mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-12-11 06:55:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

update documentation for shallow fusion

Browse Source
This commit is contained in:
marcoyang 2023-06-28 16:53:09 +08:00
parent 2ada280379
commit 3207ceab46
1 changed files with 26 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
docs/source/decoding-with-langugage-models/shallow-fusion.rst
View File

@ -51,7 +51,7 @@ To test the model, let's have a look at the decoding results without using LM. T
The following WERs are achieved on test-clean and test-other: The following WERs are achieved on test-clean and test-other:
.. code-block:: bash .. code-block:: text
$ For test-clean, WER of different settings are: $ For test-clean, WER of different settings are:
$ beam_size_4 3.11 best for test-clean $ beam_size_4 3.11 best for test-clean
@ -81,6 +81,7 @@ To use shallow fusion for decoding, we can execute the following command:
$ exp_dir=./icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/exp $ exp_dir=./icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/exp
$ lm_dir=./icefall-librispeech-rnn-lm/exp $ lm_dir=./icefall-librispeech-rnn-lm/exp
$ lm_scale=0.29
$ ./pruned_transducer_stateless7_streaming/decode.py \ $ ./pruned_transducer_stateless7_streaming/decode.py \
--epoch 99 \ --epoch 99 \
--avg 1 \ --avg 1 \
@ -95,7 +96,7 @@ To use shallow fusion for decoding, we can execute the following command:
--lm-type rnn \ --lm-type rnn \
--lm-exp-dir $lm_dir \ --lm-exp-dir $lm_dir \
--lm-epoch 99 \ --lm-epoch 99 \
--lm-scale 0.29 \ --lm-scale $lm_scale \
--lm-avg 1 \ --lm-avg 1 \
--rnn-lm-embedding-dim 2048 \ --rnn-lm-embedding-dim 2048 \
--rnn-lm-hidden-dim 2048 \ --rnn-lm-hidden-dim 2048 \
@ -118,7 +119,7 @@ between ``rnn`` or ``transformer``. The following three arguments are associated
The decoding result obtained with the above command are shown below. The decoding result obtained with the above command are shown below.
.. code-block:: bash .. code-block:: text
$ For test-clean, WER of different settings are: $ For test-clean, WER of different settings are:
$ beam_size_4 2.77 best for test-clean $ beam_size_4 2.77 best for test-clean
@ -137,8 +138,30 @@ A few parameters can be tuned to further boost the performance of shallow fusion
The number of active paths in the search beam. It controls the trade-off between decoding efficiency and accuracy. The number of active paths in the search beam. It controls the trade-off between decoding efficiency and accuracy.
Here, we also show how `--beam-size` effect the WER and decoding time:
.. list-table:: WERs and decoding time (on test-clean) of shallow fusion with different beam sizes
:widths: 25 25 50
:header-rows: 1
* - Beam size
- test-clean
- test-other
- Decoding time on test-clean (s)
* - 4
- 2.77
- 7.08
- 262
* - 8
- 2.62
- 6.65
- 352
* - 12
- 2.58
- 6.65
- 488
As we see, a larger beam size during shallow fusion improves the WER, but is also slower.