StanfordVL / GibsonEnv

Gibson Environments: Real-World Perception for Embodied Agents
http://gibsonenv.stanford.edu/
MIT License
875 stars 146 forks source link

An Issue about Environment on my Computer #54

Closed Berk035 closed 6 years ago

Berk035 commented 6 years ago

Hello everyone,

I am a novice ubuntu and github user. Nowadays, I am studying Master Degree in EEE at my university. So, I want to use GibsonEnv. for my final thesis. I installed packages and requirements according to the instructions. Then, I could work first example. Unfortunately, Other examples which consist camera and RGB don't work and are closed themselves. I've read some issues about cuda render or something else.

First example works as following; working_1

Whereas other examples have problem.. And the output is given below;

`(py35) deepsrv@deepsrv-System-Product-Name:~/GibsonEnv$ python examples/demo/play_husky_camera.py Error: cuda renderer is not loaded, rendering will not work pybullet build time: Oct 17 2018 10:24:42 pygame 1.9.4 Hello from the pygame community. https://www.pygame.org/contribute.html /home/deepsrv/GibsonEnv/examples/demo/../configs/play/play_husky_camera.yaml startThreads creating 1 threads. starting thread 0 started thread 0 argc=2 argv[0] = --unused argv[1] = --start_demo_name=Physics Server ExampleBrowserThreadFunc started X11 functions dynamically loaded using dlopen/dlsym OK! Creating context Created GL 3.0 context Direct GLX rendering context obtained Making context current GL_VENDOR=NVIDIA Corporation GL_RENDERER=GeForce GTX 780/PCIe/SSE2 GL_VERSION=3.2.0 NVIDIA 410.48 GL_SHADING_LANGUAGE_VERSION=1.50 NVIDIA via Cg compiler pthread_getconcurrency()=0 Version = 3.2.0 NVIDIA 410.48 Vendor = NVIDIA Corporation Renderer = GeForce GTX 780/PCIe/SSE2 b3Printf: Selected demo: Physics Server startThreads creating 1 threads. starting thread 0 started thread 0 MotionThreadFunc thread started ven = NVIDIA Corporation killing None Error in sys.excepthook: Traceback (most recent call last): File "/home/deepsrv/GibsonEnv/gibson/envs/env_modalities.py", line 642, in camera_multi_excepthook self.r_camera_mul.terminate() AttributeError: 'NoneType' object has no attribute 'terminate'

Original exception was: Traceback (most recent call last): File "examples/demo/play_husky_camera.py", line 17, in env = HuskyNavigateEnv(config=args.config, gpu_idx = args.gpu) File "/home/deepsrv/GibsonEnv/gibson/envs/husky_env.py", line 40, in init self.robot_introduce(Husky(self.config, env=self)) File "/home/deepsrv/GibsonEnv/gibson/envs/env_modalities.py", line 349, in robot_introduce self.setup_rendering_camera() File "/home/deepsrv/GibsonEnv/gibson/envs/env_modalities.py", line 375, in setup_rendering_camera self.setup_camera_multi() File "/home/deepsrv/GibsonEnv/gibson/envs/env_modalities.py", line 671, in setup_camera_multi self.r_camera_mul = subprocess.Popen(shlex.split(render_main), shell=False) File "/home/deepsrv/anaconda3/envs/py35/lib/python3.5/subprocess.py", line 676, in init restore_signals, start_new_session) File "/home/deepsrv/anaconda3/envs/py35/lib/python3.5/subprocess.py", line 1289, in _execute_child raise child_exception_type(errno_num, err_msg) FileNotFoundError: [Errno 2] No such file or directory: './depth_render' numActiveThreads = 0 stopping threads stopThreads: Thread 0 used: 1 Thread with taskId 0 exiting Thread TERMINATED destroy semaphore semaphore destroyed destroy main semaphore main semaphore destroyed finished numActiveThreads = 0 btShutDownExampleBrowser stopping threads stopThreads: Thread 0 used: 1 Thread with taskId 0 exiting Thread TERMINATED destroy semaphore semaphore destroyed destroy main semaphore main semaphore destroyed Error in atexit._run_exitfuncs: Traceback (most recent call last): File "/home/deepsrv/anaconda3/envs/py35/lib/python3.5/site-packages/gym/utils/closer.py", line 67, in close closeable.close() File "/home/deepsrv/GibsonEnv/gibson/envs/env_modalities.py", line 487, in _close self.r_camera_mul.terminate() AttributeError: 'NoneType' object has no attribute 'terminate' `

fxia22 commented 6 years ago

Looks like you are compiling from source, but it is not successfully compiled. Can you show the output of build script? (the output of ./build.sh build_local)

Berk035 commented 6 years ago

Yes, I built it from source. I send the command which you said me and the output is attached below. error_output.txt

hzyjerry commented 6 years ago

Hi @Berk035! Are you using ubuntu 18.04?

It looks like your GLM/GLX library is not building correctly, which has been similarly reported by other developers on 18.04.

Here is a quick fix (original) by adding 1 line to GibsonEnv/gibson/core/channels/CMakeLists.txt

add_definitions(-D GLM_ENABLE_EXPERIMENTAL)

Berk035 commented 6 years ago

Hi @Berk035! Are you using ubuntu 18.04?

It looks like your GLM/GLX library is not building correctly, which has been similarly reported by other developers on 18.04.

Here is a quick fix (original) by adding 1 line to GibsonEnv/gibson/core/channels/CMakeLists.txt

add_definitions(-D GLM_ENABLE_EXPERIMENTAL)

Hi @hzyjerry, yes I am using Ubuntu 18.04. Thank you for your advice, I'll try it and give a feedback.

Berk035 commented 6 years ago

@hzyjerry Hi again, I have tried the code now and there is still same problem. Is there any different ideas? error output is given below. error.txt

fxia22 commented 6 years ago

Looks like the compilation of depth_renderer is still not successful. Can you show the log of compilation?

Berk035 commented 6 years ago

@fxia22 Hi again, Sorry for my late response. I viewed that the log file syslog and observed the last line as: "Nov 1 09:59:08 deepsrv-System-Product-Name gnome-shell[2498]: setup_framebuffers: assertion 'width > 0' failed" Also, I observed the log of compilation from dmesg as fallows: dmesg_output.txt

fxia22 commented 6 years ago

@Berk035 sorry for the confusion, by log of compilation I mean the output of ./build.sh build_local

Berk035 commented 6 years ago

@fxia22 I've complied and attached new output file below now: output_2.txt

Berk035 commented 6 years ago

Is there any solution for now? Did you find any opportunity for look up the problem? I want to develop my programme on GibsonEnv. If there is no solution for Ubuntu 18.04, I will try it on Ubuntu 16.

fxia22 commented 6 years ago

Sorry for the delay, looks like depth_render compilation was successful, however it is not in GibsonEnv/gibson/core/channels/depth_render, can you show ls GibsonEnv/gibson/core/channels/depth_render? Did you run pip install -e . after the compilation?

Berk035 commented 6 years ago

Hello again, Yes I ran pip install -e. after installation and saw Gibson is installed succesfully. Also you can see files in directory below: render

Thanks for attention.

hzyjerry commented 6 years ago

@Berk035 hi, it looks like your depth_render is successfully compiled.

If your Gibson rendering is blacked out, I think this could be possibly due to your CUDA version (it looks like you're using CUDA 10?)

I recommend using CUDA 8 since it's well tested with Gibson. CUDA 7.5 is the version on my computer and it also works.

Berk035 commented 6 years ago

@hzyjerry hi, I suppose that I would downgrade my drivers and operating system for working. Thanks for advice.

hzyjerry commented 6 years ago

I just realize that CUDA 8 and below are not supported on Ubuntu 18.04. I think installing CUDA 9.0 is worth trying if you'd like to avoid reinstalling the OS.

Berk035 commented 6 years ago

@hzyjerry Thank you for your respond sir, I will try somethings to figure out. Also, you predicted true about my cuda version which is CUDA 10 and it doesn't work properly. I reinstalled OS to another computer as Ubuntu 16.04 without any upgrades and I hope it will work this time. Thank you all for your concern again.

hzyjerry commented 6 years ago

Sounds good! Let us know how it goes

Berk035 commented 6 years ago

@hzyjerry Hi again, I was able to run the demo with camera on Ubuntu 16.04 and CUDA 7.5 Now, I will develop my learning algorithm as next step but fps values are very low and process stucks sometimes. Is there any advice? Thank you for your helps. fps

fxia22 commented 6 years ago

@Berk035 Are you using a GTX 780? It makes sense that you get such frame rate, because Gibson is designed for modern GPUs, like 9/10 series and needs a lot of computation power to run(each of our frames goes through a neural network). If you think your frame rate is too low, you can reduce the resolution to 128x128 and design your learning algorithm based on that.

Berk035 commented 6 years ago

@fxia22 I use the GTX1050 now. But, the other machine uses GTX780 double graphic cards. I tried the programme on both of them and changing some parameters for optimal work. I will change resolution for much FPS. Thank you for everything again.

zubair-irshad commented 4 years ago

@Berk035 Did you solve the low frame rate issue? I am getting the same issue on my end

Berk035 commented 4 years ago

@zubair-irshad Yes, it is solved. Please check your graphics options and manage your timestep&frame_skip relations in config file.

zubair-irshad commented 4 years ago

Awesome. I am curious to know, are there any optimum graphic options settings?