Error when starting pgdrive: "ValueError: matmul: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (n?,k),(k,m?)->(n?,m?) (size 275 is different from 274)"

[rondey]

Hello, I tried to run "python3 -m pgdrive.examples.enjoy_expert" and this is the log that I obtain:

"Successfully registered the following environments: ['PGDrive-test-v0', 'PGDrive-validation-v0', 'PGDrive-v0', 'PGDrive-10envs-v0', 'PGDrive-1000envs-v0', 'PGDrive-training0-v0', 'PGDrive-training1-v0', 'PGDrive-training2-v0'].
Known pipe types:
glxGraphicsPipe
(1 aux display modules not yet loaded.)
:display(warning): FrameBufferProperties available less than requested.
requested: depth_bits=1 color_bits=3 red_bits=1 green_bits=1 blue_bits=1 alpha_bits=1 multisamples=8 back_buffers=1 force_hardware
got: depth_bits=32 color_bits=24 red_bits=8 green_bits=8 blue_bits=8 alpha_bits=8 multisamples=4 back_buffers=1 force_hardware force_software
:device(warning): /dev/input/event4 is not readable, some features will be unavailable.
WARNING:root:Since your screen is too small (1920, 929), we resize the window to (1114, 836).
Closing the environment!
Episode count 0, Success rate: 0, Average reward: 0
Traceback (most recent call last):
File "/usr/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/usr/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/home/davide/pgdrive/pgdrive/examples/enjoy_expert.py", line 19, in
action = expert(obs)
File "/home/davide/pgdrive/pgdrive/examples/ppo_expert/numpy_expert.py", line 32, in expert
x = np.matmul(obs, weights["default_policy/fc_1/kernel"]) + weights["default_policy/fc_1/bias"]
ValueError: matmul: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (n?,k),(k,m?)->(n?,m?) (size 275 is different from 274)"

The OS is Ubuntu 20.04, can you help me? Thanks

[QuanyiLi]

Hi rondey, thanks for your feedback. It happens due to the observation dimension mismatch, but we will not fix it in PGDrive. Actually, we refactor the whole architecture of PGDrive and move the updated version to a new repo, named MetaDrive, which also has more features than PGDrive. In the future, we will maintain the MetaDrive instead of PGDrive, so you can use MetaDrive to support your project.