howtoosee / cs5340_uncertainty_modelling_in_ai

CS5340 AY2023/2024 Project Group 4

CS5340_Uncertainty_Modelling_in_AI

CS5340 AY2023/2024 Project Group 4

Changes made from OOD code

In openood.utils, dataset_config.num_workers, .gpu, .num_machines have been commented out. The latter 2 is used in deciding if a sampler from torch is necessary, and the former is used for determining number of subprocesses to use when loading in data. For more information, visit https://pytorch.org/docs/stable/data.html

Preprocessors are specific to datapipelines in the paper, for now we set default to base preprocessor and testpreprocessor.

Data

Data from openood is hosted in their wiki page. For this implementation please specifiy the absolute path of the data folder when calling dataloaders(will change later if have time).