search

timeseriesAI / tsai

Time series Timeseries Deep Learning Machine Learning Python Pytorch fastai | State-of-the-art Deep Learning library for Time Series and Sequences in Pytorch / fastai
https://timeseriesai.github.io/tsai/
Apache License 2.0
4.95k stars 625 forks source link

Endlesss OpenBLAS Warning #735

Closed weizhenhit closed 1 year ago

weizhenhit commented 1 year ago

When I try to run the training function clf.fit_one_cycle(num_epoch, lr), endless

"OpenBLAS Warning : Detect OpenMP Loop and this application may hang. Please rebuild the library with USE_OPENMP=1 option. "

would be printed. the same problem with the function clf.lr_find() and clf.plot_metrics(). Is there any solution? Thanks!

oguiza commented 1 year ago

Hi @1180300318, With the information you have provided, I'm afraid I won't be able to help. Could you run my_setup()and paste the output here? I have not seen that issue before. But it seems you may need to rebuild OpenBLAS as suggested here.

weizhenhit commented 1 year ago

The outputs of my_setup() are as follows:

os : Linux-5.4.0-144-generic-x86_64-with-glibc2.31 python : 3.9.16 tsai : 0.3.5 fastai : 2.7.11 fastcore : 1.5.29 torch : 1.12.1 cpu cores : 44 threads per cpu : 2 RAM : 157.26 GB GPU memory : [11.77, 11.77] GB

I tried to rebuild OpenBLAS as suggested, but I can't find the OpenBLAS folder in tsai. Or should I install it myself?

oguiza commented 1 year ago

OpenBLAS is not part of tsai. It comes with your operating system. I don't know what's causing this warning. Maybe others can help?

weizhenhit commented 1 year ago

@oguiza Ok, thanks for your reply. I will try some other solutions.

oguiza commented 1 year ago

Ok, @1180300318. I'm sorry I cannot provide any help on this. I will leave the issue open for a while to see if you or anybody else comes up with a solution.

weizhenhit commented 1 year ago

I turn to run the code on the google colab rather than on my lab server, and it runs properly now! So there may be a problem with the configuration of the environment on the server.