Dzmitry Bahdanau
Dzmitry Bahdanau
https://github.com/mila-udem/blocks/issues/701
Because Blocks-examples and NMT demo have quite different purposes. Blocks-examples is supposed to contain minimalistic toy examples of using Blocks. We would like to keep it small, especially in terms...
This URL is just a placeholder to start the discussion. Maybe this should be @mila-udem repository, I don't know yet. On 23 November 2015 at 11:32, Orhan Firat [email protected] wrote:...
We could do all compilation in `before_training`, that's where the algorithm compilation is currently done. I like the proposal, by the way.
I think it would be definitely great if machine translation demo was able to actually translate, that is use beam search to find the most likely candidate. Honestly, I thought...
How long did you train `reverse_words`? Did you try to use `beam_search` mode?
The error in beam search has just been fixed. The default number of iterations is 10000, not 100, but I guess this is what you ran. I will try to...
I am not sure I understand what kind of support from Fuel you guys want. Yes, statistics can be computed by iterating dataset. Indeed, `DatasetEvaluator` from Blocks might be a...
Hold on guys, what will happen if the user creates two iterators from a transformer that uses a stateful data stream, like e.g. a datastream over a `TextFile`? It seems...
Sorry for the late reply and thanks for the PR! I am a bit hesitant to merge it though. There is so many ways in which `Printing` could be customized,...