merged master

Signed-off-by: ftgreat <[email protected]>
FlagAI-Open · Aug 24, 2023 · 1e70db5 · 1e70db5
2 parents fdb485d + 33a5afa
commit 1e70db5
Show file tree

Hide file tree

Showing 15 changed files with 99,796 additions and 28 deletions.
diff --git a/examples/Aquila/Aquila-chat/README.md b/examples/Aquila/Aquila-chat/README.md
@@ -28,13 +28,13 @@
 
 悟道·天鹰Aquila系列模型将持续开源更优版本，大家可以先删除原来目录下的`checkpoints_in/aquilachat-7b`，再下载新权重，其他使用方式不变。
 
-- 2023/07/24 ：发布权重文件 v0.9，开源了 AquilaCode-multi、AquilaCode-py。 AquilaChat-7B和Aquila-7B权重无更新, AquilaCode-7B-NV和AquilaCode-7B-TS权重暂时不会有更新计划。
+- 2023-08-15 ：发布权重文件 v1.0，开源了 AquilaCode-multi、AquilaCode-py。 AquilaChat-7B和Aquila-7B权重无更新, AquilaCode-7B-NV和AquilaCode-7B-TS权重暂时不会有更新计划。
   - Aquila-7B md5: 5b56d31c8154c9184a38ff7bc6b4d887
-  - AquilaChat-7B md5: 465683009c8b536ef4cca85febb0227c
+  - AquilaChat-7B md5: 883e83286ee309dbb624016256e30d4c
   - AquilaCode-multi md5：07cfce9440a0fa1ac2768b39d2cf4286
   - AquilaCode-py md5：3faa85fc03d8fda70a73064f48d02d85
 
-其中AquilaChat-7B新版本模型在 FlagEval 大模型评测中（ “主观+客观”）相比0.7的版本整体稍有提，其在Chinese-MMLU上提升10%左右，详细评测结果请通过 http://flageval.baai.ac.cn 网站查看。历史版本变更详情见：[变更日志](https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/changelog_zh.md) 。
+其中AquilaChat-7B新版本模型在 FlagEval 大模型评测中（ “客观”）相比上一版在C-Eval、MMLU、BoolQ上分别提升12.46%，10.88%和9.93%。详细评测结果请通过 http://flageval.baai.ac.cn 网站查看。历史版本变更详情见：[变更日志](https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/changelog_zh.md) 。
 
 
 <br>如有使用问题请先查看 [FAQ](https://github.com/FlagAI-Open/FlagAI/issues/371)，若不能解决，请直接提交 [issue](https://github.com/FlagAI-Open/FlagAI/issues) ~

diff --git a/examples/Aquila/Aquila-chat/README_en.md b/examples/Aquila/Aquila-chat/README_en.md
@@ -30,13 +30,13 @@ The additional details of the Aquila model will be presented in the official tec
 
 We will continue to release improved versions of Aquila model as open source. You can start by deleting the `checkpoints_in/aquilachat-7b` in the original directory and then download the new weights. Other usage methods remain unchanged. For more details, please refer to the folloing change log:
 
-- 2023/07/24 ：Released v0.8 checkpoint files，AquilaCode-multi and AquilaCode-python have been released while AquilaCode-7B-NV and AquilaCode-7B-TS are temporarily not maintained. There are no updates for the weights of Aquila-7B and AquilaChat-7B.
+- 2023-08-15 ：Released v1.0 checkpoint files，AquilaCode-multi and AquilaCode-python have been released while AquilaCode-7B-NV and AquilaCode-7B-TS are temporarily not maintained. There are no updates for the weights of Aquila-7B and AquilaChat-7B.
   - Aquila-7B md5: 5b56d31c8154c9184a38ff7bc6b4d887
-  - AquilaChat-7B md5: 465683009c8b536ef4cca85febb0227c
+  - AquilaChat-7B md5: 883e83286ee309dbb624016256e30d4c            
   - AquilaCode-multi md5：07cfce9440a0fa1ac2768b39d2cf4286
   - AquilaCode-py md5：3faa85fc03d8fda70a73064f48d02d85
 
-In the FlagEval large model evaluation ("Subjective + Objective"), AquilaChat-7B v0.8 has shown a slight overall improvement compared to version 0.7. It achieved an improvement of around 10% on the Chinese-MMLU dataset. For detailed evaluation results, please refer to the website http://flageval.baai.ac.cn.
+In the FlagEval large model evaluation ("Subjective + Objective"), AquilaChat-7B v1.0 has shown a slight overall improvement compared to last version. It achieved an improvement of around 12.46% on the C-Eval, 10.88% on the MMLU, and 9.93% on the BoolQ dataset. For detailed evaluation results, please refer to the website http://flageval.baai.ac.cn.
 For detailed version change history, see [Change Log](https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/changelog.md).
 
