-
Notifications
You must be signed in to change notification settings - Fork 0
Open
Description
Hi researchers,
Thank you for your marvellous work! I really enjoy your idea presented and am trying to get the project run on my own computer. However, I have problems running
python3 -m train_sim fit --config configs/sim.yaml --data.data_path <path/to/extracted/sim/dataset/folder> --config configs/logging_sim.yaml
even though I have already run:
python3 -m src.dataset_util.src.generate_dataset_meta <path/to/dataset> <number of samples> --n_workers <number of processes>
python3 -m src.dataset_util.src.visualize_dataset <path/to/dataset> <sample index> <filename_a.pdf> # Optional
python3 -m src.dataset_util.src.compute_cost_to_goal <path/to/dataset> --path_start_dist <distance in meter> --n_workers <number of processes> --n_paths <number of paths>
python3 -m src.dataset_util.src.visualize_dataset <path/to/dataset> <sample index> <filename_b.pdf> # Optional
The problems I have met are:
- What should
<path/to/extracted/sim/dataset/folder>
be? I have generated the dataset in the./dataset
folder. And the json files are stored in.dataset/meta/
- When I set
<path/to/extracted/sim/dataset/folder>
asdataset
. I got FileNotFound Error: "No such file or directory: '/xxx/dataset/sensor_polar_0/008007.png'". I do not have asensor_polar_0
folder in mydataset
folder. I wonder if there is something missing for theSimulation Dataset
in the README guidance? - When I set
<path/to/extracted/sim/dataset/folder>
asdataset/meta
, I got ValueError: "ValueError('num_samples should be a positive integer value, but got num_samples=0')". I think this indicates that<path/to/extracted/sim/dataset/folder>
should be set to thedataset
folder?
To conclude, I think there is something missing in the date generation process about how how convert the json files into pictures that pytorch lightening can take as training data. I would really appreciate it if you can provide me with some guidelines about how to tackle the issue.
Thank you in advance!
Metadata
Metadata
Assignees
Labels
No labels