GrapChi: Rating2 Under Toolkit

User 199 | 4/4/2014, 3:35:35 PM

Hi, how do we use ./toolkits/collaborative_filtering/rating2 - I get the error message

rating2: ../../src/util/cmdopts.hpp:67: std::string graphchi::getconfigoption_string(const char*): Assertion `false' failed. ERROR: could not find option algorithm from config.Aborted (core dumped)

I am able to use Lanczos SVD without any problem. Do I need extra configuration for rating2?

Comments

User 6 | 4/4/2014, 3:37:53 PM

Please send us the full command line you used. This error says you did not specify the algorithm type (using --algorithm=svd)


User 199 | 4/4/2014, 3:45:40 PM

OK, that option fixed one problem and now I have another:

Full cmd line

/toolkits/collaborativefiltering/rating --algorithm=svd --training=smallnetflixmm .. ... === REPORT FOR sharder() === [Timings] edataflush: 0.121151s (count: 26, min: 0.000802s, max: 0.004877, avg: 0.00465965s) executesharding: 0.315104 s finishshard.sort: 0.068732 s preprocessing: 0.668263 s shardfinal: 0.23095 s [Other] app: sharder INFO: sharder.hpp(done:903): Created 1 shards, for 3298163 edgesSuccessfully finished sharding for smallnetflixmm INFO: io.hpp(convertmatrixmarket:592): Created 1 shards. fopen failed: No such file or directory FATAL: io.hpp(openfile:36): Failed to open filesmallnetflixmm_U.mm terminate called after throwing an instance of 'char const*'

The command creates all SVD output, it cannot move on to the next step. I am not sure why it tries to open "filesmallnetflixmmU.mm" instead of "smallnetflixmmU.mm".


User 6 | 4/4/2014, 7:07:23 PM

The output message is garbled. The rating application does not find the output of svd. Please send us the full command for running svd.