Skip to content

Commit

Permalink
update model link
Browse files Browse the repository at this point in the history
  • Loading branch information
@pooruss
pooruss committed Oct 2, 2023
1 parent a46de28 commit ce54183
Show file tree
Hide file tree
Showing 2 changed files with 11 additions and 11 deletions.
12 changes: 6 additions & 6 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -36,11 +36,11 @@
*Read this in [中文](README_ZH.md).*

## What's New
- **[2023/9/29]** A new version **ToolEval** which is more stable and covers more models including GPT4! Please refer to [ToolEval](https://github.com/OpenBMB/ToolBench/tree/master/toolbench/tooleval) for more details.
- **[2023/9/29]** A new version ToolEval which is more stable and covers more models including GPT4! Please refer to [**ToolEval**](https://github.com/OpenBMB/ToolBench/tree/master/toolbench/tooleval) for more details. Besides, [**ToolLLaMA-2-7b-v2**](https://huggingface.co/ToolBench/ToolLLaMA-2-7b-v2) is released with stronger tool-use capabilities. Please use the ToolLLaMA-2-7b-v2 model to reproduce our latest experimental results with the new version ToolEval.

- **[2023/8/30]** Data updation, with more than **120,000** solution path annotations and **intact reasoning thoughts**! Please find `data.zip` on [Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J). *Notice that `data_0801` is the old version data.*
- **[2023/8/30]** Data updation, with more than **120,000** solution path annotations and **intact reasoning thoughts**! Please find `data.zip` on [Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J).

- **[2023/8/8]** No more hallucination! [**ToolLLaMA-2-7b**](https://huggingface.co/ToolBench/ToolLLaMA-2-7b) (fine-tuned from LLaMA-2-7b) is released with lower API hallucination than ChatGPT.
- **[2023/8/8]** No more hallucination! [**ToolLLaMA-2-7b-v1**](https://huggingface.co/ToolBench/ToolLLaMA-2-7b-v1) (fine-tuned from LLaMA-2-7b) is released with lower API hallucination than ChatGPT.

- **[2023/8/4]** We provide **RapidAPI backend service** to free you from using your own RapidAPI key and subscribing the APIs. Please fill out our [form](https://forms.gle/oCHHc8DQzhGfiT9r6). We will review it as soon as possible and send you the ToolBench key to get start on it!

Expand Down Expand Up @@ -111,7 +111,7 @@ ToolBench contains both single-tool and multi-tool scenarios. The multi-tool sce

### Data Release

Please download our dataset using the following link: [Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J) or [Tsinghua Cloud](https://cloud.tsinghua.edu.cn/f/c9e50625743b40bfbe10/).
Please download our dataset using the following link: [Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J) or [Tsinghua Cloud](https://cloud.tsinghua.edu.cn/f/c9e50625743b40bfbe10/). *Notice that `data_0801` is the old version data.*
The file structure is as follows:
```
├── /data/
Expand Down Expand Up @@ -142,7 +142,7 @@ Here is some descriptions for the `data` directory:

## 🤖Model

We release the [ToolLLaMA-7b](https://huggingface.co/ToolBench/ToolLLaMA-7b), [ToolLLaMA-7b-LoRA](https://huggingface.co/ToolBench/ToolLLaMA-7b-LoRA) and [ToolLLaMA-2-7b](https://huggingface.co/ToolBench/ToolLLaMA-2-7b) models, which are both trained on the released dataset in a multi-task fashion. We also release the [tool retriever](https://huggingface.co/ToolBench/ToolBench_IR_bert_based_uncased) trained under our experimental setting.
We release the [ToolLLaMA-2-7b-v2](https://huggingface.co/ToolBench/ToolLLaMA-2-7b-v2) which is trained on the latest version data, and [ToolLLaMA-7b-v1](https://huggingface.co/ToolBench/ToolLLaMA-7b-v1), [ToolLLaMA-7b-LoRA-v1](https://huggingface.co/ToolBench/ToolLLaMA-7b-LoRA-v1) which are trained on the 0801 version data. All models are trained on the released dataset in a multi-task fashion. We also release the [tool retriever](https://huggingface.co/ToolBench/ToolBench_IR_bert_based_uncased) trained under our experimental setting.

## 🚀Fine-tuning
### Install
Expand Down Expand Up @@ -609,7 +609,7 @@ Please refer to [ToolEval](https://github.com/OpenBMB/ToolBench/tree/master/tool
### 📊 Model Experiments Results


In our main experiments, ToolLLaMA demonstrates a compelling capability to handle both single-tool and complex multi-tool instructions, which on a par with ChatGPT.
In our main experiments, ToolLLaMA(v2) demonstrates a compelling capability to handle both single-tool and complex multi-tool instructions, which on a par with ChatGPT.
Below are the main results. Win rate for each model is compared with ChatGPT-ReACT.


Expand Down
10 changes: 5 additions & 5 deletions README_ZH.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -36,11 +36,11 @@
*英文[README](README.md)链接.*

## 最新支持
- **[2023/9/29]** 更稳定的更新版本**ToolEval**， 加上更多模型比如GPT4的结果！更多细节请参考[ToolEval](https://github.com/OpenBMB/ToolBench/tree/master/toolbench/tooleval).
- **[2023/9/29]** 更稳定的更新版本**ToolEval**， 加上更多模型比如GPT4的结果！更多细节请参考[ToolEval](https://github.com/OpenBMB/ToolBench/tree/master/toolbench/tooleval). 除此之外，工具使用能力更强的[**ToolLLaMA-2-7b-v2**](https://huggingface.co/ToolBench/ToolLLaMA-2-7b-v2)模型已经开放，请使用这版模型和更新的tooleval来复现最新的实验结果。

- **[2023/8/30]** 数据更新，拥有超过**12万**解路径标注和**完整的推理thoughts**！请在 [Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J) 上找到`data.zip`。*请注意：`data_0801.zip`是老版本数据。*
- **[2023/8/30]** 数据更新，拥有超过**12万**解路径标注和**完整的推理thoughts**！请在 [Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J) 上找到`data.zip`。

- **[2023/8/8]** 告别幻觉！[**ToolLLaMA-2-7b**](https://huggingface.co/ToolBench/ToolLLaMA-2-7b) (从LLaMA-2-7b微调而来)模型已发布，比ChatGPT有着更少的API幻觉现象.
- **[2023/8/8]** 告别幻觉！[**ToolLLaMA-2-7b-v1**](https://huggingface.co/ToolBench/ToolLLaMA-2-7b-v1) (从LLaMA-2-7b微调而来)模型已发布，比ChatGPT有着更少的API幻觉现象.

- **[2023/8/4]** 我们提供RapidAPI后端服务，以免您使用自己的RapidAPI私钥去订阅API。填写[表单](https://forms.gle/oCHHc8DQzhGfiT9r6)后，我们会尽快审核并给您发送ToolBench key去请求该后端服务!

Expand Down Expand Up @@ -117,7 +117,7 @@ ToolBench包含单工具和多工具场景。多工具场景可以进一步分

### 数据发布

请使用以下链接下载我们的数据集：[Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J)或者[清华云盘](https://cloud.tsinghua.edu.cn/f/c9e50625743b40bfbe10/).
请使用以下链接下载我们的数据集：[Google Drive](https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J)或者[清华云盘](https://cloud.tsinghua.edu.cn/f/c9e50625743b40bfbe10/).*请注意：`data_0801.zip`是老版本数据。*
文件结构如下:
```
├── /data/
Expand Down Expand Up @@ -145,7 +145,7 @@ ToolBench包含单工具和多工具场景。多工具场景可以进一步分
- `toolllama_G123_dfs_train.json` 和 `toolllama_G123_dfs_eval.json`：预处理数据，可用于直接训练 toolllama 并复现我们的结果。对于预处理细节，我们将 G1、G2 和 G3 数据分别分为训练、评估和测试部分，合并各数据集的训练数据进行训练。

## 🤖模型
我们发布了全参数微调版本[ToolLLaMA-7b](https://huggingface.co/ToolBench/ToolLLaMA-7b)和lora版本[ToolLLaMA-7b-LoRA](https://huggingface.co/ToolBench/ToolLLaMA-7b-LoRA)，都是在发布的数据集上以多任务方式训练的。我们也发布在实验设置下训练的[tool retriever](https://huggingface.co/ToolBench/ToolBench_IR_bert_based_uncased).
我们发布了在最新数据上全参数微调的[ToolLLaMA-2-7b-v2](https://huggingface.co/ToolBench/ToolLLaMA-2-7b-v2)，还有在0801版本数据上全参数微调的[ToolLLaMA-7b-v1](https://huggingface.co/ToolBench/ToolLLaMA-7b-v1)和lora版本[ToolLLaMA-7b-LoRA-v1](https://huggingface.co/ToolBench/ToolLLaMA-7b-LoRA-v1)，所有模型都是以多任务方式训练的。我们也发布在实验设置下训练的[tool retriever](https://huggingface.co/ToolBench/ToolBench_IR_bert_based_uncased).
## 🚀精调
### 安装
克隆这个仓库并进入ToolBench文件夹。
Expand Down

0 comments on commit ce54183

Please sign in to comment.