Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add prefix beam search and corresponding decoding methods #1786

Merged
merged 3 commits into from
Oct 30, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 6 additions & 3 deletions egs/librispeech/ASR/RESULTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -153,6 +153,7 @@ You can use <https://github.com/k2-fsa/sherpa> to deploy it.
| decoding method | test-clean | test-other | comment |
|--------------------------------------|------------|------------|---------------------|
| ctc-greedy-decoding | 2.57 | 5.95 | --epoch 50 --avg 25 |
| ctc-prefix-beam-search | 2.52 | 5.85 | --epoch 50 --avg 25 |

The training command using 2 32G-V100 GPUs is:
```bash
Expand Down Expand Up @@ -184,7 +185,7 @@ export CUDA_VISIBLE_DEVICES="0,1"
The decoding command is:
```bash
export CUDA_VISIBLE_DEVICES="0"
for m in ctc-greedy-search; do
for m in ctc-greedy-search ctc-prefix-beam-search; do
./zipformer/ctc_decode.py \
--epoch 50 \
--avg 25 \
Expand Down Expand Up @@ -212,6 +213,7 @@ You can use <https://github.com/k2-fsa/sherpa> to deploy it.
| decoding method | test-clean | test-other | comment |
|--------------------------------------|------------|------------|---------------------|
| ctc-greedy-decoding | 2.12 | 4.62 | --epoch 50 --avg 24 |
| ctc-prefix-beam-search | 2.1 | 4.61 | --epoch 50 --avg 24 |

The training command using 4 32G-V100 GPUs is:
```bash
Expand All @@ -238,7 +240,7 @@ export CUDA_VISIBLE_DEVICES="0,1,2,3"
The decoding command is:
```bash
export CUDA_VISIBLE_DEVICES="0"
for m in ctc-greedy-search; do
for m in ctc-greedy-search ctc-prefix-beam-search; do
./zipformer/ctc_decode.py \
--epoch 50 \
--avg 24 \
Expand All @@ -262,6 +264,7 @@ You can use <https://github.com/k2-fsa/sherpa> to deploy it.
| decoding method | test-clean | test-other | comment |
|--------------------------------------|------------|------------|---------------------|
| ctc-greedy-decoding | 2.03 | 4.37 | --epoch 50 --avg 26 |
| ctc-prefix-beam-search | 2.02 | 4.35 | --epoch 50 --avg 26 |

The training command using 2 80G-A100 GPUs is:
```bash
Expand Down Expand Up @@ -292,7 +295,7 @@ export CUDA_VISIBLE_DEVICES="0,1"
The decoding command is:
```bash
export CUDA_VISIBLE_DEVICES="0"
for m in ctc-greedy-search; do
for m in ctc-greedy-search ctc-prefix-beam-search; do
./zipformer/ctc_decode.py \
--epoch 50 \
--avg 26 \
Expand Down
Loading
Loading