John S. Dvorak comments

Results 9 comments of


                                            John S. Dvorak

TF model doesn't stay in memory with --save-every > 1

What is the elegant solution that someone might helpfully offer as a pull request?

TF model doesn't stay in memory with --save-every > 1

No problem! You are doing all the lifting, I was just curious if I could help. On Wed, Jun 7, 2017 at 9:44 AM Arthur Goldberg wrote: > Hey @JDvorak...

TF model doesn't stay in memory with --save-every > 1

Isn't the idiomatic meaning that an epoch is a complete lap through the training set? Do you really need to load everything into memory or just the immediate next minibatch?...

TF model doesn't stay in memory with --save-every > 1

It might be premature optimization, but if you want to engage in such premature designs: consider that there is a processing time for some n images likely greater than the...

makes middlewares pluggable. i.e for custom oauth flows or analytics etc.

Recommendation: Speeding up non-faiss distance matrix calculatios

Is there a similar way to batch out the combine_patches call?

Allow `DEFERRABLE` or `modifyEnd/modifyStart` to constraints

Is there a way to hack around this in the meanwhile using raw sql within a transaction?

Words Per Line

You can read more than one word at a time without moving your eye. There is little difference in eye movement between reading the word 'distinction' and 'cats eat mice'...