The examples.sh file gives the commands to run epic training with various settings and the poison_path is an input argument in the command. Based on the poison_path the dataset index is chosen internally in train_poison_epic.py but there are no examples of the poison_path provided or what its format should be. Without the correct format, the code will invariably throw an error. Can some examples and details on the structure of the poison path and its contents be provided for clarity?
While some precomputed poisoned datasets are provided, how do I compute the poisons for any other custom dataset and store them in the appropriate directory structure?
The
examples.shfile gives the commands to run epic training with various settings and thepoison_pathis an input argument in the command. Based on thepoison_paththe dataset index is chosen internally intrain_poison_epic.pybut there are no examples of thepoison_pathprovided or what its format should be. Without the correct format, the code will invariably throw an error. Can some examples and details on the structure of the poison path and its contents be provided for clarity?While some precomputed poisoned datasets are provided, how do I compute the poisons for any other custom dataset and store them in the appropriate directory structure?