated to
300
- using artificial intelligence technology to create the next generation of revolutionary robots, allowing people to break
301
- free from repetitive physical labor and making human work and life more intelligent and enjoyable. Through technology,
302
- OrionStar aims to make society and the world a better place.
303
 
304
- OrionStar possesses fully self-developed end-to-end artificial intelligence technologies, such as voice interaction and
305
- visual navigation. It integrates product development capabilities and technological application capabilities. Based on
306
- the Orion robotic arm platform, it has launched products such as OrionStar AI Robot Greeting, AI Robot Greeting Mini,
307
- Lucki, Coffee Master, and established the open platform OrionOS for Orion robots. Following the philosophy of "Born for
308
- Truly Useful Robots", OrionStar empowers more people through AI technology.
309
 
310
- # Declarations, License
311
-
312
- ## Declarations
313
 
314
  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.
315
  Additionally, we request users not to use the Orion-14B model for internet services without proper security review and filing.
@@ -320,15 +311,32 @@ problems arise due to the use of the Orion-14B open-source model, including but
320
  issues, public opinion risks, or any risks and issues arising from the model being misled, abused, disseminated, or
321
  improperly utilized, we will not assume any responsibility.
322
 
323
- ## License
324
 
325
  Community use of the Orion-14B series models
326
  - For code, please comply with [Apache License Version 2.0](./LICENSE)<br>
327
  - For model, please comply with [【Orion-14B Series】 Models Community License Agreement](./ModelsCommunityLicenseAgreement)
328
 
329
- # Contact Us
330
 
331
- Email: ai@orionstar.com
332
 
333
- ![](./assets/imgs/wechat_group.jpg)
 
 
 
334
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
 
12
  <!-- markdownlint-disable first-line-h1 -->
13
  <!-- markdownlint-disable html -->
14
+ <div align="center">
15
+ <img src="./assets/imgs/orion_start.PNG" alt="logo" width="50%" />
16
+ </div>
17
 
18
  <div align="center">
19
  <h1>
 
28
  <p>
29
  <b>🌐English</b> |
30
  <a href="https://huggingface.co/OrionStarAI/Orion-14B-Chat-RAG/blob/main/README_cn.md">🇨🇳中文</a><br><br>
31
+ 🤗 <a href="https://huggingface.co/OrionStarAI" target="_blank">HuggingFace Mainpage</a> | 🤖 <a href="https://modelscope.cn/organization/OrionStarAI" target="_blank">ModelScope Mainpage</a><br>🎬 <a href="https://huggingface.co/spaces/OrionStarAI/Orion-14B-App-Demo" target="_blank">HuggingFace Demo</a> | 🎫 <a href="https://modelscope.cn/studios/OrionStarAI/Orion-14B-App-Demo/summary" target="_blank">ModelScope Demo</a><br>📖 <a href="https://github.com/OrionStarAI/Orion/blob/master/doc/Orion14B_v3.pdf" target="_blank">Tech Report</a>
32
  <p>
33
  </h4>
34
 
 
42
  - [🔗 Model Download](#model-download)
43
  - [🔖 Model Benchmark](#model-benchmark)
44
  - [📊 Model Inference](#model-inference)
 
45
  - [📜 Declarations & License](#declarations-license)
46
+ - [🥇 Company Introduction](#company-introduction)
47
 
48
+ # 1. Model Introduction
49
 
50
  - **Orion-14B-Chat-RAG:** A chat-model fine-tuned on a custom retrieval augmented generation dataset, achieving superior performance in retrieval augmented generation tasks.
51
 
52
+ - 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. For details, please refer to [tech report](https://github.com/OrionStarAI/Orion/blob/master/doc/Orion14B_v3.pdf).
53
 
54
  - The Orion-14B series models exhibit the following features:
55
  - Among models with 20B-parameter scale level, Orion-14B-Base model shows outstanding performance in comprehensive evaluations.
 
57
  - The fine-tuned models demonstrate strong adaptability, excelling in human-annotated blind tests.
58
  - The long-chat version supports extremely long texts, extending up to 200K tokens.
59
  - The quantized versions reduce model size by 70%, improve inference speed by 30%, with performance loss less than 1%.
60
+ <div align="center">
61
+ <img src="./assets/imgs/model_cap_en.png" alt="model_cap_en" width="50%" />
62
+ </div>
63
 
64
  - Orion-14B series models including:
65
  - **Orion-14B-Base:** A multilingual large language foundational model with 14 billion parameters, pretrained on a diverse dataset of 2.5 trillion tokens.
 
70
  - **Orion-14B-Base-Int4:** A quantized base model utilizing 4-bit integer weights. It significantly reduces the model size by 70% and increases the inference speed by 30% while incurring a minimal performance loss of only 1%.
71
  - **Orion-14B-Chat-Int4:** A quantized chat model utilizing 4-bit integer weights.
72
 
73
+ # 2. Model Download
74
 
75
  Model release and download links are provided in the table below:
76
 
 
84
  | 💼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) |
85
  | 📦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) |
86
 
87
+ # 3. Model Benchmarks
88
 
89
+ ## 3.1. Base Model Orion-14B-Base Benchmarks
90
+ ### 3.1.1. LLM evaluation results on examination and professional knowledge
91
  | Model | C-Eval | CMMLU | MMLU | AGIEval | Gaokao | BBH |
92
  |--------------------|----------|----------|----------|----------|----------|----------|
93
  | LLaMA2-13B | 41.4 | 38.4 | 55.0 | 30.9 | 18.2 | 45.6 |
 
97
  | InternLM-20B | 58.8 | 59.0 | 62.1 | 44.6 | 45.5 | 52.5 |
98
  | **Orion-14B-Base** | **72.9** | **70.6** | **69.9** | **54.7** | 62.1 | **56.5** |
99
 
100
+ ### 3.1.2. LLM evaluation results on language understanding and common knowledge
101
  | Model |RACE-middle|RACE-high |HellaSwag | PIQA | Lambada | WSC |
102
  |--------------------|----------|----------|----------|----------|----------|----------|
103
  | LLaMA 2-13B | 63.0 | 58.9 | 77.5 | 79.8 | 76.5 | 66.3 |
 
107
  | InternLM-20B | 86.4 | 83.3 | 78.1 | **80.3** | 71.8 | 68.3 |
108
  | **Orion-14B-Base** | **93.3** | **91.3** | 78.5 | 79.5 | **78.9** | **70.2** |
109
 
110
+ ### 3.1.3. LLM evaluation results of OpenCompass testsets
111
  | Model | Average | Examination | Language | Knowledge | Understanding | Reasoning |
112
  |------------------|----------|----------|----------|----------|----------|----------|
113
  | LLaMA 2-13B | 47.3 | 45.2 | 47.0 | 58.3 | 50.9 | 43.6 |
 
