Questions tagged [neural-mt]

11 questions
33
votes
1 answer

Docker build from Dockerfile with more memory

How to docker build from Dockerfile with more memory? This is a different question from this Allow more memory when docker build a Dockerfile When installing the software natively, there is enough memory to successfully build and install the marian…
alvas
  • 115,346
  • 109
  • 446
  • 738
7
votes
2 answers

Tensorflow ResourceExhaustedError after first batch

Summary and Test Cases The core issue is that Tensorflow throws OOM allocations on a batch that is not the first, as I would expect. Therefore, I believe there is a memory leak since all memory is clearly not being freed after each batch. num_units:…
Evan Weissburg
  • 1,564
  • 2
  • 17
  • 38
1
vote
1 answer

Implemenet attention in vanilla encoder-decoder architecture

I have tried a vanila enc-dec arch as following (english to french NMT) I want to know how to integrate keras attention layer here. Either from the keras docs or any other attention module from third party repo is also welcome. I just need to…
1
vote
1 answer

Estimated release date for the dictionary feature in V3 NMT

My department is planning to switching to using NMT V3 soon, but we will need the dictionary feature (adding glossary and Do-Not-Translate list) for training. This might have been asked before but I don't see any recent post asking for it, would you…
1
vote
1 answer

How to find and run the largest batch in a dataset before starting training

Question In Tensorflow, I frequently run into OOM errors during the first epoch of training. However, the large nature of the network causes the first epoch to take around an hour, far to long to test new hyper-parameters quickly. Ideally, I'd like…
0
votes
1 answer

Is it okay to compare Test BLEU score between NMT models while using a slightly modified standard test sets?

I am using tst2013.en found here, as my test sets to get the Test BLEU score to compare to other previous models. However, I have to filter out some sentences that are longer than 100 words otherwise I won't have the resource to run the model. But…
Minh Ung
  • 13
  • 5
0
votes
1 answer

How to use BLEU score to compare your model to existing models?

So I am using the BLEU score metric to compare my NMT model's performance with existing models. However, I'm wondering how many settings do I have to match with the other models. Settings like dev sets, test sets and hyperparameters I think are…
Minh Ung
  • 13
  • 5
0
votes
1 answer

Dynamic dictionary feature in NMT V3

I have heard that it should be possible to add a dynamic dictionary, holding a set of terms or word list when you train your model. This should be possible in the Translation Hob today. How to do this using V3 or NMT in…
0
votes
1 answer

Can you use dictionary with the generalnn category in MS Text API?

Is there a way to upload a terminology dictionary to the Hub, that would be applied when translating using generalnn category (not a trained engine)? The workaround that seem to work is to use the tag e.g.: Let's do a…
Andrew Poblocki
  • 153
  • 1
  • 6
0
votes
0 answers

Microsoft translator API - notranslate option not working with Neural Network MT

Does anyone experience notranslate option is ignored when sending to Microsoft Neural Network MT? The same string works with SMT. I am sending this with "text/html". My example string:

For example, it is important to install the latest

0
votes
0 answers

matlab neural network delete specific neuron

I am using matlab neural network functions with feedforward network. The algorithm by default creates fully connected network, is it possible to create a network which is not fully connected and is it possible to delete some specific neuron…