update tooleval requirements.txt

README.md

@@ -119,6 +119,10 @@ Install Package (python>=3.9)
 ```bash
 pip install -r requirements.txt
 ```
+or for ToolEval only
+```bash
+pip install -r toolbench/tooleval/requirements.txt
+```
 
 Prepare the data and tool environment:
 ```bash

README_ZH.md

@@ -123,6 +123,10 @@ cd ToolBench
 ```bash
 pip install -r requirements.txt
 ```
+或者仅安装ToolEval需要的包
+```bash
+pip install -r toolbench/tooleval/requirements.txt
+```
 
 准备数据和工具环境:
 ```bash

toolbench/tooleval/README.md

@@ -11,6 +11,11 @@ Our automatic evaluator, developed using ChatGPT, demonstrates a significant cor
 We also obtain the agreement among different human annotators **83.54%**, and the agreement between humans and our evaluator **80.21%**.
 
 ## 🚀Usage
+### Install
+Install Package (python>=3.9)
+```bash
+pip install -r requirements.txt
+```
 
 ### Reproduce our Results
 

toolbench/tooleval/README_ZH.md

@@ -14,6 +14,12 @@
 
 ## 🚀用法
 
+### 安装
+安装包，要求(python>=3.9)
+```bash
+pip install -r requirements.txt
+```
+
 ### 复现结果
 
 要在测试集（如G1-Inst.）上评估模型，可以执行以下命令：

toolbench/tooleval/evaluators/evaluator.py

@@ -4,23 +4,8 @@
 import math
 from typing import List,Union,Dict,Any
 from copy import deepcopy
-from tenacity import retry, wait_random_exponential, stop_after_attempt
-import numpy as np
-import os
+from .utils import OpenaiPoolRequest
 
-class OpenaiPoolRequest:
-    def __init__(self, pool_json_file):
-        self.api_key = os.environ.get('OPENAI_KEY')
-
-    @retry(wait=wait_random_exponential(multiplier=1, max=30), stop=stop_after_attempt(10),reraise=True)
-    def request(self,messages,**kwargs):
-        import openai
-        kwargs['api_key'] = self.api_key
-        return openai.ChatCompletion.create(messages=messages,**kwargs)
-
-    def __call__(self,messages,**kwargs):
-        return self.request(messages,**kwargs)
-
 def process_answer(answer:Dict):
     answer['final_answer'] = answer['final_answer'][:1000]
     answer['answer_details'] = answer['answer_details'][:3000]
@@ -31,6 +16,7 @@ def process_tools(tools:List[Dict]):
         tool.pop('description',None)
         tool.pop('parameters',None)
     return tools
+
 class AutomaticEvaluator:
     def __init__(self,eval_config,template):
         self.eval_config = eval_config

toolbench/tooleval/evaluators/utils.py

@@ -0,0 +1,17 @@
+from tenacity import retry, wait_random_exponential, stop_after_attempt
+import os
+
+
+class OpenaiPoolRequest:
+    def __init__(self, pool_json_file):
+        self.api_key = os.environ.get('OPENAI_KEY')
+
+    @retry(wait=wait_random_exponential(multiplier=1, max=30), stop=stop_after_attempt(10),reraise=True)
+    def request(self,messages,**kwargs):
+        import openai
+        kwargs['api_key'] = self.api_key
+        return openai.ChatCompletion.create(messages=messages,**kwargs)
+
+    def __call__(self,messages,**kwargs):
+        return self.request(messages,**kwargs)
+

toolbench/tooleval/requirements.txt

@@ -0,0 +1,6 @@
+tqdm
+numpy
+pandas
+pydantic
+tenacity
+openai