Skip to content
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

Load time signal data #133

Merged
merged 4 commits into from
Jul 16, 2019
Merged

Conversation

mennetob
Copy link
Contributor

This pull request enables to load the wav data directly from kaldi ark files.

It also handles the issue of different dimensions in the input sequence length and label sequence length.

The changes are supposed to be completely backwards compatible to previous versions. But has only been tested on a small number of cases. Thus the whole implementation is kept in run_nn_refac01. But after more rigorous testing could potentially replace run_nn.

For the implementation to work the cfg_proto and cfg_proto_chunk need to be adjusted. For now the respective files have been added as separate files *_refac01.proto

menne added 4 commits July 16, 2019 17:21
* loading the raw time signal directly from wav-copy ark output
* handling different sized chunking of input and output sequence
* chunking information is currently handed over through config option
  max_seq_length_*
* added option to use max_sequence_length as carrier for the asynchron
  chunking information.
* Implemention is backwards compatible with standard use
* this is needed if the feature and label sequences are of different
  length and the respective sub sampling is done by the network
@mravanelli mravanelli merged commit 0b7b603 into mravanelli:master Jul 16, 2019
@mennetob mennetob deleted the loadTimeSignal branch July 17, 2019 13:48
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants