pial08 / semvuldet Goto Github PK

Once I train this model and get the model.bin file, how can I run predictions on new pieces of code?

Hello, great job.
I'm having some difficulty reproducing your code and would sincerely appreciate your help.
When generating the "SVG" mentioned in your paper, I noticed that you seem to be working directly on the subword IDs based on the sliding window, and the question is that this doesn't seem to be generating the data flow graphs and sequential flow graphs mentioned in the "SVG"?
Also when generating "PF" edges, it seems that the generated "PF" edges are generated by Tokens, which doesn't work during training.
These problems are in the "build_graph" function in the file "\code\modelGNN_updates.py".

Hi, pial08. Thanks for sharing the wonderful work. In the data processing, how to convert the source code to the jsonl file?

Meanwhile, what is the meaning of the number before "/n" in multiclass40_final.jsonl (VulF)? Is it the line number?

Looking forward for your reply.

I followed the method provided in this link for preprocessing, but I was unable to obtain the test. jsonl file
link：【 https://github.com/microsoft/CodeXGLUE/tree/main/Code-Code/Defect-detection/dataset】