Unable to test objnav_benchmark.py

[Athirakr94]

Dear Authors, The work seems to be interesting and wanted try this out. To start I have followed the commands in the repo. I am able to use mp3d in policy evaluation it works. But with this command, python objnav_benchmark.py --checkpoint=/home/athira/Pixel-Navigator/checkpoints/pixelnav_C.ckpt I am getting error. Screen shot shared for reference. The val split of hm3d dataset is downloaded to /data/athira/hm3d Please help.

[Athirakr94]

That was a path issue I could resolve that error. But now a new index error is coming. Any idea

[wzcai99]

Hi, I checked my code and remember that I comment one line of the habitat-lab code, which is in the line 106 of the habitat-lab/habitat-lab/habitat/core/env.py. It seems that only comments this line can make the scene_dataset_config path correct. Maybe there should be a better way to solve this.

[wzcai99]

And you can try whether this can solve your problem @Athirakr94.

[Athirakr94]

Thanks I tried commenting line 106, but then its throwing a validation error as, raise ValidationError("Cannot convert '$VALUE_TYPE' to string: '$VALUE'") omegaconf.errors.ValidationError: Cannot convert 'tuple' to string: '()' full_key: habitat.simulator.scene_dataset reference_type=SimulatorConfig object_type=SimulatorConfig So tried commenting lines 105-107, then old error is coming again ![Uploading image.png…]()

[wzcai99]

Sorry, I cannot see the recent uploaded image, can you upload the error message again?

[Athirakr94]

Thanks for the help I was able to run the script after ensuring draw_goal_aabbs=False. Hope this is fine