117
  | InternLM-20B | 59.4 | 62.5 | 55.0 | **60.1** | 67.3 | 54.9 |
118
  |**Orion-14B-Base**| **64.4** | **71.4** | **55.0** | 60.0 | **71.9** | **61.6** |
119
 
120
+ ### 3.1.4. Comparison of LLM performances on Japanese testsets
121
  | Model |**Average**| JCQA | JNLI | MARC | JSQD | JQK | XLS | XWN | MGSM |
122
  |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------|
123
  | PLaMo-13B | 52.3 | 56.7 | 42.8 | 95.8 | 70.6 | 71.0 | 8.70 | 70.5 | 2.40 |
 
130
  | Yi-34B | 67.1 | 83.8 | 61.2 | 95.2 | **86.1** | 78.5 | **27.2** | 69.2 | 35.2 |
131
  | **Orion-14B-Base** | **69.1** | **88.2** | **75.8** | 94.1 | 75.7 | **85.1** | 17.3 | **78.8** | **38.0** |
132
 
133
+ ### 3.1.5. Comparison of LLM performances on Korean testsets. n = 0 and n = 5 stand for n-shot prompts used in the evaluation
134
  |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|
135
  |------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
136
  | 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 |
 
141
  | 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 |
142
  |**Orion-14B-Chat**|**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**|
143
 
144
+ ### 3.1.6. Multilingual evaluation
145
  | Model | Train Lang | Japanese | Korean | Chinese | English |
146
  |--------------------|------------|----------|----------|----------|----------|
147
  | PLaMo-13B | En,Jp | 52.3 | * | * | * |
 
157
  | **Orion-14B-Chat** | Multi | **69.1** | **79.5** | **67.9** | 67.3 |
158
 
159
 
160
+ ## 3.2. Chat Model Orion-14B-Chat Benchmarks
161
+ ### 3.2.1. Chat model subjective evaluation of MTBench
162
  | Model | First-Turn | Second-Turn | **Average** |
163
  |----------------------|----------|----------|----------|
164
  | Baichuan2-13B-Chat | 7.05 | 6.47 | 6.76 |
 
168
  | **Orion-14B-Chat** | **7.68** | **7.07** | **7.37** |
169
  \* use vllm for inference
170
 
171
+ ### 3.2.2. Chat model subjective evaluation of AlignBench
172
  | Model | Math. | Logi. | Basic. | Chi. | Comp. | Writ. | Role. | Prof. |**Avg.**|
173
  |--------------------|--------|--------|--------|--------|--------|--------|--------|--------|--------|
174
  | Baichuan2-13B-Chat | 3.76 | 4.07 | 6.22 | 6.05 | 7.11 | 6.97 | 6.75 | 6.43 | 5.25 |
 
179
 
180
  \* use vllm for inference
181
 
182
+ ## 3.3. LongChat Model Orion-14B-LongChat Benchmarks
183
+ ### 3.3.1. LongChat evaluation of LongBench
184
  | Model | NarrativeQA|MultiFieldQA-en|MultiFieldQA-zh| DuReader | QMSum | VCSUM | TREC | TriviaQA | LSHT |RepoBench-P|
185
  |--------------------------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|
186
  | GPT-3.5-Turbo-16k | **23.60** | **52.30** | **61.20** | 28.70 | 23.40 | **16.00** | 68.00 | **91.40** | 29.20 | 53.60 |
 
190
  | Orion-14B-LongChat | 19.47 | 48.11 | 55.84 | **37.02** | **24.87** | 15.44 | **77.00** | 89.12 | **45.50** | 54.31 |
191
 
192
 
193
+ ## 3.4. Chat RAG Model Benchmarks
194
+ ### 3.4.1. LLM evaluation results of self-built RAG testsets
195
  |Model|Effectiveness of Response(Keyword)|*Effectiveness of Response(subjective evaluation)|Quoting Ability|Fallback Ability|*AutoQA|*Data Extraction|
196
  |---------------------|------|------|------|------|------|------|
197
  | Baichuan2-13B-Chat | 85 | 76 | 1 | 0 | 69 | 51 |
 
201
  | Orion-14B-Chat-RAG | 86 | 87 | 91 | 97 | 73 | 71 |
202
  \* means manual assessment
203
 
204
+ ## 3.5. Chat Plugin Model Orion-14B-Chat-Plugin Benchmarks
205
+ ### 3.5.1. LLM evaluation results of self-built plugin testsets
206
  |Model |Intent Recognition with Full Params |Intent Recognition with Missing Params |Non-Plugin Invocation Recognition |
207
  |-----------------------|--------|-----------|--------|
208
  | Baichuan2-13B-Chat | 25 | 0 | 0 |
 
210
  | GPT-4 | **95** | 52.38 | 70 |
211
  | Orion-14B-Chat-Plugin | 92.5 | **60.32** | **90** |
212
 
213
+ ## 3.6. Quantized Model Orion-14B-Base-Int4 Benchmarks
214
+ ### 3.6.1. Comparison of before and after quantization
215
  |Model |Size(GB)|Inference Speed(tokens/s)|C-Eval|CMMLU|MMLU|RACE|HellaSwag|
216
  |-------------------------|-------|-----|------|------|------|------|------|
217
  | OrionStar-14B-Base | 28.0 | 135 | 72.8 | 70.6 | 70.0 | 93.3 | 78.5 |
218
  | OrionStar-14B-Base-Int4 | 8.3 | 178 | 71.8 | 69.8 | 69.2 | 93.1 | 78.0 |
219
 
220
+ # 4. Model Inference
221
 
222
  Model weights, source code, and configuration needed for inference are published on Hugging Face, and the download link
223
  is available in the table at the beginning of this document. We demonstrate various inference methods here, and the
224
  program will automatically download the necessary resources from Hugging Face.
225
 
226
+ ## 4.1. Python Code
227
 
228
  ```python
229
  import torch
 
244
  In the above Python code, the model is loaded with `device_map='auto'` to utilize all available GPUs. To specify the
245
  device, you can use something like `export CUDA_VISIBLE_DEVICES=0,1` (using GPUs 0 and 1).
246
 
247
+ ## 4.2. Command Line Tool
248
 
249
  ```shell
250
  CUDA_VISIBLE_DEVICES=0 python cli_demo.py
 
