Can try code as long text data.
#1
by
win10
- opened
You can try to use code as long text data, try to use programming language to enhance the automatic programming ability of the model, use a complete C language project or a complete Python project.
Hi @win10 that's a great suggestion! The LCC dataset goes into that direction (512-10k tokens), although afaik the documents are still based on single code files and not entire projects.