yi-01-ai commited on
Commit
de636e3
1 Parent(s): add4ead

Auto Sync from git://github.com/01-ai/Yi.git/commit/9bc9255729d150fd2496c1f4f65e7cd486c6c8bf

Browse files
Files changed (1) hide show
  1. README.md +14 -18
README.md CHANGED
@@ -150,7 +150,7 @@ pipeline_tag: text-generation
150
 
151
  Yi models come in multiple sizes and cater to different use cases. You can also fine-tune Yi models to meet your specific requirements.
152
 
153
- If you want to deploy Yi models, see [software and hardware requirements](#deployment)
154
 
155
  ### Chat models
156
 
@@ -331,7 +331,7 @@ This tutorial guides you through every step of running **Yi-34B-Chat locally on
331
 
332
  #### Step 0: Prerequistes
333
 
334
- - Make sure Python 3.10 or later version is installed.
335
 
336
  - If you want to run other Yi models, see [software and hardware requirements](#deployment)
337
 
@@ -833,8 +833,8 @@ python eval_quantized_model.py --model /quantized_model --trust_remote_code
833
  <div align="right"> [ <a href="#building-the-next-generation-of-open-source-and-bilingual-llms">Back to top ⬆️ </a> ] </div>
834
 
835
  ### Deployment
836
- <details>
837
- <summary> Software and hardware requirements of deploying Yi models ⬇️</summary>
838
 
839
  #### Software requirements
840
 
@@ -845,7 +845,6 @@ Before using Yi quantized models, make sure you've installed the correct softwar
845
  Yi 4-bit quantized models | [AWQ and CUDA](https://github.com/casper-hansen/AutoAWQ?tab=readme-ov-file#install-from-pypi)
846
  Yi 8-bit quantized models | [GPTQ and CUDA](https://github.com/PanQiWei/AutoGPTQ?tab=readme-ov-file#quick-installation)
847
 
848
-
849
  #### Hardware requirements
850
 
851
  Before deploying Yi in your environment, make sure your hardware meets the following requirements.
@@ -881,12 +880,12 @@ Below are detailed minimum VRAM requirements under different batch use cases.
881
  | Yi-34B | 72 GB | 4 x RTX 4090 <br> A800 (80 GB) |
882
  | Yi-34B-200K | 200 GB | 4 x A800 (80 GB) |
883
 
884
- </details>
885
-
886
  ### Learning hub
 
887
  <details>
888
- <summary> Learning materials of Yi ⬇️</summary>
889
  <br>
 
890
  Welcome to the Yi learning hub!
891
 
892
  Whether you're a seasoned developer or a newcomer, you can find a wealth of helpful educational resources to enhance your understanding and skills with Yi models, including insightful blog posts, comprehensive video tutorials, hands-on guides, and more.
@@ -897,7 +896,7 @@ At the same time, we also warmly invite you to join our collaborative effort by
897
 
898
  With all these resources at your fingertips, you're ready to start your exciting journey with Yi. Happy learning! 🥳
899
 
900
- ##### Tutorials
901
 
902
  | Type | Deliverable | Date | Author |
903
  |-------------|--------------------------------------------------------|----------------|----------------|
@@ -1008,14 +1007,13 @@ If you're seeking to explore the diverse capabilities within Yi's thriving famil
1008
  - [📊 Base model performance](#-base-model-performance)
1009
 
1010
  ### 📊 Chat model performance
1011
- 🎯 Performance evaluation
1012
- - Yi-34B-chat stands out, doing better than most big models in almost all tests.
1013
- - Both Yi-34B-chat and its variant, Yi-34B-Chat-8bits (GPTQ), take the top spots in tests including MMLU, CMMLU, BBH, and GSM8k.
1014
 
1015
  ![Chat model performance](./assets/img/benchmark_chat.png)
1016
 
1017
  <details>
1018
- <summary>🎯 Evaluation methods and challenges ⬇️ </summary>
1019
 
1020
  - **Evaluation methods**: we evaluated various benchmarks using both zero-shot and few-shot methods, except for TruthfulQA.
1021
  - **Zero-shot vs. few-shot**: in chat models, the zero-shot approach is more commonly employed.
@@ -1026,15 +1024,13 @@ If you're seeking to explore the diverse capabilities within Yi's thriving famil
1026
  </details>
1027
 
1028
  ### 📊 Base model performance
1029
- 🎯 Performance evaluation
1030
- - Yi-34B stands out as the top performer among the big models, beating others like LLaMA2-70B and Falcon-180B in most tests.
1031
- - Yi-34B ranks first in MMLU, CMMLU, BBH, and common-sense reasoning.
1032
- - Yi-34B-200K ranks first C-Eval, GAOKAO, and reading comprehension.
1033
 