  <br>If you have any question, please refer to the [FAQ](https://github.com/FlagAI-Open/FlagAI/issues/371) first. If you cannot solve them, please submit an [issue](https://github.com/FlagAI-Open/FlagAI/issues) directly.

diff --git a/examples/Aquila/Aquila-code/README.md b/examples/Aquila/Aquila-code/README.md
@@ -28,9 +28,9 @@
 
 悟道·天鹰Aquila系列模型将持续开源更优版本，大家可以先删除原来目录下的code模型路径，再下载新权重，其他使用方式不变。
 
-- 2023/07/24 ：发布权重文件 v0.9，开源了 AquilaCode-multi、AquilaCode-py。 AquilaChat-7B和Aquila-7B权重无更新, AquilaCode-7B-NV和AquilaCode-7B-TS权重暂时不会有更新计划。
-  - Aquila-7B md5: 18eac56434db0198494b22b321633785
-  - AquilaChat-7B md5: 465683009c8b536ef4cca85febb0227c
+- 2023-08-15 ：发布权重文件 v1.0，开源了 AquilaCode-multi、AquilaCode-py。 AquilaChat-7B和Aquila-7B权重无更新, AquilaCode-7B-NV和AquilaCode-7B-TS权重暂时不会有更新计划。
+  - Aquila-7B md5: 5b56d31c8154c9184a38ff7bc6b4d887
+  - AquilaChat-7B md5: 883e83286ee309dbb624016256e30d4c
   - AquilaCode-multi md5：07cfce9440a0fa1ac2768b39d2cf4286
   - AquilaCode-py md5：3faa85fc03d8fda70a73064f48d02d85
 

diff --git a/examples/Aquila/Aquila-code/README_en.md b/examples/Aquila/Aquila-code/README_en.md
@@ -30,9 +30,9 @@ The additional details of the Aquila model will be presented in the official tec
 
 We will continue to release improved versions of Aquila model as open source. You can start by deleting the checkpoint file in the original directory and then download the new weights. Other usage methods remain unchanged. For more details, please refer to the folloing change log:
 
-- 2023/07/24 ：Released v0.8 checkpoint files，AquilaCode-multi and AquilaCode-python have been released while AquilaCode-7B-NV and AquilaCode-7B-TS are temporarily not maintained. There are no updates for the weights of Aquila-7B and AquilaChat-7B.
-  - Aquila-7B md5: 18eac56434db0198494b22b321633785
-  - AquilaChat-7B md5: 465683009c8b536ef4cca85febb0227c
+- 2023-08-15 ：Released v1.0 checkpoint files，AquilaCode-multi and AquilaCode-python have been released while AquilaCode-7B-NV and AquilaCode-7B-TS are temporarily not maintained. There are no updates for the weights of Aquila-7B and AquilaChat-7B.
+  - Aquila-7B md5: 5b56d31c8154c9184a38ff7bc6b4d887
+  - AquilaChat-7B md5: 883e83286ee309dbb624016256e30d4c
   - AquilaCode-multi md5：07cfce9440a0fa1ac2768b39d2cf4286
   - AquilaCode-py md5：3faa85fc03d8fda70a73064f48d02d85
 

diff --git a/examples/Aquila/Aquila-pretrain/README.md b/examples/Aquila/Aquila-pretrain/README.md
@@ -28,13 +28,13 @@
 
 悟道·天鹰Aquila系列模型将持续开源更优版本，大家可以先删除原来目录下的`checkpoints_in/aquila-7b`，再下载新权重，其他使用方式不变。
 
-- 2023/07/24 ：发布权重文件 v0.9，开源了 AquilaCode-multi、AquilaCode-py。 AquilaChat-7B和Aquila-7B权重无更新, AquilaCode-7B-NV和AquilaCode-7B-TS权重暂时不会有更新计划。
-  - Aquila-7B md5: 18eac56434db0198494b22b321633785
-  - AquilaChat-7B md5: 465683009c8b536ef4cca85febb0227c
+- 2023-08-15 ：发布权重文件 v1.0，开源了 AquilaCode-multi、AquilaCode-py。 AquilaChat-7B和Aquila-7B权重无更新, AquilaCode-7B-NV和AquilaCode-7B-TS权重暂时不会有更新计划。
+  - Aquila-7B md5: 5b56d31c8154c9184a38ff7bc6b4d887
+  - AquilaChat-7B md5: 883e83286ee309dbb624016256e30d4c
   - AquilaCode-multi md5：07cfce9440a0fa1ac2768b39d2cf4286
   - AquilaCode-py md5：3faa85fc03d8fda70a73064f48d02d85
 
