ToolBench_toolllama_G123_dfs

Dataset mentioned for ToolBench project https://github.com/OpenBMB/ToolBench

They were in the google drive data.zip https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J

These two json are already processed by the original author. Just plugin into the ToolBnech repo deepseed arguments.

```
--data_path ./toolllama_G123_dfs_train.json \
--eval_data_path ./toolllama_G123_dfs_eval.json \
```

~~My objective is to tailer the training data to 1/100 size and used them for the LLaMA-Factory project. https://github.com/hiyouga/LLaMA-Factory~~

So that more open source models could benifit from function calling dataset.

## Edit

The objective is obtained by using another dataset instead: https://huggingface.co/datasets/Yhyu13/glaive-function-calling-v2-llama-factory-convert
It is smaller and better.

本数据集为ToolBench项目（https://github.com/OpenBMB/ToolBench）所使用的数据集。其存储于Google Drive的data.zip压缩包中，访问链接为：https://drive.google.com/drive/folders/1yBUQ732mPu-KclJnuQELEhtKakdXFc3J。
上述两份JSON文件已由原作者完成预处理，可直接集成至ToolBench仓库的DeepSpeed参数配置中，示例命令如下：

--data_path ./toolllama_G123_dfs_train.json
--eval_data_path ./toolllama_G123_dfs_eval.json

~~我的原目标是将训练数据裁剪至原规模的1/100，并将其适配至LLaMA-Factory项目（https://github.com/hiyouga/LLaMA-Factory）中~~
以期让更多开源模型能够从函数调用（function calling）数据集获益。

## 调整说明
本项目现已更换目标数据集，采用https://huggingface.co/datasets/Yhyu13/glaive-function-calling-v2-llama-factory-convert 数据集，该数据集规模更小且效果更优。