252
 
253
  This command-line tool is designed for chat scenarios, and thus, it does not support calling the base model.
254
 
255
+ ## 4.3. Direct Script Inference
256
 
257
  ```shell
258
 
 
264
 
265
  ```
266
 
267
+ ## 4.4 Example Output
268
 
269
+ ### 4.4.1. Casual Chat
270
 
271
  `````
272
  User: Hello
273
  Orion-14B: Hello! How can I help you today? If you have any questions or need assistance, feel free to ask.
274
  User: What functions do you have?
275
  Orion-14B: As an AI language model, I have a wide range of functions and capabilities, including but not limited to:
276
+ Answering general knowledge questions: I can provide information on various topics such as history, science, geography, and more.
277
+ Providing definitions and explanations: I can explain complex concepts and provide definitions for words and phrases.
278
+ Assisting with tasks: I can help with simple tasks like setting reminders, converting units of measurement, and even ordering food from restaurants.
279
+ Generating text: I can write short stories, essays, articles, and other types of content based on prompts given to me.
280
+ Translating languages: I can translate text from one language to another.
281
+ Chatting: I can engage in conversation with users and respond to their queries in a conversational manner.
282
+ Summarizing long texts: I can summarize longer texts like articles, books, and research papers.
283
+ Generating random numbers and names: I can generate random numbers within a specified range and also suggest names based on certain criteria.
284
+ Playing games: I can play simple games like Hangman, Trivia, and more.
285
+ Providing recommendations: I can recommend movies, music, books, and other forms of entertainment based on user preferences.
286
+ These are just some of the many functions that I am capable of performing. If you have a specific task in mind, feel free to ask me about it.
 
287
  User: Tell me a joke.
288
  Orion-14B: Sure, here's a classic one-liner: Why don't scientists trust atoms? Because they make up everything.
289
  `````
290
 
291
+ ### 4.4.2. Japanese & Korean Chat
292
 
293
  `````
294
  User:自己を紹介してください
 
297
  Orion-14B:안녕하세요, 제 이름은 ChatMax입니다. OrionStar에서 개발한 AI 어시스턴트입니다. 저는 사용자들에게 정확하고 객관적이며 유용한 답변을 제공하여 문제를 해결하고 정보를 얻는 데 도움을 줄 수 있습니다.
298
  `````
299
 
 
 
 
 
 
 
300
 
301
+ # 5. Declarations, License
 
 
 
 
302
 
303
+ ## 5.1. Declarations
 
 
304
 
305
  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.
306
  Additionally, we request users not to use the Orion-14B model for internet services without proper security review and filing.
 
311
  issues, public opinion risks, or any risks and issues arising from the model being misled, abused, disseminated, or
312
  improperly utilized, we will not assume any responsibility.
313
 
314
+ ## 5.2. License
315
 
316
  Community use of the Orion-14B series models
317
  - For code, please comply with [Apache License Version 2.0](./LICENSE)<br>
318
  - For model, please comply with [【Orion-14B Series】 Models Community License Agreement](./ModelsCommunityLicenseAgreement)
319
 
 
320
 
321
+ # 6. Company Introduction
322
 
323
+ OrionStar is a leading global service robot solutions company, founded in September 2016. OrionStar is dedicated to
324
+ using artificial intelligence technology to create the next generation of revolutionary robots, allowing people to break
325
+ free from repetitive physical labor and making human work and life more intelligent and enjoyable. Through technology,
326
+ OrionStar aims to make society and the world a better place.
327
 
328
+ OrionStar possesses fully self-developed end-to-end artificial intelligence technologies, such as voice interaction and
329
+ visual navigation. It integrates product development capabilities and technological application capabilities. Based on
330
+ the Orion robotic arm platform, it has launched products such as OrionStar AI Robot Greeting, AI Robot Greeting Mini,
331
+ Lucki, Coffee Master, and established the open platform OrionOS for Orion robots. Following the philosophy of "Born for
332
+ Truly Useful Robots", OrionStar empowers more people through AI technology.
333
+
334
+ **The core strengths of OrionStar lies in possessing end-to-end AI application capabilities,** including big data preprocessing, large model pretraining, fine-tuning, prompt engineering, agent, etc. With comprehensive end-to-end model training capabilities, including systematic data processing workflows and the parallel model training capability of hundreds of GPUs, it has been successfully applied in various industry scenarios such as government affairs, cloud services, international e-commerce, and fast-moving consumer goods.
335
+
336
+ Companies with demands for deploying large-scale model applications are welcome to contact us.
337
+ **Enquiry Hotline: 400-898-7779**<br>
338
+ **E-mail: ai@orionstar.com**
339
+
340
+ <div align="center">
341
+ <img src="./assets/imgs/wechat_group.jpg" alt="wechat" width="40%" />
342
+ </div>
README_cn.md CHANGED
@@ -11,7 +11,9 @@ pipeline_tag: text-generation
11
 
12
  <!-- markdownlint-disable first-line-h1 -->
13
  <!-- markdownlint-disable html -->
14
- ![](./assets/imgs/orion_start.PNG)
 
 
15
 
16
  <div align="center">
17
  <h1>
@@ -26,7 +28,7 @@ pipeline_tag: text-generation
26
  <p>
27
  <b>🇨🇳中文</b> |
28
  <a href="https://huggingface.co/OrionStarAI/Orion-14B-Chat-RAG/blob/main/README.md">🌐English</a><br><br>
29
- 🤗 <a href="https://huggingface.co/OrionStarAI" target="_blank">HuggingFace主页</a> | 🤖 <a href="https://modelscope.cn/organization/OrionStarAI" target="_blank">ModelScope主页</a><br>🎬 <a href="https://huggingface.co/spaces/OrionStarAI/Orion-14B-App-Demo" target="_blank">HuggingFace在线试用</a> | 🎫 <a href="https://modelscope.cn/studios/OrionStarAI/Orion-14B-App-Demo/summary" target="_blank">ModelScope在线试用</a>
30
  <p>
31
  </h4>
32
 
@@ -40,14 +42,14 @@ pipeline_tag: text-generation
40
  - [🔗 下载路径](#下载路径)
41
  - [🔖 评估结果](#评估结果)
42
  - [📊 模型推理](#模型推理)
43
- - [🥇 企业介绍](#企业介绍)
44
  - [📜 声明协议](#声明协议)
 
45
 
46
- # 模型介绍
47
 
48
  - **Orion-14B-Chat-RAG:** 在一个定制的检索增强生成数据集上进行微调的聊天模型,在检索增强生成任务中取得了卓越的性能。
49
 
50
- - Orion-14B-Base是一个具有140亿参数的多语种大模型,该模型在一个包含2.5万亿token的多样化数据集上进行了训练,涵盖了中文、英语、日语、韩语等多种语言。在多语言环境下的一系列任务中展现出卓越的性能。在主流的公开基准评测中,Orion-14B系列模型表现优异,多项指标显著超越同等参数基本的其他模型。
51
 
52
  - Orion-14B系列大模型有以下几个特点:
53
  - 基座20B参数级别大模型综合评测效果表现优异
@@ -55,7 +57,9 @@ pipeline_tag: text-generation
55
  - 微调模型适应性强,在人类标注盲测中,表现突出
56
  - 长上下文版本支持超长文本,长达200k token
57
  - 量化版本模型大小缩小70%,推理速度提升30%,性能损失小于1%
58
- ![](./assets/imgs/model_cap_zh.png)
 
