mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-26 18:24:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

Minor fixes.

Browse Source
This commit is contained in:
Fangjun Kuang 2021-12-22 23:47:24 +08:00
parent 32d96002d9
commit c24b610389
3 changed files with 5 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.github/workflows/run-pretrained-transducer-stateless.yml vendored
View File

@ -24,7 +24,7 @@ on:
types: [labeled] types: [labeled]
jobs: jobs:
run_pre_trained_conformer_ctc: run_pre_trained_transducer_stateless:
if: github.event.label.name == 'ready' || github.event_name == 'push' if: github.event.label.name == 'ready' || github.event_name == 'push'
runs-on: ${{ matrix.os }} runs-on: ${{ matrix.os }}
strategy: strategy:

2
README.md
View File

@ -80,7 +80,7 @@ We provide a Colab notebook to run a pre-trained RNN-T conformer model: [![Open
Using Conformer as encoder. The decoder consists of 1 embedding layer Using Conformer as encoder. The decoder consists of 1 embedding layer
and 1 convolutional layer. and 1 convolutional layer.
The best WER with beam search with beam size 4 is: The best WER using beam search with beam size 4 is:
| | test-clean | test-other | | | test-clean | test-other |
|-----|------------|------------| |-----|------------|------------|

3
egs/librispeech/ASR/transducer_stateless/decoder.py
View File

@ -75,6 +75,9 @@ class Decoder(nn.Module):
Args: Args:
y: y:
A 2-D tensor of shape (N, U) with blank prepended. A 2-D tensor of shape (N, U) with blank prepended.
need_pad:
True to left padding the input. Should be True during training.
False to not pad the input. Used only during inference.
Returns: Returns:
Return a tensor of shape (N, U, embedding_dim). Return a tensor of shape (N, U, embedding_dim).
""" """