which parameter to change to run demo for n epochs?

[muthiyanbhushan]

Hello,

I had gone through all of the code. I could not get how to change the #epochs or #iter parameter and also #threads parameter like in Google-C code.

Can you give little glimpse on it.

I found n_workers parameter but by changing that parameter I could not get much more timing difference. Changing n_workers from 4 to 16 only improved 5 sec of training time.

Please let me know.

Thanks.

[jdeng]

IIRC the code relies on OpenMP for parallel training so the n_workers param is probably not useful. If you compile with OpenMP support you should be able to use all the CPUs. You can take a look at the train() function to add extra loops to train more iterations.

[muthiyanbhushan]

Hello @jdeng, Thanks I changed the code for more number of iterations by adding a for loop. Now there are 3 nested for loops in the train function. I have few questions:

1. Does one iteration means training over whole 17 Million words in text8 corpus?
2. Also, wanted to know what batch size is being used in the code.
3. Was your original github code training for just 1 iterations ?
4. Does the concept #iteraions in original word2vec code same as this one?

https://github.com/svn2github/word2vec

Please, let me know.

Thanks.