Skip to content
This repository was archived by the owner on Jul 7, 2023. It is now read-only.

Conversation

@yynil
Copy link
Contributor

@yynil yynil commented Jun 29, 2017

First of all, I apologize for the previous error that I misused the len operator.
Then I figure out if the beam is set to the value that's greater than 1 will still cause the Out of index error.

For Class_Label_Modality search , beam is set to 1 to enable greedy_infere seems to get better results.
@lukaszkaiser

  1. for beam search use the shape.ndims to avoid out of index error
  2. for greedy search, still use the shape.ndims to avoid the out of index error. Before that I misuse the slice operation:-(

1. for beam search use the shape.ndims to avoid out of index error
2. for greedy search, still use the shape.ndims to avoid the out of index error. Before that I misuse the slice operation:-(
Copy link
Contributor

@lukaszkaiser lukaszkaiser left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks! I might overwrite parts of it soon, but let's have it for now.

@lukaszkaiser lukaszkaiser merged commit 06df1d4 into tensorflow:master Jun 29, 2017
@yynil
Copy link
Contributor Author

yynil commented Jun 30, 2017

I confirm 1.0.9 T2T class modality decoding works out of box and produces exact the same results I got in this fix.
Thanks a lot!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants