edge-prediction-based-CQA-model

  • a0_259231
    了解作者
  • 2.6KB
    文件大小
  • zip
    文件格式
  • 0
    收藏次数
  • VIP专享
    资源类型
  • 0
    下载次数
  • 2022-05-23 17:18
    上传日期
“基于边缘预测的CQA模型” 目录结构:层:编码器:编码层。 输入:段落,对话历史记录,当前问题。 输出:段落,对话历史记录,当前问题以及每个单词的嵌入推理器:推理层。 输入:编码器层的输出输出:每个词都带有嵌入的证据预测器:预测层。 输入:推理机层的输出输出:答案(答案类型,开始位置,结束位置)接口:定义每个层的输出模型:我们使用的模型类(例如BERT)和我们提出的模型。 proposal_model.py将调用layer.pytrainer.py的函数:在minibatch中进行训练,评估和测试。 调用proposal_model.py main.py:从cmd中读取args并启动该过程,然后调用trainer.py。utils:多层或整个模型使用的通用utils。config.py:可变配置,例如epochs = 5,output_dir data: data log:
edge-prediction-based-CQA-model-main.zip
  • edge-prediction-based-CQA-model-main
  • layers
  • interface
  • encoder_output.py
    992B
  • trainer.py
    0B
  • README.md
    1.3KB
  • config.py
    0B
  • main.py
    0B
  • requirement.txt
    0B
内容介绍
"# edge-prediction-based-CQA-model" Directory structure: layers: encoder: the encoding layer. Input: paragraph, conversation history, current question. output: paragraph, conversation history, current question with embeddings of each words reasoner: the reasoning layer. Input: the output of encoder layer output: evidence with embeddings of each words predicter: the prediction layer. Input: the output of reasoner layer output: answer (answer type, start position, end position) interface: define the output of each layer models: the models class we use like BERT and our proposed model. proposed_model.py will call functions of layer.py trainer.py: do the training, evaluation, testing in the minibatch. call proposed_model.py main.py: read args from cmd and start the process call the trainer.py utils: the common utils used by multi layers or the whole model config.py: mutable configuration like epochs=5, output_dir data: data log: log model: trained model parameters. save them here. requirement.txt: required packages
评论
    相关推荐