New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improved feature inference #2103
Conversation
Hey @anderleich |
I fixed some bugs in the code. It should be ready now |
When feature merge type is rnn_size: 512 I get the following error, as it is expecting 512 size tensors and not 512+16=528 size tensors
|
I found setting However, OpenNMT-py/onmt/utils/parse.py Line 257 in a55f246
I guess we should remove that constraint |
This constraint is inherent to the Transformer architecture. |
Thanks! That's what I finally did. I guess we are ready to merge, don't we? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A few comments.
Also, isn't there anything to add to the docs/FAQ?
I've tried fixing what you mentioned. Note: I've got another PR for the server part but I'll submit it in a new PR after this one is accepted. |
I've improved the feature inference pipeline to allow prior tokenization with joiners. Features can be dumped to a file now for debugging purposes during the vocabulary building phase.
It assumes source and features can be fed in two different ways:
The second case might enable a more customized feature map.