path in training is not the same as in preprocessing
Maybe I'm missing something, but :
I have a dataset_config like this :
data_root: DATA/DataSample_Workshop_DNADNA/
dataset_name: BacterialDemoSelection
scenario_params_path: DATA/DataSample_Workshop_DNADNA/SPIBact_param_1000
data_source:
format: dnadna
filename_format: "scenario_{scenario:05}/{dataset_name}_{scenario:05}_{replicate:03}.npz"
init
and preprocess
worked well. It creates the config and run files or folder under BacterialDemoSelection
(folder that dnadna
created, in my working dir). Yet, train
returns an error like that:
.../Workshop_dnadna/BacterialDemoSelection/DATA/DataSample_Workshop_DNADNA/scenario_06335/BacterialDemoSelection_06335_057.npz": FileNotFoundError
Which indeed does not exist, but .../Workshop_dnadna/DATA/DataSample_Workshop_DNADNA/scenario_06335/BacterialDemoSelection_06335_057.npz"
exists.
Why does it add the dataset_name
(BacterialDemoSelection) in the path ?