Problems with Pytorch Env

[Leonardodeheinz]

Hello,

i am trying to get this repo to run on my local macbook (m3). I have installed all the neccessary software to get it running. But some questions remain open. Since macos does not really cuda support anymore i wonder if it is essential to get the simulation working. Because encounter to this problem in your tutorial notebook number 3.

The code:

env = GPUDriveTorchEnv(
config=env_config,
scene_config=scene_config,
render_config=render_config,
max_cont_agents=MAX_NUM_OBJECTS, # Maximum number of agents to control per scenario
device=device,
)

the Error :

---

AssertionError Traceback (most recent call last)
Cell In[6], line 2
1 # MAKE ENV
----> 2 env = GPUDriveTorchEnv(
3 config=env_config,
4 scene_config=scene_config,
5 render_config=render_config,
6 max_cont_agents=MAX_NUM_OBJECTS, # Maximum number of agents to control per scenario
7 device=device,
8 )

File ~/Documents/studium/deep_rl/gpu_drive_project/Deep_RL_Project/gpudrive/pygpudrive/env/env_torch.py:61, in GPUDriveTorchEnv.init(self, config, scene_config, max_cont_agents, device, action_type, render_config, backend)
55 self.num_valid_controlled_agents_across_worlds = (
56 self.cont_agent_mask.sum().item()
57 )
59 # Setup action and observation spaces
60 self.observation_space = Box(
---> 61 low=-np.inf, high=np.inf, shape=(self.get_obs().shape[-1],)
62 )
63 self._setup_action_space(action_type)
64 self.info_dim = 5 # Number of info features

File ~/Documents/studium/deep_rl/gpu_drive_project/Deep_RL_Project/gpudrive/pygpudrive/env/env_torch.py:400, in GPUDriveTorchEnv.get_obs(self)
393 def get_obs(self):
...
312 raise AssertionError(
313 "libcudart functions unavailable. It looks like you have a broken build?"
314 )

AssertionError: Torch not compiled with CUDA enabled

The tests that you provided through "pytest" works without any problem. I will try to get it working on a ubuntu workstation with an nvida card next week anyway at my university, since I need for training anyway for my course project.

Best regards