crozai/vllm-benchmark-coding
收藏Hugging Face2025-03-05 更新2025-04-26 收录
下载链接：
https://hf-mirror.com/datasets/crozai/vllm-benchmark-coding
下载链接
链接失效反馈官方服务：
资源简介：
这是一个用于轻松基准测试部署在服务模式下的LLM（大型语言模型）的数据集，设计上与vLLM的benchmark_serving.py兼容。数据集来源于四个不同的数据集，经过处理和过滤，只保留了涉及Java、TypeScript、React、Spring、Kafka等关键词的对话。数据集包含两个特征字段：from和value，都是字符串类型，并且划分为训练集。训练集共有36894个样本。
This is a dataset for easy benchmarking of deployed LLMs in serving mode, designed to be compatible with vLLMs benchmark_serving.py. The dataset is sourced from four different datasets, processed and filtered to keep only conversations involving keywords such as Java, TypeScript, React, Spring, Kafka. The dataset includes two feature fields: from and value, both of which are strings, and is split into a training set. The training set contains a total of 36,894 samples.
提供机构：
crozai
