MemoryError

[teknogeek]

I seem to be getting a MemoryError when trying to generate a story and im not quite sure how to fix it.

Traceback (most recent call last):
  File "D:/Users/teknogeek/Documents/Code/NeuralStoryteller/test.py", line 3, in <module>
    z = generate.load_all()
  File "D:\Users\teknogeek\Documents\Code\NeuralStoryteller\generate.py", line 105, in load_all
    stv = skipthoughts.load_model(path_to_skmodels, path_to_sktables)
  File "D:\Users\teknogeek\Documents\Code\NeuralStoryteller\skipthoughts.py", line 63, in load_model
    utable, btable = load_tables(path_to_tables)
  File "D:\Users\teknogeek\Documents\Code\NeuralStoryteller\skipthoughts.py", line 81, in load_tables
    utable = numpy.load(path_to_tables + 'utable.npy')
  File "D:\Python27\lib\site-packages\numpy\lib\npyio.py", line 406, in load
    pickle_kwargs=pickle_kwargs)
  File "D:\Python27\lib\site-packages\numpy\lib\format.py", line 638, in read_array
    array = pickle.load(fp, **pickle_kwargs)
MemoryError

The code that i am using to get this is

import generate, os

z = generate.load_all()
generate.story(z, os.path.join("images", "ex1.jpg"))

Any help?

[ryankiros]

How much memory do you have available?

This model requires loading 2 large .npy files (utable and btable) both of which are > 2GB. These consist of almost 1 million word embeddings (each of which are 620 dimensional, single precision floats) as part of the skip-thoughts model.

Given that these embeddings are only used for MS COCO training sentences, it's possible I can create a reduced vocabulary version, where only ~30K words are necessary instead of 1M. I'll look into this.

[teknogeek]

I will test this now. Currently I am starting the test with 8GB of RAM free

[teknogeek]

still failed with a MemoryError after only getting to 9.4GB used of 16

[ryankiros]

I haven't done anything different yet. I'll update once I have, probably next week (I've got a paper deadline this week)

[curryli]

Hi, ryankiros, We have also met the memmory error:

In [2]: z = generate.load_all()
/data/romance.npz
Loading skip-thoughts...
/usr/lib/python2.7/site-packages/theano/scan_module/scan.py:1019: Warning: In the strict mode, all neccessary shared variables must be passed as a part of non_sequences
'must be passed as a part of non_sequences', Warning)
Loading decoder...
Problem occurred during compilation with the command line below:
/usr/bin/g++ -shared -g -O3 -fno-math-errno -Wno-unused-label -Wno-unused-variable -Wno-write-strings -Wl,-rpath,/usr/lib64 -march=nocona -mcx16 -msahf -mno-movbe -maes -mno-pclmul -mpopcnt -mno-abm -mno-lwp -mno-fma -mno-fma4 -mno-xop -mno-bmi -mno-bmi2 -mno-tbm -mno-avx -mno-avx2 -msse4.2 -msse4.1 -mno-lzcnt -mno-rtm -mno-hle -mno-rdrnd -mno-f16c -mno-fsgsbase -mno-rdseed -mno-prfchw -mno-adx -mfxsr -mno-xsave -mno-xsaveopt --param l1-cache-size=32 --param l1-cache-line-size=64 --param l2-cache-size=20480 -mtune=nocona -D NPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION -m64 -fPIC -I/usr/lib64/python2.7/site-packages/numpy/core/include -I/usr/include/python2.7 -I/usr/lib/python2.7/site-packages/theano/gof -fvisibility=hidden -o /root/.theano/compiledir_Linux-3.10-el7.x86_64-x86_64-with-centos-7.0.1406-Core-x86_64-2.7.5-64/tmpSPHNG7/de61f4c333d06430fc7dc014bf04eb25.so /root/.theano/compiledir_Linux-3.10-el7.x86_64-x86_64-with-centos-7.0.1406-Core-x86_64-2.7.5-64/tmpSPHNG7/mod.cpp -L/usr/lib64 -L/usr/lib64 -lpython2.7 -lopenblas
ERROR (theano.gof.cmodule): [Errno 12] Cannot allocate memory

We are running on a machine with 8Gb, but it failed.
I'm wondering how many memory do we need?

Thanks.