1034
  ![Base model performance](./assets/img/benchmark_base.png)
1035
 
1036
  <details>
1037
- <summary>🎯 Evaluation methods ⬇️</summary>
1038
 
1039
  - **Disparity in Results**: while benchmarking open-source models, a disparity has been noted between results from our pipeline and those reported by public sources like OpenCompass.
1040
  - **Investigation Findings**: a deeper investigation reveals that variations in prompts, post-processing strategies, and sampling techniques across models may lead to significant outcome differences.
 
150
 
151
  Yi models come in multiple sizes and cater to different use cases. You can also fine-tune Yi models to meet your specific requirements.
152
 
153
+ If you want to deploy Yi models, make sure you meet the [software and hardware requirements](#deployment).
154
 
155
  ### Chat models
156
 
 
331
 
332
  #### Step 0: Prerequistes
333
 
334
+ - Make sure Python 3.10 or a later version is installed.
335
 
336
  - If you want to run other Yi models, see [software and hardware requirements](#deployment)
337
 
 
833
  <div align="right"> [ <a href="#building-the-next-generation-of-open-source-and-bilingual-llms">Back to top ⬆️ </a> ] </div>
834
 
835
  ### Deployment
836
+
837
+ If you want to deploy Yi models, make sure you meet the software and hardware requirements.
838
 
839
  #### Software requirements
840
 
 
845
  Yi 4-bit quantized models | [AWQ and CUDA](https://github.com/casper-hansen/AutoAWQ?tab=readme-ov-file#install-from-pypi)
846
  Yi 8-bit quantized models | [GPTQ and CUDA](https://github.com/PanQiWei/AutoGPTQ?tab=readme-ov-file#quick-installation)
847
 
 
848
  #### Hardware requirements
849
 
850
  Before deploying Yi in your environment, make sure your hardware meets the following requirements.
 
880
  | Yi-34B | 72 GB | 4 x RTX 4090 <br> A800 (80 GB) |
881
  | Yi-34B-200K | 200 GB | 4 x A800 (80 GB) |
882
 
 
 
883
  ### Learning hub
884
+
885
  <details>
886
+ <summary> If you want to learn Yi, you can find a wealth of helpful educational resources here ⬇️</summary>
887
  <br>
888
+
889
  Welcome to the Yi learning hub!
890
 
891
  Whether you're a seasoned developer or a newcomer, you can find a wealth of helpful educational resources to enhance your understanding and skills with Yi models, including insightful blog posts, comprehensive video tutorials, hands-on guides, and more.
 
896
 
897
  With all these resources at your fingertips, you're ready to start your exciting journey with Yi. Happy learning! 🥳
898
 
899
+ #### Tutorials
900
 
901
  | Type | Deliverable | Date | Author |
902
  |-------------|--------------------------------------------------------|----------------|----------------|
 
1007
  - [📊 Base model performance](#-base-model-performance)
1008
 
1009
  ### 📊 Chat model performance
1010
+
1011
+ Yi-34B-Chat model demonstrates exceptional performance, ranking first among all existing open-source models in the benchmarks including MMLU, CMMLU, BBH, GSM8k, and more.
 
1012
 
1013
  ![Chat model performance](./assets/img/benchmark_chat.png)
1014
 
1015
  <details>
1016
+ <summary> Evaluation methods and challenges ⬇️ </summary>
1017
 
1018
  - **Evaluation methods**: we evaluated various benchmarks using both zero-shot and few-shot methods, except for TruthfulQA.
1019
  - **Zero-shot vs. few-shot**: in chat models, the zero-shot approach is more commonly employed.
 
1024
  </details>
1025
 
1026
  ### 📊 Base model performance
1027
+
1028
+ The Yi-34B and Yi-34B-200K models stand out as the top performers among open-source models, especially excelling in MMLU, CMML, common-sense reasoning, reading comprehension, and more.
 
 
1029
 
1030
  ![Base model performance](./assets/img/benchmark_base.png)
1031
 
1032
  <details>
1033
+ <summary> Evaluation methods ⬇️</summary>
1034
 
1035
  - **Disparity in Results**: while benchmarking open-source models, a disparity has been noted between results from our pipeline and those reported by public sources like OpenCompass.
1036
  - **Investigation Findings**: a deeper investigation reveals that variations in prompts, post-processing strategies, and sampling techniques across models may lead to significant outcome differences.