Problem with binary format

[Soma Juhász]

Hello,

I'm trying to predict the Ml-10M dataset. I augmented the rows with the implicit feedback and wrote it to a .txt in the following form:

3 20812:1 7018:1 3254:1 804:3.1814e-02 1594:3.1814e-02 3426:3.1814e-02...

Then I used the tools to convert it to binary format. I have the .x, .y, .xt files for train and test. When I try to use the libFM it fails with no error. It just finished work very fast and creates no output text. When I fiddle around with cache size (i.e. the command below) it writes additional info but usually it is not the case.

libFM -task r -train TimeSVDpp_train -test TimeSVDpp_test -dim ’1,1,8’ -cache_size 100000000 -verbosity 1 -iter 56 -out out_test.txt

Loading train...

has x = 0

has xt = 1

data transpose... num entries in cache=12499895 num rows in cache=209

num_cases=9000048       num_values=4268116901   num_features=71568      min_target=0.5  max_target=5

Loading test...

has x = 0

has xt = 1

data transpose... num entries in cache=12499058 num rows in cache=1884

num_cases=1000006       num_values=474691623    num_features=71568      min_target=0.5  max_target=5

#relations: 0

Loading meta data...

#attr=71568     #groups=1

#attr_in_group[0]=71568

num_attributes=71568

use w0=0

use w1=1

dim v =8

reg_w0=0

reg_w=0

reg_v=0

init ~ N(0,0.1)

task=0

min_target=0.5

max_target=5

do_multilevel=1

do_sampling=1

num_eval_cases=1000006

I have 16 Gb RAM and I use Win10.

Thank you in advance.