Depth image dimension mismatch in parkour code

[tony-hj]

Hi,
Thank you very much for your team's contributions. I've encountered some issues while running the parkour section of the code, and I was hoping you could provide some guidance.
When runing this code:
python scripts/instinct_rl/train.py --headless --task=Instinct-Parkour-Target-Amp-G1-v0
The code produced an error:
File "/home/train/instinct_rl/instinct_rl/modules/parallel_layer.py", line 80, in _build_one_block
model = Conv2dHeadModel(
^^^^^^^^^^^^^^^^
File "/home/train/instinct_rl/instinct_rl/modules/conv2d.py", line 107, in init
c, h, w = image_shape
^^^^^^^
ValueError: not enough values to unpack (expected 3, got 1)

This format looks like an image format, requiring three dimensions. The code uses depth images, but there is only one dimension.

Is this issue related to modifying the depth image configuration or to changes in the code? Thank you again for your team's contribution to the community, and we hope to receive your reply as soon as possible.

[zst1406217]

It seems like you forgot to add encoder_configs to PolicyCfg. Could you please provide the config file so I can take a closer look?