 
59
 
60
  - 具体而言,Orion-14B系列大语言模型包含:
61
  - **Orion-14B-Base:** 基于2.5万亿令牌多样化数据集训练处的140亿参数量级的多语言基座模型。
@@ -66,7 +70,7 @@ pipeline_tag: text-generation
66
  - **Orion-14B-Base-Int4:** 一个使用4位整数进行量化的基座模型。它将模型大小显著减小了70%,同时提高了推理速度30%,仅引入了1%的最小性能损失。
67
  - **Orion-14B-Chat-Int4:** 一个使用4位整数进行量化的对话模型。
68
 
69
- # 下载路径
70
 
71
  发布模型和下载链接见下表:
72
 
@@ -80,10 +84,11 @@ pipeline_tag: text-generation
80
  | 💼 基座Int4量化模型 | [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) |
81
  | 📦 对话Int4量化模型 | [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) |
82
 
83
- # 评估结果
84
 
85
- ## 1. 基座模型评估
86
- ### 专业知识与试题评估结果
 
87
  | 模型名称 | C-Eval | CMMLU | MMLU | AGIEval | Gaokao | BBH |
88
  |--------------------|----------|----------|----------|----------|----------|----------|
89
  | LLaMA2-13B | 41.4 | 38.4 | 55.0 | 30.9 | 18.2 | 45.6 |
@@ -93,7 +98,7 @@ pipeline_tag: text-generation
93
  | InternLM-20B | 58.8 | 59.0 | 62.1 | 44.6 | 45.5 | 52.5 |
94
  | **Orion-14B-Base** | **72.9** | **70.6** | **69.9** | **54.7** | 62.1 | **56.5** |
95
 
96
- ### 理解与通识评估结果
97
  | 模型名称 |RACE-middle|RACE-high| HellaSwag| PIQA | Lambada | WSC |
98
  |--------------------|----------|----------|----------|----------|----------|----------|
99
  | LLaMA 2-13B | 63.0 | 58.9 | 77.5 | 79.8 | 76.5 | 66.3 |
@@ -103,7 +108,7 @@ pipeline_tag: text-generation
103
  | InternLM-20B | 86.4 | 83.3 | 78.1 | **80.3** | 71.8 | 68.3 |
104
  | **Orion-14B-Base** | **93.3** | **91.3** | 78.5 | 79.5 | **78.9** | **70.2** |
105
 
106
- ### OpenCompass评测集评估结果
107
  | 模型名称 | Average | Examination | Language | Knowledge | Understanding | Reasoning |
108
  |------------------|----------|----------|----------|----------|----------|----------|
109
  | LLaMA 2-13B | 47.3 | 45.2 | 47.0 | 58.3 | 50.9 | 43.6 |
@@ -113,7 +118,7 @@ pipeline_tag: text-generation
113
  | InternLM-20B | 59.4 | 62.5 | 55.0 | **60.1** | 67.3 | 54.9 |
114
  |**Orion-14B-Base**| **64.4** | **71.4** | **55.0** | 60.0 | **71.9** | **61.6** |
115
 
116
- ### 日语测试集评估结果
117
  | 模型名称 |**Average**| JCQA | JNLI | MARC | JSQD | JQK | XLS | XWN | MGSM |
118
  |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------|
119
  | PLaMo-13B | 52.3 | 56.7 | 42.8 | 95.8 | 70.6 | 71.0 | 8.70 | 70.5 | 2.40 |
@@ -126,7 +131,7 @@ pipeline_tag: text-generation
126
  | Yi-34B | 67.1 | 83.8 | 61.2 | 95.2 | **86.1** | 78.5 | **27.2** | 69.2 | 35.2 |
127
  | **Orion-14B-Base** | **69.1** | **88.2** | **75.8** | 94.1 | 75.7 | **85.1** | 17.3 | **78.8** | **38.0** |
128
 
129
- ### 韩语测试集n-shot评估结果
130
  | 模型名称 | **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|
131
  |------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
132
  | 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 |
@@ -137,7 +142,7 @@ pipeline_tag: text-generation
137
  | 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 |
138
  |**Orion-14B-Base**|**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**|
139
 
140
- ### 多语言评估结果
141
  | 模型名称 | Train Lang | Japanese | Korean | Chinese | English |
142
  |--------------------|------------|----------|----------|----------|----------|
143
  | PLaMo-13B | En,Jp | 52.3 | * | * | * |
@@ -152,8 +157,8 @@ pipeline_tag: text-generation
152
  | Yi-34B | Multi | 67.1 | 72.2 | 58.7 | **68.8** |
153
  | **Orion-14B-Base** | Multi | **69.1** | **79.5** | **67.9** | 67.3 |
154
 
155
- ## 2. 对话模型评估
156
- ### 对话模型MTBench主观评估
157
  | 模型名称 | 第一轮 | 第二轮 | **平均** |
158
  |----------------------|----------|----------|----------|
159
  | Baichuan2-13B-Chat | 7.05 | 6.47 | 6.76 |
@@ -163,7 +168,7 @@ pipeline_tag: text-generation
163
  | **Orion-14B-Chat** | **7.68** | **7.07** | **7.37** |
164
  \*这里评测使用vllm进行推理
165
 
166
- ### 对话模型AlignBench主观评估
167
  | 模型名称 | 数学能力 | 逻辑推理 | 基本能力 | 中文理解 | 综合问答 | 写作能力 | 角色扮演 | 专业知识 | **平均** |
168
  |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------|
169
  | Baichuan2-13B-Chat | 3.76 | 4.07 | 6.22 | 6.05 | 7.11 | 6.97 | 6.75 | 6.43 | 5.25 |
@@ -173,8 +178,8 @@ pipeline_tag: text-generation
173
  | **Orion-14B-Chat** | 4.00 | 4.24 | 6.18 | **6.57** | 7.16 | **7.36** | **7.16** | **6.99** | 5.51 |
174
  \*这里评测使用vllm进行推理
175
 
176
- ## 3. 长上下文模型评估
177
- ### 长上下文模型LongBench评估
178
  | 模型名称 | NarrativeQA| MultiFieldQA-en| MultiFieldQA-zh | DuReader | QMSum | VCSUM | TREC | TriviaQA | LSHT | RepoBench-P |
179
  |--------------------------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|
180
  | GPT-3.5-Turbo-16k | **23.60** | **52.30** | **61.20** | 28.70 | 23.40 | **16.00** | 68.00 | **91.40** | 29.20 | 53.60 |
@@ -183,8 +188,8 @@ pipeline_tag: text-generation
183
  | Yi-6B-200K | 14.11 | 36.74 | 22.68 | 14.01 | 20.44 | 8.08 | 72.00 | 86.61 | 38.00 | **63.29** |
184
  | Orion-14B-LongChat | 19.47 | 48.11 | 55.84 | **37.02** | **24.87** | 15.44 | **77.00** | 89.12 | **45.50** | 54.31 |
185
 
