Py |
您所在的位置:网站首页 › irt训练 › Py |
Py-IRT的使用(MacOS)
环境 pip 22.2 python 3.10 MacOS 12.3+ 1. 安装pytorch官网地址: https://pytorch.org/get-started/locally/ MacOS要求: MPS acceleration is available on MacOS 12.3+ 使用pip包管理器安装: pip3 install torch torchvision torchaudio使用Anaconda包管理器安装: conda install pytorch::pytorch torchvision torchaudio -c pytorch 2. 安装Pyro官网地址 https://pyro.ai/ 通过pip包管理工具安装 pip3 install pyro-ppl 3. 安装py-irtgithub地址 https://github.com/nd-ball/py-irt 使用pip包管理器安装: pip install py-irt 4. py_irt的使用首先需要一份jsonlines格式的学生答题情况表,其数据结构为: {"subject_id": "pedro", "responses": {"q1": 1, "q2": 0, "q3": 1, "q4": 0}} {"subject_id": "pinguino", "responses": {"q1": 1, "q2": 1, "q3": 0, "q4": 0}} {"subject_id": "ken", "responses": {"q1": 1, "q2": 1, "q3": 1, "q4": 1}} {"subject_id": "burt", "responses": {"q1": 0, "q2": 0, "q3": 0, "q4": 0}}subject_id是学生ID,responses包含每一道题的正确与否 运行命令: py-irt train 1pl minitest.jsonlines test-1pl/ --lr 0.02 --epochs 100训练命令你可以直接在终端上跑,也可以用jupyter 命令用法: Usage: py-irt train [OPTIONS] MODEL_TYPE DATA_PATH OUTPUT_DIR Arguments: MODEL_TYPE [required] DATA_PATH [required] OUTPUT_DIR [required] Options: --epochs INTEGER --priors TEXT --dims INTEGER --lr FLOAT --lr-decay FLOAT --device TEXT [default: cpu] --initializers TEXT --config-path TEXT --dropout FLOAT [default: 0.5] --hidden INTEGER [default: 100] --help Show this message and exit. We then run the py-irt package to fit a 1PL model and save it to a specified output directory.训练结束,得到两个文件: best_parameters.json parameters.json 其结构如下: { "ability": [ 0.202667236328125, -0.95220547914505, 0.9865020513534546, 0.21425879001617432 ], "diff": [ 0.5277286767959595, -0.19620391726493835, -0.7283022403717041, -0.011020521633327007 ], "irt_model": "1pl", "item_ids": { "0": "q4", "1": "q3", "2": "q1", "3": "q2" }, "subject_ids": { "0": "pedro", "1": "burt", "2": "ken", "3": "pinguino" } }从json文件中就能得到学生能力值和题目难度 以上就是py-irt的用法 |
今日新闻 |
点击排行 |
|
推荐新闻 |
图片新闻 |
|
专题文章 |
CopyRight 2018-2019 实验室设备网 版权所有 win10的实时保护怎么永久关闭 |