Toy pipeline on all paper reproduction with code available

What’s needed to reproduce a work?

• get the data downloaded

• Get the environment setup

• Get the work training

• Get the same results

Data could start with a tiny dataset, to make sure the env is consistent and code can be run

Compare the results=>

• start with the raw data?

• Set random seed

Is the random seed making sure all the operations consistent?