-
Notifications
You must be signed in to change notification settings - Fork 23
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
Make tf flow functions usable for onnx #431
Conversation
This will be used e.g. in rwth-i6/i6_experiments#155. |
33a9753
to
74b1abe
Compare
74b1abe
to
52e8dd1
Compare
The number of threads is now moved to the post config and the hashing issue for flow post configs (#436) is resolved. |
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.
(ONNX) Recognition with this worked fine, also from the code I see no issues. Potential TF changes should be covered by the AppTek-tests (@michelwi ?), so I guess this is all good.
No, we do not use this function. |
I tested the changes for one of my tf precomputed hybrid setups. |
To set up a recognition using RASR's
tensorflow-forward
flow node, we have helper functions inreturnn.flow.py
. I'd like to generalize them to also support theonnx-forward
node.This involves generalizing the existing function to create the precomputed feature flow and adding separate functions for tf and onnx with the specific needs for their configs. The function to integrate it into the base flow is basically just renamed and a version keeping the tf behavior is added.