186
- ## 4. 检索增强模型评估
187
- ### 自建检索增强测试集评估结果
188
  |模型名称|回复效果(关键字)|*回复效果(主观打分)|引用能力|兜底能力|*AutoQA|*抽取数据|
189
  |---------------------|------|------|------|------|------|------|
190
  | Baichuan2-13B-Chat | 85 | 76 | 1 | 0 | 69 | 51 |
@@ -194,8 +199,8 @@ pipeline_tag: text-generation
194
  | Orion-14B-Chat-RAG | 86 | 87 | 91 | 97 | 73 | 71 |
195
  \* 表示人工评判结果
196
 
197
- ## 5. 插件模型评估
198
- ### 自建插件测试集评估结果
199
  | 模型名称 | 全参数意图识别 | 缺参数意图识别 | 非插件调用识别 |
200
  |-----------------------|--------|-----------|--------|
201
  | Baichuan2-13B-Chat | 25 | 0 | 0 |
@@ -203,20 +208,20 @@ pipeline_tag: text-generation
203
  | GPT-4 | **95** | 52.38 | 70 |
204
  | Orion-14B-Chat-Plugin | 92.5 | **60.32** | **90** |
205
 
206
- ## 6. 量化模型评估
207
- ### 量化前后整体对比
208
  |模型名称|模型大小(GB)|推理速度(令牌数/秒)|C-Eval |CMMLU |MMLU |RACE | HellaSwag|
209
  |-------------------------|------|-----|------|------|------|------|------|
210
  | OrionStar-14B-Base | 28.0 | 135 | 72.8 | 70.6 | 70.0 | 93.3 | 78.5 |
211
  | OrionStar-14B-Base-Int4 | 8.3 | 178 | 71.8 | 69.8 | 69.2 | 93.1 | 78.0 |
212
 
213
 
214
- # 模型推理
215
 
216
  推理所需的模型权重、源码、配置已发布在 Hugging Face,下载链接见本文档最开始的表格。我们在此示范多种推理方式。程序会自动从
217
  Hugging Face 下载所需资源。
218
 
219
- ## Python 代码方式
220
 
221
  ```python
222
  import torch
@@ -237,7 +242,7 @@ print(response)
237
  在上述两段代码中,模型加载指定 `device_map='auto'`
238
  ,会使用所有可用显卡。如需指定使用的设备,可以使用类似 `export CUDA_VISIBLE_DEVICES=0,1`(使用了0、1号显卡)的方式控制。
239
 
240
- ## 命令行工具方式
241
 
242
  ```shell
243
  CUDA_VISIBLE_DEVICES=0 python cli_demo.py
@@ -245,7 +250,7 @@ CUDA_VISIBLE_DEVICES=0 python cli_demo.py
245
 
246
  本命令行工具是为 Chat 场景设计,因此我们不支持使用该工具调用 Base 模型。
247
 
248
- ## 脚本直接推理
249
 
250
  ```shell
251
  # base model
@@ -256,9 +261,9 @@ CUDA_VISIBLE_DEVICES=0 python demo/text_generation.py --model OrionStarAI/Orion-
256
 
257
  ```
258
 
259
- ## 示例输出
260
 
261
- ## 闲聊
262
 
263
  `````
264
  用户:你好,你叫什么名字
@@ -266,26 +271,26 @@ Orion-14B:你好!我是一个人工智能助手,没有固定的名字。你可
266
  用户:讲个故事
267
  Orion-14B:从前有一个小男孩叫杰克,他住在一个小村庄里。杰克非常喜欢冒险和探索,总是梦想着能够发现一些神奇的事情。
268
 
269
- 有一天,杰克听说在附近的森林深处有一座传说中的魔法城堡,里面住着一个邪恶的巨人。这个巨人的名字叫格里姆,他被困在城堡中,无法离开。
270
 
271
- 杰克决定要去解救这个巨人,因为他相信格里姆一定有自己的原因被困在那里。于是,他带着勇气和决心,踏上了前往魔法城堡的旅程。
272
 
273
- 经过长时间的跋涉,杰克终于来到了城堡的大门前。他小心翼翼地推开大门,进入了一个充满神秘气息的房间。
274
 
275
- 房间里弥漫着黑暗的气息,但杰克并没有被吓倒。他勇敢地向城堡深处走去,最终找到了格里姆。
276
 
277
- 格里姆告诉杰克,他是被一个邪恶的女巫诅咒的,只有通过解开咒语才能获得自由。杰克决定要帮助格里姆,他开始寻找线索和解开咒语的方法。
278
 
279
- 经过一番努力,杰克终于找到了一条通往女巫的秘密实验室的道路。他在那里发现了女巫留下的魔法书,书中记载了解开咒语的方法。
280
 
281
- 杰克按照书上所写的步骤进行操作,成功地解开了格里姆身上的诅咒。巨人感激不已,他答应帮助杰克实现他的愿望。
282
 
283
- 从此以后,杰克和格里姆成为了好朋友,他们一起经历了许多惊险刺激的冒险,并用他们的勇气和智慧战胜了各种困难。
284
 
285
- 这个故事告诉我们,只要有勇气和决心,我们就能克服一切困难,实现自己的梦想。
286
  `````
287
 
288
- ## 日韩
289
 
290
  `````
291
  用户:自己を紹介してください
@@ -294,24 +299,10 @@ Orion-14B:こんにちは、私の名前はChatMaxで、OrionStarによって
294
  Orion-14B:안녕하세요, 제 이름은 ChatMax입니다. OrionStar에서 개발한 AI 어시스턴트입니다. 저는 사용자들에게 정확하고 객관적이며 유용한 답변을 제공하여 문제를 해결하고 정보를 얻는 데 도움을 줄 수 있습니다.
295
  `````
296
 
