Final model for the 2018 Recys challenge by the team STORMtroopers
Our solutions can be reproduced as follows:
- Move all of these files to the directory "~/Spotify"
- Store all the csv files in the data directory
- Run transformer.py, transformer2.py, ... , transformer10.py (in that order!)
- Move all csv files to the data directory
- Run Transformer11.py
- Run make_data.py
These steps significantly reduce the disk space required creating a variety of dictionaries. We are now ready to run the models
- Run Challenge1.py
- Run Challenge2_exp.py ...
- Run Challenge10_exp.py
(N.B. We ran these models on large nodes. Some of these models request 25 cores and need 400GB of RAM - You can however reduce the amout of data used to reduce the computational ressources required.)
We can now match all these csv files into one final submission:
- Run CreateSubmission_alternative.py
This is it! Please do not hesitate to get in touch should you have any questions.