unrecognized arguments: --dataset_kwargs {

[mmitterma]

Hi there,

we have an error: unrecognized arguments. Do you guys have any ideas why this occurs?

We are also confused, which train_model.py script is the newest one?

• python train_model.py --n_iter 50000 --path_dataset_csv data/csvs/dna/train.csv --dataset_kwargs '{"transform_signal": ["fnet.transforms.Normalize()", "fnet.transforms.Resizer((1, 0.37241, 0.37241))"], "transform_target": ["fnet.transforms.Normalize()", "fnet.transforms.Resizer((1, 0.37241, 0.37241))"] }' --bpds_kwargs '{"buffer_size": 1 }' --path_run_dir saved_models/dna --gpu_ids 0 usage: train_model.py [-h] [--batch_size BATCH_SIZE] [--bpds_kwargs BPDS_KWARGS] [--buffer_size BUFFER_SIZE] [--buffer_switch_frequency BUFFER_SWITCH_FREQUENCY] [--class_dataset CLASS_DATASET] [--gpu_ids GPU_IDS [GPU_IDS ...]] [--interval_save INTERVAL_SAVE] [--iter_checkpoint ITER_CHECKPOINT [ITER_CHECKPOINT ...]] [--lr LR] [--n_iter N_ITER] [--nn_kwargs NN_KWARGS] [--nn_module NN_MODULE] [--patch_size PATCH_SIZE [PATCH_SIZE ...]] [--path_dataset_csv PATH_DATASET_CSV] [--path_dataset_val_csv PATH_DATASET_VAL_CSV] [--path_run_dir PATH_RUN_DIR] [--seed SEED] [--shuffle_images] [--transform_signal TRANSFORM_SIGNAL [TRANSFORM_SIGNAL ...]] [--transform_target TRANSFORM_TARGET [TRANSFORM_TARGET ...]] train_model.py: error: unrecognized arguments: --dataset_kwargs {"transform_signal": ["fnet.transforms.Normalize()", "fnet.transforms.Resizer((1, 0.37241, 0.37241))"], "transform_target": ["fnet.transforms.Normalize()", "fnet.transforms.Resizer((1, 0.37241, 0.37241))"] }

[counkomol]

You can see from the error message that the training script does not support the --dataset_kwargs argument. Can I ask where did you get this from?

[mmitterma]

Thank you for your reply! I just tried your code, and wanted to train a model, but in the current version is no train_model.py (not in the folder "scripts") provided. So, after I startet the script train_model.sh it stopped, because there was no train_model.py. Then I used the train_model.py file from the "original paper" release branche1. Now, the train_model.sh started and the train_model.py was started by the shell script, but it stopped again due to the error message train_model.py: error: unrecognized arguments: --dataset_kwargs {"transform_signal": . Do you have any ideas how to solve this issue? Thank you very much in advance!

[counkomol]

You'll need to stick to one snapshot of the repository. For example, release_1. If you mix and match different files from different repository states, then there is no way for us to know what to expect.

We suggest you try a fresh copy of release_1 again and let us know if you find anything.

[mmitterma]

Thank you! I will stick to one repository. The issue is that there is no train_model.py file in the newest repository. Could your please provide it?

[counkomol]

The train_model.py file does not exist because that is intended. With that said, the newest version of the repository is in active development and is currently not meant for outside use.

[mmitterma]

Ok, thank you!