Commits · b7d44d97af5778012817bce06da7eec08ec2ffc3 · OpenDAS / Torchaudio

17 Aug, 2021 1 commit
- RNNT loss resolve null gradient (#1707) · 0f603eb9
  Caroline Chen authored Aug 17, 2021
  
  0f603eb9
03 Aug, 2021 2 commits
- Remove fused_log_softmax option from RNNT Loss (#1615) · d74d0604
  Caroline Chen authored Aug 03, 2021
  
  d74d0604
- Remove reuse_logits_for_grads option for RNNTL (#1610) · 16f3b2f9
  Caroline Chen authored Aug 03, 2021
  
  16f3b2f9
15 Jul, 2021 1 commit
- Refactor RNNT Loss Unit Tests (#1630) · 0ea6d10d
  Caroline Chen authored Jul 15, 2021
  
  0ea6d10d
28 Jun, 2021 1 commit
- Rename transducer to RNNT (#1603) · a9623854
  Caroline Chen authored Jun 28, 2021
  
  a9623854
24 Jun, 2021 1 commit
- Add reduction parameter for RNNT loss (#1590) · bac32ec1
  Caroline Chen authored Jun 24, 2021
  
  bac32ec1
04 Jun, 2021 1 commit
- Add autograd gradcheck test for RNN transducer loss (#1532) · d4d09074
  Vincent QB authored Jun 04, 2021
```
* autograd test from carolineechen/audio#2

* fix numpy backward: be careful to not modify inplace.
```
  d4d09074
19 May, 2021 1 commit
- Add torchscript support to RNNT Loss (#1507) · af7eb4d6
  Caroline Chen authored May 19, 2021
  
  af7eb4d6
06 May, 2021 1 commit
- Add GPU RNNT Loss (#1483) · 5417e4fb
  Caroline Chen authored May 06, 2021
  
  5417e4fb
30 Apr, 2021 1 commit

Caroline Chen authored Apr 30, 2021

Replace the prototype RNNT implementation (using warp-transducer) with one without external library dependencies

0c263a93