Digital-Humans-23 / a2

4 stars 0 forks source link

[ex.3] Joint angle cannot pass the test #2

Closed dehezhang2 closed 1 year ago

dehezhang2 commented 1 year ago

May I ask a question about the joint angle calculation? It seems that my calculation has a small error:

Screen Shot 2023-03-18 at 10 02 19 PM

In the task description, the number of joints is 43, but in the ObservationData class the size of the joint_angles array is 44. Is it possible that this mismatch caused the error?

steinraf commented 1 year ago

I am also getting a similarly offset of 0.009954251707404079.

As a side question, does the calculation of your torque work for the dog gaussian and the human reward?

dehezhang2 commented 1 year ago

The torque part can pass the test.

yitongx commented 1 year ago

I also encountered the same issue.

Obtained: 0.00995425170740408 Expected: 0.009990833552124415 ± 1.0e-08

MiguelZamoraM commented 1 year ago

There was a small error creating the test case for Bob. I'm training Bob again to make sure that everything works.
I will update that test case during the day.

MiguelZamoraM commented 1 year ago
dehezhang2 commented 1 year ago

Should we also change the directory in line 8 of the file ./src/python/tests/03_test_human_reward.py?

MiguelZamoraM commented 1 year ago

Yes. Please update the variable pickle_path in line 8 of src/python/tests/03_test_human_reward.py to reflect the changes i.e: pickle_path = '/pickled_files/test_data_bob1679334591.29026.pickle'

drlxj commented 1 year ago

Do you also change the number of joints from 43 to 44? I pass the test if I choose 44, but in Readme the variable is requested to be 43.

MiguelZamoraM commented 1 year ago

@drlxj The actual number of joints is 44. I just fixed the typo on the README.