Hello, I uploaded a single docker image on Sept 3rd (before the deadline) and submitted it to both toy and challenge tasks. Today I got feedback emails. The toy sample task worked fine: "Your toy-task scores for the sample-level task are: -brain: 1.00000 (runtime: 1468.70 sec for all samples) -abdominal: 1.00000 (runtime: 1806.31 sec for all samples)" However the next email, which I believe was for the challenge sample task (though it is titled "Mood toy result" and refers to "toy-task") says the testing failed: "Your toy-task scores for the sample-level task are: -brain: failed Log: Traceback (most recent call last): File "/workspace/pred_simple.py", line 328, in predict_folder_pixel_abs(input_dir, output_dir) File "/workspace/pred_simple.py", line 223, in predict_folder_pixel_abs checkpoint.load_model() File "/workspace/oodd/models/checkpointing.py", line 37, in load_model self.model = oodd.models.load_model(self.path, device=device) File "/workspace/oodd/models/base_module.py", line 24, in load_model raise RuntimeError(f'Name of class of model to load not specified and not saved in checkpoint: {path}') RuntimeError: Name of class of model to load not specified and not saved in checkpoint: ./models/mood_10mix_1000epochs_no-cool_no-nats [...] Your submission predicted 0 / 605 brain samples and 0 / 423 abdominal samples successfully." Is there any difference in how our code is called for the toy vs. challenge tasks that might explain why our code works for the toy task but not the challenge (test) task?

Created by Tanya Schmah tschmah
Thank you
Hi, run the exact same call for the toy and the challenge dataset (only changing the path and the cwd might change). However, as you probably can understand, we restrict some access to the system. This is why we recommend only using absolute paths and have an extra point for that under the "Common Errors" section on the website. Cheers, David

Difference between toy and challenge tasks page is loading…