search

danpovey / pocolm

Small language toolkit for creation, interpolation and pruning of ARPA language models
Other
90 stars 48 forks source link

Modify optimize_metaparameters.py to ensure at least a finite distanc… #75

Closed danpovey closed 7 years ago

danpovey commented 7 years ago

…e between discounting values (prevents certain errors later on)

vince62s commented 7 years ago

Dan, Just for my full understanding. in your above fix, you write the parameters with more decimals when very close to the limit. However when you read the parameter and convert string to float don't you think we will lose this precision ?

danpovey commented 7 years ago

I think there should be no loss of precision in reading-- AFAIK python uses doubles throughout and there would be no reason to truncate.

On Fri, Oct 7, 2016 at 4:46 AM, vince62s notifications@github.com wrote:

Dan, Just for my full understanding. in your above fix, you write the parameters with more decimals when very close to the limit. However when you read the parameter and convert string to float don't you think we will lose this precision ?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/danpovey/pocolm/pull/75#issuecomment-252185681, or mute the thread https://github.com/notifications/unsubscribe-auth/ADJVu6iA0SAyg7z8tLdUC_SxndPnmikOks5qxgbsgaJpZM4KPPIk .

vince62s commented 7 years ago

I reran, works fine, you can merge.

danpovey commented 7 years ago

Thanks!

On Sun, Oct 9, 2016 at 2:46 AM, vince62s notifications@github.com wrote:

I reran, works fine, you can merge.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/danpovey/pocolm/pull/75#issuecomment-252468505, or mute the thread https://github.com/notifications/unsubscribe-auth/ADJVu5mokd9mhQeQ4FRz14YTNJ_iXG_wks5qyI3HgaJpZM4KPPIk .