Commits

Breadcrumbs

History for

ondevelop

User selector

All users

All time

Commit History

Commits on May 2, 2021

• 'funcionality' -> 'functionality'

  jim-hays-root

  committed
  6152964

  Copy full SHA for 6152964

Commits on Mar 8, 2021

• Fix special transfers inside the toolkit

  Gastron

  committed
  87888a2

  Copy full SHA for 87888a2

Commits on Feb 14, 2021

• updated alignment/aligner.py docs

  nauman-daw

  committed
  9ff5f31

  Copy full SHA for 9ff5f31

Commits on Jan 20, 2021

• fixed typos in docstrings

  

  mirco

  committed
  5be94bb

  Copy full SHA for 5be94bb

Commits on Jan 15, 2021

• Update init.py to work as a pip package

  pplantinga

  committed
  d277af4

  Copy full SHA for d277af4

Commits on Dec 7, 2020

• Checkpointer doesn't support default args so perhaps best not to use anywhere

  Gastron

  committed
  21e89dc

  Copy full SHA for 21e89dc

Commits on Nov 16, 2020

• fixed tests

  mravanelli

  committed
  2a4e3eb

  Copy full SHA for 2a4e3eb

• convert align

  mravanelli

  committed
  a438c6a

  Copy full SHA for a438c6a

Commits on Oct 14, 2020

• Fix checkpoint loading for DDP

  pplantinga

  committed
  2eb3695

  Copy full SHA for 2eb3695

Commits on Oct 2, 2020

• remove old decoders.py and put undo_padding in utils.data_utils

  mravanelli

  committed
  886d7b5

  Copy full SHA for 886d7b5

• fixing .to(device), remove rnn_flatten, avoid rev if fails

  mravanelli

  committed
  5abe122

  Copy full SHA for 5abe122

Commits on Aug 2, 2020

• solving conflicts

  

  mirco

  committed
  05b7ab5

  Copy full SHA for 05b7ab5

• Save alignments with checkpoints

  Gastron

  committed
  150de6a

  Copy full SHA for 150de6a

Commits on Jul 10, 2020

• fix use_lexicon expected test output

  elenaras

  committed
  8f664ab

  Copy full SHA for 8f664ab

• changed forward algorithm to store alpha_matrix and make identity overwrite when seq length exceeded

  elenaras

  committed
  7c6acb6

  Copy full SHA for 7c6acb6

• fix _get_flat_start_batch to support large states_per_phoneme

  elenaras

  committed
  ac36db8

  Copy full SHA for ac36db8

• replace constants with self.neg_inf where necessary

  elenaras

  committed
  49516fd

  Copy full SHA for 49516fd

Commits on Jul 6, 2020

• put class HMMAligner... at the top

  elenaras

  committed
  ef67523

  Copy full SHA for ef67523

• remove commented code

  elenaras

  committed
  74289c1

  Copy full SHA for 74289c1

• update docstring in HMMAligner init

  elenaras

  committed
  bd08c16

  Copy full SHA for bd08c16

• changed default batch_reduction so tests will work

  elenaras

  committed
  1738bd8

  Copy full SHA for 1738bd8

• pass loss reduction parameters to HMMAligner class

  elenaras

  committed
  2f330a7

  Copy full SHA for 2f330a7

Commits on Jul 2, 2020

• change map_inds_to_intersect example test to deterministic example

  elenaras

  committed
  ea67762

  Copy full SHA for ea67762

Commits on Jul 1, 2020

• allow training with any ground truth phn_set

  elenaras

  committed
  24b4129

  Copy full SHA for 24b4129

Commits on Jun 30, 2020

• remove phn_set parameter in use_lexicon method

  elenaras

  committed
  2dc3522

  Copy full SHA for 2dc3522

Commits on Jun 29, 2020

• use trans_prob, pi_prob, final states from lexicon processing

  elenaras

  committed
  35a31a6

  Copy full SHA for 35a31a6

Commits on Jun 28, 2020

• change phn_set in alignment example to 61

  elenaras

  committed
  15c164e

  Copy full SHA for 15c164e

• add use of lexicon (most of lexicon dict and transition matrix sections written by Loren)

  elenaras

  committed
  50cb509

  Copy full SHA for 50cb509

Commits on Jun 19, 2020

• add docstrings for accuracy calculation methods

  elenaras

  committed
  e9b651b

  Copy full SHA for e9b651b

• convert 3 state alignments to 1 state alignments so repeating phones dont get merged

  elenaras

  committed
  666053c

  Copy full SHA for 666053c

Commits on Jun 18, 2020

• add docstring for expand_phns_by_states_per_phoneme

  elenaras

  committed
  f3f19d4

  Copy full SHA for f3f19d4

• add way to specify reduction of losses

  elenaras

  committed
  9205a8d

  Copy full SHA for 9205a8d

• remove for loop when upsampling predicted alignments_

  elenaras

  committed
  9faba13

  Copy full SHA for 9faba13

• fix accidental states_per_phoneme hard-coding

  elenaras

  committed
  e2cdd45

  Copy full SHA for e2cdd45

Commits on Jun 16, 2020

• remove torch.set_printoptions code that I used for debugging

  elenaras

  committed
  ff85dec

  Copy full SHA for ff85dec