georgeliu233 / scene-rep-transformer Goto Github PK

How much video memory is needed for training?

Hi, I was wondering, what was the dataset you used to train Transformer?

How should I train? I don't see the corresponding train.py

I am trying to reproduce your experiments but I am facing some difficulties.

• I am not able to build the scenarios.
• I have run the code using the scenario provided by smarts (previously built) and I get the following error: "FileNotFoundError: [Errno 2] No such file or directory: './test_results/rb_left_turn_new_test.json'"
  It looks like some files are missing in the repository.
• When I fix this by creating the missing files I get a new error.

Maybe I am doing something wrong or maybe the repository is not complete yet.

Please let me know if you can help me to run your experiments.

Thank you.

Traceback (most recent call last):
File "test.py", line 142, in
testing(args)
File "test.py", line 131, in testing
runner.evaluate_policy_continuously()
File "../envs/runners/off_policy_trainer.py", line 610, in evaluate_policy_continuously
res = self.evaluate_policy(total_steps=0,plot_map_mode=plot_map_mode)
File "../envs/runners/off_policy_trainer.py", line 643, in evaluate_policy
obs,ego,map_s = self.neighbor_obs(obs,test=True)
File "../envs/runners/off_policy_trainer.py", line 190, in neighbor_obs
n_obs = self.process_neighbors(neighbors_id,test)
File "../envs/runners/off_policy_trainer.py", line 179, in process_neighbors
self._test_smarts.attach_sensors_to_vehicles(self.test_neighbor_spec, ids)
File "/home/hwb/miniconda3/envs/smart/lib/python3.7/site-packages/smarts/core/smarts.py", line 434, in attach_sensors_to_vehicles
self, agent_spec.interface, vehicle_ids
AttributeError: 'AgentInterface' object has no attribute 'interface'

[Errno 2] No such file or directory: '../envs/smarts/smarts_scenarios/left_turn'

Hello, I'd like to inquire about the required Python version and operating system to run your code. While attempting to install the listed requirements, I encountered errors related to dependencies and libraries.

Unfortunately, my level of competence is limited, I tried for a long time to write a training file, but failed, and I am very eager to train the results, so it is very presumptuous to ask you, how did you train at that time, since you have a trained model, you should have a training file, right?

Hi, I'm using Ubuntu 20.04 and Python 3.8.10. I have tried many times but still can't succeed in doing 'pip -r requirements.txt'. Could you please post your environment information?
1.before split requirements into SMARTS and Scene-Rep-Transformer

(venv) de@pc# pip install  -r requirements.txt
ERROR: Double requirement given: matplotlib==3.4.2 (from -r requirements.txt (line 97)) (already in matplotlib==3.4.3 (from -r requirements.txt (line 4)), name='matplotlib')

2.split requirements ,just Scene-Rep-Transformer

(venv) de@pc# pip install  -r requirements.txt
ERROR: imageio 2.34.1 has requirement pillow>=8.3.2, but you'll have pillow 8.2.0 which is incompatible.
ERROR: scikit-image 0.21.0 has requirement pillow>=9.0.1, but you'll have pillow 8.2.0 which is incompatible.
ERROR: rich 13.7.1 has requirement typing-extensions<5.0,>=4.0.0; python_version < "3.9", but you'll have typing-extensions 3.7.4.3 which is incompatible.
ERROR: twisted 24.3.0 has requirement typing-extensions>=4.2.0, but you'll have typing-extensions 3.7.4.3 which is incompatible.
......
ERROR: tensorflow-gpu 2.4.0 has requirement numpy~=1.19.2, but you'll have numpy 1.24.4 which is incompatible.
ERROR: tf2rl 1.1.5 has requirement tensorflow-addons~=0.13.0, but you'll have tensorflow-addons 0.11.2 which is incompatible.

after I solve problem above,then

(venv) de@pc# cd tools;python test.py  --scenario=left_turn  --algo=scene_rep
ModuleNotFoundError: No module named 'cv2'

I know it means I need to install opencv-python and there are many other dependent packages that should be installed. I tried to solve those packages one by one many times and failed. Could you help me? Thanks!