-Aquila-7B 新版本模型 在 FlagEval 大模型评测中（ “客观”）相比0.7的版本在MMLU_Chinese、TruthfulQA、MMLU上分别提升10.07%，14.84%和7.94%。详细评测结果请通过 http://flageval.baai.ac.cn 网站查看。历史版本变更详情见：[变更日志](https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/changelog_zh.md) 。
+Aquila-7B 新版本模型在 FlagEval 大模型评测中（ “客观”）相比上一版在TruthfulQA中提升9.09%。详细评测结果请通过 http://flageval.baai.ac.cn 网站查看。历史版本变更详情见：[变更日志](https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/changelog_zh.md) 。
 
 
 <br>如有使用问题请先查看 [FAQ](https://github.com/FlagAI-Open/FlagAI/issues/371)，若不能解决，请直接提交 [issue](https://github.com/FlagAI-Open/FlagAI/issues) ~

diff --git a/examples/Aquila/Aquila-pretrain/README_en.md b/examples/Aquila/Aquila-pretrain/README_en.md
@@ -29,13 +29,13 @@ The additional details of the Aquila model will be presented in the official tec
 
 We will continue to release improved versions of Aquila model as open source. You can start by deleting the `checkpoints_in/aquila-7b` in the original directory and then download the new weights. Other usage methods remain unchanged. For more details, please refer to the folloing change log:
 
-- 2023/07/24 ：Released v0.8 checkpoint files，AquilaCode-multi and AquilaCode-python have been released while AquilaCode-7B-NV and AquilaCode-7B-TS are temporarily not maintained. There are no updates for the weights of Aquila-7B and AquilaChat-7B.
-  - Aquila-7B md5: 18eac56434db0198494b22b321633785
-  - AquilaChat-7B md5: 465683009c8b536ef4cca85febb0227c
+- 2023-08-15 ：Released v1.0 checkpoint files，AquilaCode-multi and AquilaCode-python have been released while AquilaCode-7B-NV and AquilaCode-7B-TS are temporarily not maintained. There are no updates for the weights of Aquila-7B and AquilaChat-7B.
+  - Aquila-7B md5: 5b56d31c8154c9184a38ff7bc6b4d887
+  - AquilaChat-7B md5: 883e83286ee309dbb624016256e30d4c
   - AquilaCode-multi md5：07cfce9440a0fa1ac2768b39d2cf4286
   - AquilaCode-py md5：3faa85fc03d8fda70a73064f48d02d85
 
-Aquila-7B v0.8 has shown improvements in the FlagEval large model evaluation ("Objective") compared to version 0.7. It achieved improvements of approximately 10.07% on MMLU_Chinese, 14.84% on TruthfulQA, and 7.94% on MMLU datasets. For detailed evaluation results, please refer to the website http://flageval.baai.ac.cn.
+Aquila-7B has shown improvements in the FlagEval large model evaluation ("Objective") compared to last. It achieved improvements of approximately 10.07% on MMLU_Chinese, 14.84% on TruthfulQA, and 7.94% on MMLU datasets. For detailed evaluation results, please refer to the website http://flageval.baai.ac.cn.
 For detailed version change history, see [Change Log](https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/changelog.md).
 
  <br>If you have any question, please refer to the [FAQ](https://github.com/FlagAI-Open/FlagAI/issues/371) first. If you cannot solve them, please submit an [issue](https://github.com/FlagAI-Open/FlagAI/issues) directly.

diff --git a/examples/Aquila/Aquila-tokenizer-hf/special_tokens_map.json b/examples/Aquila/Aquila-tokenizer-hf/special_tokens_map.json
@@ -0,0 +1,6 @@
+{
+  "bos_token": "[CLS]",
+  "eos_token": "</s>",
+  "unk_token": "<|endoftext|>",
+  "pad_token": "<|endoftext|>"
+}
diff --git a/examples/Aquila/Aquila-tokenizer-hf/vocab.json.llama.cpp b/examples/Aquila/Aquila-tokenizer-hf/vocab.json.llama.cpp