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.

Sign up for GitHub

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

Jump to bottom

Implementation of CTC in pure theano with custom gradient #108

Open
wants to merge 8 commits into
base: master
Choose a base branch
from
Open

Implementation of CTC in pure theano with custom gradient #108

wants to merge 8 commits into from

Commits on Aug 26, 2017

  1. Implementation of CTC in pure theano with custom gradient (which shou… 

    …ld hopefully be more robust to precision issues)
    @nlgranger
    nlgranger committed Aug 26, 2017
    Configuration menu
    Copy the full SHA
    9ca152d View commit details
    Browse the repository at this point in the history

Commits on Dec 15, 2017

  1. fix error with empty sequences+added low-level test for forward backw… 

    …ard passes
    @nlgranger
    nlgranger committed Dec 15, 2017
    Configuration menu
    Copy the full SHA
    96b8d68 View commit details
    Browse the repository at this point in the history

  2. more fixes for precision issues

    @nlgranger
    nlgranger committed Dec 15, 2017
    Configuration menu
    Copy the full SHA
    afcf4b7 View commit details
    Browse the repository at this point in the history

Commits on Jan 8, 2018

  1. Improved ctc gradient stability 

    The CTC loss function now takes predictions in log space (before
softmax) to avoid precision issues.
    @nlgranger
    nlgranger committed Jan 8, 2018
    Configuration menu
    Copy the full SHA
    c7ce022 View commit details
    Browse the repository at this point in the history

Commits on Jan 9, 2018

  1. test for more stability in computations

    @nlgranger
    nlgranger committed Jan 9, 2018
    Configuration menu
    Copy the full SHA
    0f739a8 View commit details
    Browse the repository at this point in the history

Commits on Jan 12, 2018

  1. fixes errors and precision issues, adds demos and tests.

    @nlgranger
    nlgranger committed Jan 12, 2018
    Configuration menu
    Copy the full SHA
    51a3585 View commit details
    Browse the repository at this point in the history

Commits on Jan 24, 2018

  1. fixed error in logsumexp, ctc gradient is now equal to tensorflow's

    @nlgranger
    nlgranger committed Jan 24, 2018
    Configuration menu
    Copy the full SHA
    f58174f View commit details
    Browse the repository at this point in the history

Commits on Jun 24, 2018

  1. split ctc op in two to avoid redundant computation: no improvement th… 

    …ough...
    @nlgranger
    nlgranger committed Jun 24, 2018
    Configuration menu
    Copy the full SHA
    324f2a9 View commit details
    Browse the repository at this point in the history