RNNT Doesn't backpropa with BPTT? #167

Open

Assignees

opened

Title

RNNT Doesn't backpropa with BPTT?

Description

i see, Trainer code and model loss output, RNN-T does't BPTT is that right? and why?

resolved #

to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Assignees

sooftware

Labels

No labels

No labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests