search

Waikato / meka

Multi-label classifiers and evaluation procedures using the Weka machine learning framework.
http://waikato.github.io/meka/
GNU General Public License v3.0
200 stars 76 forks source link

java.lang.OutOfMemoryError: GC overhead limit exceeded #60

Closed haokeliu closed 5 years ago

haokeliu commented 5 years ago

Hello, I am a beginner of Meka. when I use the package meka.classifiers.multilabel.incremental.RTUpdateable，an error has occurred. This is the wrong details： image This is my command line argument： image This is my data (a total of 100million generated by MOA)： image My computer configuration information, it has 8G memory. image This situation also occurs in the BRU and PSU. I would be grateful if I could give me any advice.

fracpete commented 5 years ago

Try to train with less data (if you're using the GUI, it will load all of the data into memory). Start with 100,000 rows. If that works, try 1,000,000. etc.

haokeliu commented 5 years ago

尝试使用较少的数据进行训练（如果您使用的是GUI，它会将所有数据加载到内存中）。从100,000行开始。如果可行，请尝试1,000,000。等等

Thank you for your prompt reply, I will try it out.

haokeliu commented 5 years ago

I changed the data to 100K according to the method you provided, but similar errors still occurred. Can you give me some help? image image

fracpete commented 5 years ago

Are you using batch incremental or prequential as evaluation method in the Explorer? If not, then no incremental training is occurring.

haokeliu commented 5 years ago

Are you using batch incremental or prequential as evaluation method in the Explorer? If not, then no incremental training is occurring.

I used Meavn's local repository before. I used the GUI under your suggestion. The 10k data can run normally, but the data will reach 100million and there will be an error. I will increase the amount of data according to the suggestions you gave. Thank you for your suggestion. This is the wrong details： image