This is a QA model for Chinese medicine and it can be consumed by Seq2SeqSharp (https://github.com/zhongkaifu/Seq2SeqSharp). You can clone Seq2SeqSharp, build it and run the model as follows:
git clone https://github.com/zhongkaifu/Seq2SeqSharp.git cd Seq2SeqSharp dotnet build Seq2SeqSharp.sln --configuration Release cd Tools cd Seq2SeqConsole cd bin ./bin/Seq2SeqConsole -Task Test -ModelFilePath medicalQA_chs.model -InputTestFile %InputFilePath% -OutputFile %OutputFilePath% -ProcessorType %Device% -MaxSrcSentLength 128 -MaxTgtSentLength 256 -DeviceIds 0 -DecodingStrategy Sampling -DecodingRepeatPenalty 1
%InputFilePath% is file path of medical questions in Chinese. It's one question per line. It needs to be tokenized by SentencePiece encoder using chsMedSpm.model in this repo
%OutputFilePath% is file path of answers output by the model. It's one answer per line. You can decode them by SentencePiece decoder using chsMedSpm.model in this repo
%Device% is device type for model inference, it can be GPU, CPU and CPU_MKL(For Windows only).