RomeoV / ml4co-competition

Winner of NeurIPS 2021 student leaderboard. Self-bootstrapping bayesian optimization for SCIP configuration using GNNs.
BSD 3-Clause "New" or "Revised" License
13 stars 5 forks source link

MVP: DataLoader for instances #5

Closed RomeoV closed 3 years ago

RomeoV commented 3 years ago

The dataloader needs to load the pickled problem specifications as well as parse the csv file (probably just using pandas) and extract the necessary information.

Given the pandas dataframe, it should be easy to coordinate that part. The output will be some kind of tuple of the configuration input, the milp problem and the label (TimeConstrPrimalDualIntegral).