一、目录

1. 步骤
2. demo
3. rouge-n 含义

二、实现

1. 步骤

离线使用evaluate: 1. 下载evaluate 文件：https://github.com/huggingface/evaluate/tree/main
                 2. 离线使用  路径+/evaluate-main/metrics/rouge

2. demo

import evaluate
'''
离线使用evaluate: 1. 下载evaluate 文件：https://github.com/huggingface/evaluate/tree/main
                 2. 离线使用  路径+/evaluate-main/metrics/rouge
'''
rouge=evaluate.load("/app/tensorrt_llm/examples/qwen2/evaluate-main/metrics/rouge")
predictions = ["Transformers Transformers are fast plus efficient",
               "Good Morning", "I am waiting for new Transformers"]
references = [
              ["HuggingFace Transformers are fast efficient plus awesome",
              "Transformers are awesome because they are fast to execute"],

              ["Good Morning Transformers", "Morning Transformers"],

              ["People are eagerly waiting for new Transformer models",
               "People are very excited about new Transformers"]
results = rouge.compute(predictions=predictions, references=references)
print(results)

3. rouge-n 含义
   ROUGE-N: 在 N-gram 上计算召回率
   ROUGE-L: 考虑了机器译文和参考译文之间的最长公共子序列
   ROUGE-W: 改进了ROUGE-L，用加权的方法计算最长公共子序列