How does one calculate the Stride multiplication factor for calculating word timestamps for different models? #5389
Unanswered
ishansharma1320
asked this question in
Q&A
Replies: 1 comment
-
Window_stride is in the preprocessor config. Stride factor depends on model, there's no predefined section that defines it. It depends on the encoder, not the decoder - Conformer is usually 4x, ContextNet is 8x. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
As stated in this comment, the combination of stride multiplication factor along with model_stride is needed for the calculation of word-based timestamps.
Can someone please let me know how to calculate the aforementioned multiplication factor?
Lastly, is the stride multiplication factor dependent on the model architecture?
if, yes, then do conformer_transducer_large and contextnet have the same value since their underlying architectures are the same?
Beta Was this translation helpful? Give feedback.
All reactions