search

sisl / ngsim_env

Learning human driver models from NGSIM data with imitation learning.
https://arxiv.org/abs/1803.01044
MIT License
172 stars 80 forks source link

Do some package have something build in your own system? Like a building error showing your path. #20

Open Ericonaldo opened 5 years ago

Ericonaldo commented 5 years ago

The setting are strictly follows your guide and the system is ubuntu16.04. The errors are as follows:

(rllab3) xxx@XXX:~/ngsim_env/scripts/imitation$ python validate.py --n_proc 1 --exp_dir ../../data/experiments/NGSIM-gail/ --params_filename itr_1000.npz --random_seed 42 50 vehicles with H = 200 2019-07-03 10:57:27.242208: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA pid: 0 traj: 58 / 2062 signal (11): Segmentation fault in expression starting at no file:0 GetResult at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Modules/_ctypes/callproc.c:911 [inlined] _ctypes_callproc at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Modules/_ctypes/callproc.c:1186 PyCFuncPtr_call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Modules/_ctypes/_ctypes.c:3856 PyObject_Call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/abstract.c:2165 do_call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4936 [inlined] call_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4732 [inlined] PyEval_EvalFrameEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:3236 fast_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4803 [inlined] call_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4730 [inlined] PyEval_EvalFrameEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:3236 fast_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4803 [inlined] call_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4730 [inlined] PyEval_EvalFrameEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:3236 fast_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4803 [inlined] call_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4730 [inlined] PyEval_EvalFrameEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:3236 fast_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4803 [inlined] call_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4730 [inlined] PyEval_EvalFrameEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:3236 _PyEval_EvalCodeWithName at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4018 PyEval_EvalCodeEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4039 function_call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/funcobject.c:627 PyObject_Call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/abstract.c:2165 method_call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/classobject.c:330 PyObject_Call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/abstract.c:2165 PyObject_CallFunctionObjArgs at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/abstract.c:2445 call_attribute at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/typeobject.c:6089 [inlined] slot_tp_getattr_hook at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/typeobject.c:6131 builtin_getattr at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/bltinmodule.c:998 PyCFunction_Call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/methodobject.c:109 call_function at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4705 [inlined] PyEval_EvalFrameEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:3236 _PyEval_EvalCodeWithName at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4018 PyEval_EvalCodeEx at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Python/ceval.c:4039 function_call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/funcobject.c:627 PyObject_Call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/abstract.c:2165 method_call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/classobject.c:330 PyObject_Call at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/abstract.c:2165 PyObject_CallFunctionObjArgs at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/abstract.c:2445 call_attribute at /home/ilan/minonda/conda-bld/work/Python-3.5.2/Objects/typeobject.c:6089 [inlined]

ZhilaiShen commented 5 years ago

Hi Ericonaldo, I am actually having the exact same problem as yours. Have you been able to solve this problem?

nebneBgnahZ commented 5 years ago

Hi Ericonaldo, I am actually having the exact same problem as yours. Have you been able to solve this problem?

same issue here.

Ericonaldo commented 5 years ago

No... But maybe it is not the path problem. I think it is the error that the authors had mentioned here, which says "this is somehow related to julia processes remaining unfinished and the python script moving on. Looking in validate.py, there is a sleep() call"

ZhilaiShen commented 5 years ago

And I think he also mentions the Seg fault, which is solved by deleting the PyCall.jl, but for julia1.1, there is no longer such folder. Another issue is that the issue only occurs when I use my own dataset, while the original NGSIM dataset is still working. Are you using the original dataset or your own dataset?

Ericonaldo commented 5 years ago

Did you work well with the NGSIM dataset that the authors packaged up? I encountered this error when I use the packaged NGSIM dataset.

nebneBgnahZ commented 5 years ago

Did you work well with the NGSIM dataset that the authors packaged up? I encountered this error when I use the packaged NGSIM dataset.

I also encountered this error using the NGSIM dataset. I am wondering if anything was wrong when I created the observations. In addition, I also encountered error when I run the bash script "run_experiment.sh". Apparently all the models for multi-agent failed. But I do have some results for single-agent scenario.

ZhilaiShen commented 5 years ago

When I try to run it with packaged NGSIM data, the only problem I encountered is that n_env from 10 to 20, the Tensorflow alerts because the parameters of layers are not cleared, and not allowed. BTW, do you guys have this problem occurring "randomly"?

ZhilaiShen commented 5 years ago

Hi all I am not sure if this works for all, but I think the problem can be solved by removing the cache file under folder "~/julia/.julia/compiled/v1.1/PyCall/****.jl". It can be the same problem here, under section trouble shooting, reason 1. The cache file path for julia 1.1 is different from path in julia 0.6, so I delete the new cache file, and it seems to be working.

Ericonaldo commented 5 years ago

Hi all I am not sure if this works for all, but I think the problem can be solved by removing the cache file under folder "~/julia/.julia/compiled/v1.1/PyCall/****.jl". It can be the same problem here, under section trouble shooting, reason 1. The cache file path for julia 1.1 is different from path in julia 0.6, so I delete the new cache file, and it seems to be working.

Hi, ZhilaiShen You can work the whole thing now?

ZhilaiShen commented 5 years ago

sorry Minghuan, not really

From: Minghuan Liu notifications@github.com Sent: Sunday, August 18, 2019 8:10:01 AM To: sisl/ngsim_env Cc: Shen, Zhilai (Z.); Comment Subject: Re: [sisl/ngsim_env] Do some package have something build in your own system? Like a building error showing your path. (#20)

Hi all I am not sure if this works for all, but I think the problem can be solved by removing the cache file under folder "~/julia/.julia/compiled/v1.1/PyCall/****.jl". It can be the same problem herehttps://github.com/sisl/ngsim_env/blob/master/docs/install.md, under section trouble shooting, reason 1. The cache file path for julia 1.1 is different from path in julia 0.6, so I delete the new cache file, and it seems to be working.

Hi, ZhilaiShen You can work the whole thing now?

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/sisl/ngsim_env/issues/20?email_source=notifications&email_token=AMI4ORP4CCI73BAZ2FAF5N3QFFQ4TA5CNFSM4H5BP32KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4RB6QI#issuecomment-522329921, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AMI4OROBJTK4IFSPGLGJLVLQFFQ4TANCNFSM4H5BP32A.

raks097 commented 4 years ago

@Ericonaldo @ZhilaiShen @nebneBgnahZ I am currently facing the same issue, Were any of you able to get this working?

Ericonaldo commented 4 years ago

No. Actually I already gave it up ...

DarrenRuan commented 4 years ago

Have you tried to use single_process_collect_trajectories rather than parallel_collect_trajectories? @Ericonaldo @raks097 @ZhilaiShen @nebneBgnahZ Then we might be able to skip the part (line 190 in validate.py):

# let the julia processes finish up
    time.sleep(10)

Although it might be much slower.

Edit: Well. Although I choose to use single_process_collect_trajectories, I still met that weird Segmentation fault.