297
- # 企业介绍
298
-
299
- 猎户星空(OrionStar)是一家全球领先的服务机器人解决方案公司,成立于2016年9月。猎户星空致力于基于人工智能技术打造下一代革命性机器人,使人们能够摆脱重复的体力劳动,使人类的工作和生活更加智能和有趣,通过技术使社会和世界变得更加美好。
300
-
301
- 猎户星空拥有完全自主开发的全链条人工智能技术,如语音交互和视觉导航。它整合了产品开发能力和技术应用能力。基于Orion机械臂平台,它推出了ORION
302
- STAR AI Robot Greeting、AI Robot Greeting Mini、Lucki、Coffee
303
- Master等产品,并建立了Orion机器人的开放平台OrionOS。通过为 **真正有用的机器人而生** 的理念实践,它通过AI技术为更多人赋能。
304
-
305
- 凭借7年AI经验积累,猎户星空已推出的大模型深度应用“聚言”,并陆续面向行业客户提供定制化AI大模型咨询与服务解决方案,真正帮助客户实现企业经营效率领先同行目标。
306
-
307
- **猎户星空具备全链条大模型应用能力的核心优势**,包括拥有从海量数据处理、大模型预训练、二次预训练、微调(Fine-tune)、Prompt
308
- Engineering 、Agent开发的全链条能力和经验积累;拥有完整的端到端模型训练能力,包括系统化的数据处理流程和数百张GPU的并行模型训练能力,现已在大政务、云服务、出海电商、快消等多个行业场景落地。
309
 
310
- ***欢迎有大模型应用落地需求的企业联系我们进行商务合作,咨询电话 400-898-7779 。***
311
 
312
- # 声明、协议
313
-
314
- ## 声明
315
 
316
  我们强烈呼吁所有使用者,不要利用 Orion-14B 模型进行任何危害国家社会安全或违法的活动。另外,我们也要求使用者不要将
317
  Orion-14B 模型用于未经适当安全审查和备案的互联网服务。
@@ -320,14 +311,29 @@ Orion-14B 模型用于未经适当安全审查和备案的互联网服务。
320
  我们已经尽我们所能,来确保模型训练过程中使用的数据的合规性。然而,尽管我们已经做出了巨大的努力,但由于模型和数据的复杂性,仍有可能存在一些无法预见的问题。因此,如果由于使用
321
  Orion-14B 开源模型而导致的任何问题,包括但不限于数据安全问题、公共舆论风险,或模型被误导、滥用、传播或不当利用所带来的任何风险和问题,我们将不承担任何责任。
322
 
323
- ## 协议
324
 
325
  社区使用Orion-14B系列模型
326
  - 代码请遵循 [Apache License Version 2.0](./LICENSE)<br>
327
  - 模型请遵循 [Orion-14B系列模型社区许可协议](./ModelsCommunityLicenseAgreement)
328
 
329
- # 联系我们
 
 
 
 
 
 
 
 
 
 
 
330
 
331
- 电子邮箱: ai@orionstar.com
 
 
332
 
333
- ![](./assets/imgs/wechat_group.jpg)
 
 
 
11
 
12
  <!-- markdownlint-disable first-line-h1 -->
13
  <!-- markdownlint-disable html -->
14
+ <div align="center">
15
+ <img src="./assets/imgs/orion_start.PNG" alt="logo" width="50%" />
16
+ </div>
17
 
18
  <div align="center">
19
  <h1>
 
28
  <p>
29
  <b>🇨🇳中文</b> |
30
  <a href="https://huggingface.co/OrionStarAI/Orion-14B-Chat-RAG/blob/main/README.md">🌐English</a><br><br>
31
+ 🤗 <a href="https://huggingface.co/OrionStarAI" target="_blank">HuggingFace主页</a> | 🤖 <a href="https://modelscope.cn/organization/OrionStarAI" target="_blank">ModelScope主页</a><br>🎬 <a href="https://huggingface.co/spaces/OrionStarAI/Orion-14B-App-Demo" target="_blank">HuggingFace在线试用</a> | 🎫 <a href="https://modelscope.cn/studios/OrionStarAI/Orion-14B-App-Demo/summary" target="_blank">ModelScope在线试用</a><br>📖 <a href="https://github.com/OrionStarAI/Orion/blob/master/doc/Orion14B_v3.pdf" target="_blank">技术报告</a>
32
  <p>
33
  </h4>
34
 
 
42
  - [🔗 下载路径](#下载路径)
43
  - [🔖 评估结果](#评估结果)
44
  - [📊 模型推理](#模型推理)
 
45
  - [📜 声明协议](#声明协议)
46
+ - [🥇 企业介绍](#企业介绍)
47
 
48
+ # 1. 模型介绍
49
 
50
  - **Orion-14B-Chat-RAG:** 在一个定制的检索增强生成数据集上进行微调的聊天模型,在检索增强生成任务中取得了卓越的性能。
51
 
52
+ - Orion-14B-Base是一个具有140亿参数的多语种大模型,该模型在一个包含2.5万亿token的多样化数据集上进行了训练,涵盖了中文、英语、日语、韩语等多种语言。在多语言环境下的一系列任务中展现出卓越的性能。在主流的公开基准评测中,Orion-14B系列模型表现优异,多项指标显著超越同等参数基本的其他模型。具体技术细节请参考[技术报告](https://github.com/OrionStarAI/Orion/blob/master/doc/Orion14B_v3.pdf)。
53
 
54
  - Orion-14B系列大模型有以下几个特点:
55
  - 基座20B参数级别大模型综合评测效果表现优异
 
57
  - 微调模型适应性强,在人类标注盲测中,表现突出
58
  - 长上下文版本支持超长文本,长达200k token
59
  - 量化版本模型大小缩小70%,推理速度提升30%,性能损失小于1%
60
+ <div align="center">
61
+ <img src="./assets/imgs/model_cap_zh.png" alt="model_cap" width="50%" />
62
+ </div>
63
 
64
  - 具体而言,Orion-14B系列大语言模型包含:
65
  - **Orion-14B-Base:** 基于2.5万亿令牌多样化数据集训练处的140亿参数量级的多语言基座模型。
 
70
  - **Orion-14B-Base-Int4:** 一个使用4位整数进行量化的基座模型。它将模型大小显著减小了70%,同时提高了推理速度30%,仅引入了1%的最小性能损失。
71
  - **Orion-14B-Chat-Int4:** 一个使用4位整数进行量化的对话模型。
72
 
73
+ # 2. 下载路径
74
 
75
  发布模型和下载链接见下表:
76
 
 
84
  | 💼 基座Int4量化模型 | [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) |
85
  | 📦 对话Int4量化模型 | [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) |
86
 
87
+ # 3. 评估结果
88
 
89
+ ## 3.1. 基座模型Orion-14B-Base评估
90
+
91
+ ### 3.1.1. 专业知识与试题评估结果
92
  | 模型名称 | C-Eval | CMMLU | MMLU | AGIEval | Gaokao | BBH |
93
  |--------------------|----------|----------|----------|----------|----------|----------|
94
  | LLaMA2-13B | 41.4 | 38.4 | 55.0 | 30.9 | 18.2 | 45.6 |
 
98
  | InternLM-20B | 58.8 | 59.0 | 62.1 | 44.6 | 45.5 | 52.5 |
99
  | **Orion-14B-Base** | **72.9** | **70.6** | **69.9** | **54.7** | 62.1 | **56.5** |
100
 
101
+ ### 3.1.2. 理解与通识评估结果
102
  | 模型名称 |RACE-middle|RACE-high| HellaSwag| PIQA | Lambada | WSC |
103
  |--------------------|----------|----------|----------|----------|----------|----------|
104
  | LLaMA 2-13B | 63.0 | 58.9 | 77.5 | 79.8 | 76.5 | 66.3 |
 
108
  | InternLM-20B | 86.4 | 83.3 | 78.1 | **80.3** | 71.8 | 68.3 |
109
  | **Orion-14B-Base** | **93.3** | **91.3** | 78.5 | 79.5 | **78.9** | **70.2** |
110
 
111
+ ### 3.1.3. OpenCompass评测集评估结果
112
  | 模型名称 | Average | Examination | Language | Knowledge | Understanding | Reasoning |
113
  |------------------|----------|----------|----------|----------|----------|----------|
114
  | LLaMA 2-13B | 47.3 | 45.2 | 47.0 | 58.3 | 50.9 | 43.6 |
 
118
  | InternLM-20B | 59.4 | 62.5 | 55.0 | **60.1** | 67.3 | 54.9 |
119
  |**Orion-14B-Base**| **64.4** | **71.4** | **55.0** | 60.0 | **71.9** | **61.6** |
120
 
121
+ ### 3.1.4. 日语测试集评估结果
122
  | 模型名称 |**Average**| JCQA | JNLI | MARC | JSQD | JQK | XLS | XWN | MGSM |
123
  |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------|
124
  | PLaMo-13B | 52.3 | 56.7 | 42.8 | 95.8 | 70.6 | 71.0 | 8.70 | 70.5 | 2.40 |
 
131
  | Yi-34B | 67.1 | 83.8 | 61.2 | 95.2 | **86.1** | 78.5 | **27.2** | 69.2 | 35.2 |
132
  | **Orion-14B-Base** | **69.1** | **88.2** | **75.8** | 94.1 | 75.7 | **85.1** | 17.3 | **78.8** | **38.0** |
133
 
134
+ ### 3.1.5. 韩语测试集n-shot评估结果
135
  | 模型名称 | **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|
136
  |------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
137
  | 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 |
 
142
  | 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 |
143
  |**Orion-14B-Base**|**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**|
144
 
145
+ ### 3.1.6. 多语言评估结果
146
  | 模型名称 | Train Lang | Japanese | Korean | Chinese | English |
147
  |--------------------|------------|----------|----------|----------|----------|
148
  | PLaMo-13B | En,Jp | 52.3 | * | * | * |
 
157
  | Yi-34B | Multi | 67.1 | 72.2 | 58.7 | **68.8** |
158
  | **Orion-14B-Base** | Multi | **69.1** | **79.5** | **67.9** | 67.3 |
159
 
160
+ ## 3.2. 对话模型Orion-14B-Chat评估
161
+ ### 3.2.1. 对话模型MTBench主观评估
162
  | 模型名称 | 第一轮 | 第二轮 | **平均** |
163
  |----------------------|----------|----------|----------|
164
  | Baichuan2-13B-Chat | 7.05 | 6.47 | 6.76 |
 
168
  | **Orion-14B-Chat** | **7.68** | **7.07** | **7.37** |
169
  \*这里评测使用vllm进行推理
170
 
171
+ ### 3.2.2. 对话模型AlignBench主观评估
172
  | 模型名称 | 数学能力 | 逻辑推理 | 基本能力 | 中文理解 | 综合问答 | 写作能力 | 角色扮演 | 专业知识 | **平均** |
173
  |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------|
174
  | Baichuan2-13B-Chat | 3.76 | 4.07 | 6.22 | 6.05 | 7.11 | 6.97 | 6.75 | 6.43 | 5.25 |
 
178
  | **Orion-14B-Chat** | 4.00 | 4.24 | 6.18 | **6.57** | 7.16 | **7.36** | **7.16** | **6.99** | 5.51 |
179
  \*这里评测使用vllm进行推理
180
 
181
+ ## 3.3. 长上下文模型Orion-14B-LongChat评估
182
+ ### 3.3.1. 长上下文模型LongBench评估
183
  | 模型名称 | NarrativeQA| MultiFieldQA-en| MultiFieldQA-zh | DuReader | QMSum | VCSUM | TREC | TriviaQA | LSHT | RepoBench-P |
184
  |--------------------------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|
185
  | GPT-3.5-Turbo-16k | **23.60** | **52.30** | **61.20** | 28.70 | 23.40 | **16.00** | 68.00 | **91.40** | 29.20 | 53.60 |
 
188
  | Yi-6B-200K | 14.11 | 36.74 | 22.68 | 14.01 | 20.44 | 8.08 | 72.00 | 86.61 | 38.00 | **63.29** |
189
  | Orion-14B-LongChat | 19.47 | 48.11 | 55.84 | **37.02** | **24.87** | 15.44 | **77.00** | 89.12 | **45.50** | 54.31 |
190
 
191
+ ## 3.4. 检索增强模型Orion-14B-Chat-RAG评估
192
+ ### 3.4.1. 自建检索增强测试集评估结果
193
  |模型名称|回复效果(关键字)|*回复效果(主观打分)|引用能力|兜底能力|*AutoQA|*抽取数据|
194
  |---------------------|------|------|------|------|------|------|
195
  | Baichuan2-13B-Chat | 85 | 76 | 1 | 0 | 69 | 51 |
 
199
  | Orion-14B-Chat-RAG | 86 | 87 | 91 | 97 | 73 | 71 |
200
  \* 表示人工评判结果
201
 
202
+ ## 3.5. 插件模型Orion-14B-Chat-Plugin评估
203
+ ### 3.5.1. 自建插件测试集评估结果
204
  | 模型名称 | 全参数意图识别 | 缺参数意图识别 | 非插件调用识别 |
205
  |-----------------------|--------|-----------|--------|
206
  | Baichuan2-13B-Chat | 25 | 0 | 0 |
 
208
  | GPT-4 | **95** | 52.38 | 70 |
209
  | Orion-14B-Chat-Plugin | 92.5 | **60.32** | **90** |
210
 
211
+ ## 3.6. 量化模型Orion-14B-Base-Int4评估
212
+ ### 3.6.1. 量化前后整体对比
213
  |模型名称|模型大小(GB)|推理速度(令牌数/秒)|C-Eval |CMMLU |MMLU |RACE | HellaSwag|
214
  |-------------------------|------|-----|------|------|------|------|------|
215
  | OrionStar-14B-Base | 28.0 | 135 | 72.8 | 70.6 | 70.0 | 93.3 | 78.5 |
216
  | OrionStar-14B-Base-Int4 | 8.3 | 178 | 71.8 | 69.8 | 69.2 | 93.1 | 78.0 |
217
 
218
 
219
+ # 4. 模型推理
220
 
221
  推理所需的模型权重、源码、配置已发布在 Hugging Face,下载链接见本文档最开始的表格。我们在此示范多种推理方式。程序会自动从
222
  Hugging Face 下载所需资源。
223
 
224
+ ## 4.1. Python 代码方式
225
 
226
  ```python
227
  import torch
 
242
  在上述两段代码中,模型加载指定 `device_map='auto'`
243
  ,会使用所有可用显卡。如需指定使用的设备,可以使用类似 `export CUDA_VISIBLE_DEVICES=0,1`(使用了0、1号显卡)的方式控制。
244
 
245
+ ## 4.2. 命令行工具方式
246
 
247
  ```shell
248
  CUDA_VISIBLE_DEVICES=0 python cli_demo.py
 
250
 
251
  本命令行工具是为 Chat 场景设计,因此我们不支持使用该工具调用 Base 模型。
252
 
253
+ ## 4.3. 脚本直接推理
254
 
255
  ```shell
256
  # base model
 
261
 
262
  ```
263
 
264
+ ## 4.4. 示例输出
265
 
266
+ ### 4.4.1. 闲聊
267
 
268
  `````
269
  用户:你好,你叫什么名字
 
271
  用户:讲个故事
272
  Orion-14B:从前有一个小男孩叫杰克,他住在一个小村庄里。杰克非常喜欢冒险和探索,总是梦想着能够发现一些神奇的事情。
273
 
274
+ 有一天,杰克听说在附近的森林深处有一座传说中的魔法城堡,里面住着一个邪恶的巨人。这个巨人的名字叫格里姆,他被困在城堡中,无法离开。
275
 
276
+ 杰克决定要去解救这个巨人,因为他相信格里姆一定有自己的原因被困在那里。于是,他带着勇气和决心,踏上了前往魔法城堡的旅程。
277
 
278
+ 经过长时间的跋涉,杰克终于来到了城堡的大门前。他小心翼翼地推开大门,进入了一个充满神秘气息的房间。
279
 
280
+ 房间里弥漫着黑暗的气息,但杰克并没有被吓倒。他勇敢地向城堡深处走去,最终找到了格里姆。
281
 
282
+ 格里姆告诉杰克,他是被一个邪恶的女巫诅咒的,只有通过解开咒语才能获得自由。杰克决定要帮助格里姆,他开始寻找线索和解开咒语的方法。
283
 
284
+ 经过一番努力,杰克终于找到了一条通往女巫的秘密实验室的道路。他在那里发现了女巫留下的魔法书,书中记载了解开咒语的方法。
285
 
286
+ 杰克按照书上所写的步骤进行操作,成功地解开了格里姆身上的诅咒。巨人感激不已,他答应帮助杰克实现他的愿望。
287
 
288
+ 从此以后,杰克和格里姆成为了好朋友,他们一起经历了许多惊险刺激的冒险,并用他们的勇气和智慧战胜了各种困难。
289
 
290
+ 这个故事告诉我们,只要有勇气和决心,我们就能克服一切困难,实现自己的梦想。
291
  `````
292
 
293
+ ### 4.4.2. 日韩
294
 
295
  `````
296
  用户:自己を紹介してください
 
299
  Orion-14B:안녕하세요, 제 이름은 ChatMax입니다. OrionStar에서 개발한 AI 어시스턴트입니다. 저는 사용자들에게 정확하고 객관적이며 유용한 답변을 제공하여 문제를 해결하고 정보를 얻는 데 도움을 줄 수 있습니다.
300
  `````
301
 
 
 
 
 
 
 
 
 
 
 
 
 
302
 
303
+ # 5. 声明、协议
304
 
305
+ ## 5.1. 声明
 
 
306
 
307
  我们强烈呼吁所有使用者,不要利用 Orion-14B 模型进行任何危害国家社会安全或违法的活动。另外,我们也要求使用者不要将
308
  Orion-14B 模型用于未经适当安全审查和备案的互联网服务。
 
311
  我们已经尽我们所能,来确保模型训练过程中使用的数据的合规性。然而,尽管我们已经做出了巨大的努力,但由于模型和数据的复杂性,仍有可能存在一些无法预见的问题。因此,如果由于使用
312
  Orion-14B 开源模型而导致的任何问题,包括但不限于数据安全问题、公共舆论风险,或模型被误导、滥用、传播或不当利用所带来的任何风险和问题,我们将不承担任何责任。
313
 
314
+ ## 5.2. 协议
315
 
316
  社区使用Orion-14B系列模型
317
  - 代码请遵循 [Apache License Version 2.0](./LICENSE)<br>
318
  - 模型请遵循 [Orion-14B系列模型社区许可协议](./ModelsCommunityLicenseAgreement)
319
 
320
+ # 6. 企业介绍
321
+
322
+ 猎户星空(OrionStar)是一家全球领先的服务机器人解决方案公司,成立于2016年9月。猎户星空致力于基于人工智能技术打造下一代革命性机器人,使人们能够摆脱重复的体力劳动,使人类的工作和生活更加智能和有趣,通过技术使社会和世界变得更加美好。
323
+
324
+ 猎户星空拥有完全自主开发的全链条人工智能技术,如语音交互和视觉导航。它整合了产品开发能力和技术应用能力。基于Orion机械臂平台,它推出了ORION
325
+ STAR AI Robot Greeting、AI Robot Greeting Mini、Lucki、Coffee
326
+ Master等产品,并建立了Orion机器人的开放平台OrionOS。通过为 **真正有用的机器人而生** 的理念实践,它通过AI技术为更多人赋能。
327
+
328
+ 凭借7年AI经验积累,猎户星空已推出的大模型深度应用“聚言”,并陆续面向行业客户提供定制化AI大模型咨询与服务解决方案,真正帮助客户实现企业经营效率领先同行目标。
329
+
330
+ **猎户星空具备全链条大模型应用能力的核心优势**,包括拥有从海量数据处理、大模型预训练、二次预训练、微调(Fine-tune)、Prompt
331
+ Engineering 、Agent开发的全链条能力和经验积累;拥有完整的端到端模型训练能力,包括系统化的数据处理流程和数百张GPU的并行模型训练能力,现已在大政务、云服务、出海电商、快消等多个行业场景落地。
332
 
333
+ ***欢迎有大模型应用落地需求的企业联系我们进行商务合作***<br>
334
+ **咨询电话:** 400-898-7779<br>
335
+ **电子邮箱:** ai@orionstar.com
336
 
337
+ <div align="center">
338
+ <img src="./assets/imgs/wechat_group.jpg" alt="wechat" width="40%" />
339
+ </div>