Fine Tune Torch-rnn for very low number of characters and long chain of charaters

[areiniger]

Hi,
I would like to run a long chain of characters around 3000000 with a very few number of characters into the RNN ... Any help how to fine tune the code?
Thank ...

[ChrisCummins]

I'm sorry I don't understand the issue here. Is the behavior you're after not supported?

[areiniger]

Hi,It is not an "issue" per se ... I will need your recommendation how to improve the results in a very specific environment ... As I explained I have a large file as input around 30 millions characters with a very small diversity of characters forming very large "patterns"/words (maximum 6000 characters per pattern/word) and the results I am obtaining are very low (loss around 0.95). Sent from Nylas Mail, the best free email app for work On May 21 2017, at 4:09 pm, Chris Cummins <notifications@github.com> wrote: I'm sorry I don't understand the issue here. Is the behavior you're after not supported? —You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or mute the thread.