Commits

Breadcrumbs

History for

on7100c33

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 21, 2022

• Add pruned RNN-T for aishell. (k2-fsa#436)

  Show description for 7100c33
  csukuangfj

  authored
  7100c33

  Copy full SHA for 7100c33

Commits on Jun 6, 2022

• Use jsonl for CutSet in the LibriSpeech recipe. (k2-fsa#397)

  Show description for f1abce7
  csukuangfj

  authored
  f1abce7

  Copy full SHA for f1abce7

Commits on Jun 4, 2022

• [Ready to merge] Pruned_transducer_stateless2 for alimeeting dataset (k2-fsa#378)

  Show description for beab229
  luomingshuang

  authored
  beab229

  Copy full SHA for beab229

Commits on May 23, 2022

• Narrower and deeper conformer (k2-fsa#330)

  Show description for 2f1e23c
  csukuangfj

  authored
  2f1e23c

  Copy full SHA for 2f1e23c

Commits on May 16, 2022

• Various fixes to support torch script. (k2-fsa#371)

  Show description for f6ce135
  csukuangfj

  authored
  f6ce135

  Copy full SHA for f6ce135

Commits on May 15, 2022

• Fix GitHub CI for decoding GigaSpeech dev/test datasets (k2-fsa#366)

  csukuangfj

  authored
  6f7860a

  Copy full SHA for 6f7860a

Commits on May 14, 2022

• Update results for libri+giga multi dataset setup. (k2-fsa#363)

  Show description for f23dd43
  csukuangfj

  authored
  f23dd43

  Copy full SHA for f23dd43

• Decode gigaspeech in GitHub actions (k2-fsa#362)

  Show description for 2d7096d
  csukuangfj

  authored
  2d7096d

  Copy full SHA for 2d7096d

Commits on May 12, 2022

• Ignore padding frames during RNN-T decoding. (k2-fsa#358)

  Show description for aeb8986
  csukuangfj

  authored
  aeb8986

  Copy full SHA for aeb8986

Commits on May 10, 2022

• Run decode.py in GitHub actions. (k2-fsa#356)

  csukuangfj

  authored
  bc284e8

  Copy full SHA for bc284e8

Commits on Apr 29, 2022

• Modified conformer with multi datasets (k2-fsa#312)

  Show description for ac84220
  csukuangfj

  authored
  ac84220

  Copy full SHA for ac84220

Commits on Apr 19, 2022

• Support computing RNN-T loss with torchaudio (k2-fsa#316)

  csukuangfj

  authored
  fce7f3c

  Copy full SHA for fce7f3c

Commits on Mar 31, 2022

• Fix CI. (k2-fsa#280)

  Show description for 2045125
  csukuangfj

  authored
  2045125

  Copy full SHA for 2045125

Commits on Mar 15, 2022

• Cache pip packages for GitHub actions (k2-fsa#253)

  Show description for a764330
  csukuangfj

  authored
  a764330

  Copy full SHA for a764330

Commits on Mar 12, 2022

• Add modified beam search for pruned rnn-t. (k2-fsa#248)

  Show description for bb7f6ed
  csukuangfj

  authored
  bb7f6ed

  Copy full SHA for bb7f6ed

Commits on Mar 2, 2022

• Add modified transducer loss for AIShell dataset (k2-fsa#219)

  Show description for 50d2281
  csukuangfj

  authored
  50d2281

  Copy full SHA for 50d2281

Commits on Mar 1, 2022

• Update result for full libri + GigaSpeech using transducer_stateless. (k2-fsa#231)

  csukuangfj

  authored
  05cb297

  Copy full SHA for 05cb297

Commits on Feb 21, 2022

• Begin to use multiple datasets in training (k2-fsa#213)

  Show description for 2332ba3
  csukuangfj

  authored
  2332ba3

  Copy full SHA for 2332ba3

Commits on Feb 7, 2022

• Use modified transducer loss in training. (k2-fsa#179)

  Show description for a815002
  csukuangfj

  authored
  a815002

  Copy full SHA for a815002

Commits on Jan 10, 2022

• Use optimized_transducer to compute transducer loss. (k2-fsa#162)

  Show description for 4c1b366
  csukuangfj

  authored
  4c1b366

  Copy full SHA for 4c1b366

Commits on Dec 27, 2021

• Remove batchnorm, weight decay, and SOS from transducer conformer encoder (k2-fsa#155)

  Show description for 14c93ad
  csukuangfj

  authored
  14c93ad

  Copy full SHA for 14c93ad

Commits on Dec 23, 2021

• Minor fixes to the RNN-T Conformer model (k2-fsa#152)

  Show description for 5b6699a
  csukuangfj

  authored
  5b6699a

  Copy full SHA for 5b6699a

Commits on Dec 22, 2021

• Increase the size of the context in the RNN-T decoder. (k2-fsa#153)

  csukuangfj

  authored
  fb6a57e

  Copy full SHA for fb6a57e

Commits on Dec 17, 2021

• RNN-T Conformer training for LibriSpeech (k2-fsa#143)

  Show description for 1d44da8
  csukuangfj

  authored
  1d44da8

  Copy full SHA for 1d44da8

Commits on Dec 7, 2021

• Apply layer normalization to the output of each gate in LSTM/GRU. (k2-fsa#139)

  Show description for 1aff64b
  csukuangfj

  authored
  1aff64b

  Copy full SHA for 1aff64b

Commits on Nov 17, 2021

• New label smoothing (k2-fsa#109)

  Show description for 336283f
  csukuangfj

  authored
  336283f

  Copy full SHA for 336283f

Commits on Oct 18, 2021

• Add MMI training with word pieces as modelling unit. (k2-fsa#6)

  Show description for 53b79fa
  csukuangfj

  authored
  53b79fa

  Copy full SHA for 53b79fa

• Extract framewise alignment information using CTC decoding (k2-fsa#39)

  Show description for 4890e27
  csukuangfj

  authored
  4890e27

  Copy full SHA for 4890e27

Commits on Oct 14, 2021

• Test pre-trained model in CI (k2-fsa#80)

  Show description for fee1f84
  csukuangfj

  authored
  fee1f84

  Copy full SHA for fee1f84

Commits on Sep 20, 2021

• Refactor decode.py to make it more readable and more modular. (k2-fsa#44)

  Show description for a80e58e
  csukuangfj

  authored
  a80e58e

  Copy full SHA for a80e58e

Commits on Sep 17, 2021

• Fix decode.py to remove the correct axis. (k2-fsa#50)

  Show description for cc77cb3
  csukuangfj

  authored
  cc77cb3

  Copy full SHA for cc77cb3

Commits on Sep 14, 2021

• update api for RaggedTensor (k2-fsa#45)

  Show description for 9a6e048
  pkufool

  authored
  9a6e048

  Copy full SHA for 9a6e048

Commits on Sep 13, 2021

• Change default value of lattice-score-scale from 1.0 to 0.5 (k2-fsa#41)

  Show description for f792b46
  csukuangfj

  authored
  f792b46

  Copy full SHA for f792b46

Commits on Sep 8, 2021

• Use new APIs with k2.RaggedTensor (k2-fsa#38)

  Show description for abadc71
  csukuangfj

  authored
  abadc71

  Copy full SHA for abadc71

Commits on Aug 25, 2021

• Add documentation about code style and creating new recipes. (k2-fsa#27)

  csukuangfj

  authored
  184dbb3

  Copy full SHA for 184dbb3

Pagination

Previous

Next