clarinsi / csmtiser Goto Github PK

I tried the command that is given in the README.md, but I get the following:

tomaz@new-tantra csmt$ sudo docker run -v ~/csmtiser:/csmtiser greegorey/csmtiser python preprocess.py
python: can't open file 'preprocess.py': [Errno 2] No such file or directory

@nljubesi, would you have any idea what is wrong, or know somebody that does?

(for some reason I also need to run with sudo, otherwise I get a permission error, this might be something to do with our installation, investigating...)

Some parts of the code make the assumption that the current directory (meaning the directory from which you start cSMTiser) is also the working directory (the working_dir given in config.yml), and break if the working directory is set to something else (which I think should be good practice...).

Two parts that I encountered:

• https://github.com/clarinsi/csmtiser/blob/master/preprocess.py#L102 creates the preprocessed LM files in the current directory, but the training script looks for them in the working directory
• https://github.com/clarinsi/csmtiser/blob/master/csmtiser/config.py#L26 assumes the paths for the training files are relative to the working directory, but paths to development and LM files are interpreted relative to the current directory