Closed Penguin0321 closed 11 months ago
here is the output: ############################################################################################### Generation 0:00:00 - On average: 00:00:00 ± 00:00:00 - Total: 00:00:00 - To completion: Unknown ############################################################################################### audi_checkpoint: starting generation 1/500 Offsprings generated: 0/300 (816 s, 0.0 /s)
That's unusual. I'll have a look into this. Did you use the default configuration of the notebook or changed it? Also how many CPU cores did you give the training?
On Thu, 9 Nov 2023, 06:34 Probe, @.***> wrote:
I ran all 3 ipynb examples but they jamed in the last step. When I set verbose=2 in sr.fit() to get more details, it seems the first offspring is not generated even minutes later. Do you have any ideas?
— Reply to this email directly, view it on GitHub https://github.com/davideferrari92/multiobjective_symbolic_regression/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQIFDKTH5YDWG5YQM52OYLYDR2OBAVCNFSM6AAAAAA7EBEC3OVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE4DIOJSHE4DGNI . You are receiving this because you are subscribed to this thread.Message ID: @.*** com>
I didnt change anything in each ipynb file (except verbose) or jupyter-lab. Does the parameter "n_jobs" stand for CPU cores I give? I left it as -1. When the first offspring is generating, the CPU usage on task manager is nearly 0.
Can you give me more info about the Python version and the packages you have installed?
In general, try to execute creating a new python environment just for this test because I also have problems with environments that I used for other applications. When I create a fresh one with python -m venv venv
and then pip install -r requirements.txt
when it worked fine.
Also make sure to have pulled the last version of the repository (I made changes about this type of error a few weeks ago and maybe it's just the matter of getting the updated version).
I downloaded the latest code and created python environment with anaconda. The python version is 3.8.18, and the details about packages are in the attach file. requirements_Penguin.txt. However, the first offspring still cannot be generated in each ipynb example.
Sorry to hear that. Happy to look into that personally. Drop me an email at @.*** so we can schedule a f2f meeting and try to fix it.
On Sat, 11 Nov 2023, 17:38 Probe, @.***> wrote:
I downloaded the latest code and created python environment with anaconda. The python version is 3.8.18, and the details about packages are in the attach file. requirements_Penguin.txt https://github.com/davideferrari92/multiobjective_symbolic_regression/files/13326057/requirements_Penguin.txt . However, the first offspring still cannot be generated in each ipynb example.
— Reply to this email directly, view it on GitHub https://github.com/davideferrari92/multiobjective_symbolic_regression/issues/1#issuecomment-1806873891, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQIFDIIWTSI2N4WYL4ONT3YD6ZZXAVCNFSM6AAAAAA7EBEC3OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMBWHA3TGOBZGE . You are receiving this because you commented.Message ID: <davideferrari92/multiobjective_symbolic_regression/issues/1/1806873891@ github.com>
I'm very grateful for your enthusiasm. However, I have fixed this issue after I thoroughly checked the error messages in the console (These messages are not output on web of Jupyter Notebook). I guess you develop this project on Mac or Linux, but I ran the ipynb examples on Windows. Since the signal mechanism is not fully supported in Windows kernel, "signal.SIGALRM" and "signal.alarm" give error in "./symbolic_regression/SymbolicRegressor.py". I have replaced "signal.SIGALRM" with "signal.SIGABRT", and "signal.alarm" with "time.sleep". The offsprings can be generated now, though ETA is more than a day. Thanks again for your help.
Ah great. Yes, I develop on Linux and will fix that. Thanks for the debug.
The ETA depends strongly on the jobs you allow and how complex the expressions get. I recommend starting with few generations to explore. Unfortunately, for to the nature of the algorithm, it's a slow process and may end up in long trainings. So far there's nothing much I could do on my end but will continue to improve. Also even I am discovering the limitations of the approach and so there may be inefficiencies that have to be fixed.
Do let me know if I can support your further and feel free to report any issues.
On Mon, 13 Nov 2023, 05:24 Probe, @.***> wrote:
I'm very grateful for your enthusiasm. However, I have fixed this issue after I thoroughly checked the error messages in the console (These messages are not output on web of Jupyter Notebook). I guess you develop this project on Mac or Linux, but I ran the ipynb examples on Windows. Since the signal mechanism is not fully supported in Windows kernel, "signal.SIGALRM" and "signal.alarm" give error in "./symbolic_regression/SymbolicRegressor.py". I have replaced "signal.SIGALRM" with "signal.SIGABRT", and "signal.alarm" with "time.sleep". The offsprings can be generated now, though ETA is more than a day. Thanks again for your help.
— Reply to this email directly, view it on GitHub https://github.com/davideferrari92/multiobjective_symbolic_regression/issues/1#issuecomment-1807497242, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQIFDJZRVXZ5TFPR3P7GY3YEGVKNAVCNFSM6AAAAAA7EBEC3OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMBXGQ4TOMRUGI . You are receiving this because you commented.Message ID: <davideferrari92/multiobjective_symbolic_regression/issues/1/1807497242@ github.com>
Now the code should be fixed to run properly on Windows too, please consider downloading the latest version. Do let me know if there's anything else.
I ran all 3 ipynb examples but they jamed in the last step. When I set verbose=2 in sr.fit() to get more details, it seems the first offspring is not generated even minutes later. Do you have any ideas?