Create_trainset error

Question

Create_trainset error

Chengaowei opened this issue 3 years ago · comments

Chengaowei commented 3 years ago

Hi,
Thanks for your work. It's excellent!
I run the demo.py file. And got an error at --'create_trainset'.

It says: 'ValueError: need at least one array to concatenate'
could you tell me how to fix it？

K L · Answer 1 · Thu Mar 25 2021 03:59:04 GMT+0800 (China Standard Time)

Hi,

I think this error appears because your config.yaml has no Data specified. Did you follow the VAME workflow guide?

What I see from the first image you attached is that your video folder name is wrong, you miss the video file format ending like .mp4 or .avi. The specification videoType=.mp4 is not adding this to the string. Therefore, I believe your config.yaml is created wrongly and under the header video_sets is no video name listed. Can you confirm this?

If this is the case you need to re-initialize the project. Just delete the current one and create an new one with the right video path string.

I hope this helps!

Chengaowei · Answer 2 · Thu Mar 25 2021 15:11:35 GMT+0800 (China Standard Time)

Thanks for your quickly reply! @KLuxem and sorry for this simple mistake.
you are right! I missed the video file format at end.

I follow the workflow guide with your example video, and now get another error at 'vame.pose_segmentation(config)'

It says: 'Program is Terminated. Because you tried to allocate too many memory regions.'

K L · Answer 3 · Thu Mar 25 2021 16:34:21 GMT+0800 (China Standard Time)

So far I haven't seen this specific problem. Can you set a flag in the pose_segmentation script at line 245 to see if the error occurs after the k-Means calculation? Just type something like print("Flag") so that we understand better where the program breaks for you. Or you can just uncomment line 250

Chengaowei · Answer 4 · Thu Mar 25 2021 17:18:16 GMT+0800 (China Standard Time)

Print-245

Print-250

K L · Answer 5 · Thu Mar 25 2021 17:31:45 GMT+0800 (China Standard Time)

So your error seems related to BLAS. Are you using or have installed anything like Dask?
We could not reproduce this error on two different machines so far. .

Check out this thread:
https://stackoverflow.com/questions/45086246/too-many-memory-regions-error-with-dask

Maybe setting export OMP_NUM_THREADS=1 could help.

Chengaowei · Answer 6 · Thu Mar 25 2021 18:51:19 GMT+0800 (China Standard Time)

I checked VAME envs, and didn't find something like Dask~~

I want to try 'export OMP_NUM_THREADS=1'. where should I do this?

AND I found a page here:
https://github.com/xianyi/OpenBLAS/wiki/faq#program-is-terminated-because-you-tried-to-allocate-too-many-memory-regions

Do you think this can help ?

K L · Answer 7 · Thu Mar 25 2021 19:15:57 GMT+0800 (China Standard Time)

You should be able to type 'export OMP_NUM_THREADS=1' just in your console and try again.

This issue seems to be more machine depended than on the VAME package. Maybe updating the numpy version can help.

What you could still check if is the Kmeans call in function same_parameterization at line 112 is working. Maybe set another flag in line 115 and line 123 to pin down where the error is happening. And another one at line 133. If you think its the Kmeans call for you try out the example from sklearn and see if you run into similar problems

Chengaowei · Answer 8 · Thu Mar 25 2021 22:28:15 GMT+0800 (China Standard Time)

I changed another PC; and the error disappeared!
Anyway, it's working now.

I'll go ahead with your example data.

K L · Answer 9 · Thu Mar 25 2021 22:35:32 GMT+0800 (China Standard Time)

Perfect! I close this issue then!
Enjoy the example ;)