Skip to content

Upload FireRedASR v2 models (AED and CTC) #3220

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.

Sign up for GitHub

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

Merged
csukuangfj merged 1 commit into from
Feb 26, 2026
Merged

Upload FireRedASR v2 models (AED and CTC) #3220

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion .github/scripts/test-nodejs-addon-npm.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,7 @@ rm -rf sherpa-onnx-funasr-nano-int8-2025-12-30

echo "----------non-streaming ASR Google MedASR CTC----------"

wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-medasr-ctc-en-int8-2025-12-25.tar.bz2
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-medasr-ctc-en-int8-2025-12-25.tar.bz2

@gemini-code-assist gemini-code-assist Bot Feb 26, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

Using -O relies on the URL's path to determine the output filename. While it works here, it's more robust to explicitly specify the output file with -o <filename>. This decouples the script from the URL structure and ensures the filename matches what subsequent commands like tar and rm expect.

Suggested change
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-medasr-ctc-en-int8-2025-12-25.tar.bz2
curl -SL -o sherpa-onnx-medasr-ctc-en-int8-2025-12-25.tar.bz2 https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-medasr-ctc-en-int8-2025-12-25.tar.bz2

tar xvf sherpa-onnx-medasr-ctc-en-int8-2025-12-25.tar.bz2
rm sherpa-onnx-medasr-ctc-en-int8-2025-12-25.tar.bz2

Expand Down
146 changes: 141 additions & 5 deletions .github/workflows/upload-models.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,12 +25,14 @@ jobs:
- uses: actions/checkout@v4

- name: Install ffmpeg
if: false
shell: bash
run: |
sudo apt-get update
sudo apt-get install -y ffmpeg

- name: Verify ffmpeg
if: false
shell: bash
run: |
ffmpeg -version
Expand All @@ -41,6 +43,140 @@ jobs:
git config --global user.email "csukuangfj@gmail.com"
git config --global user.name "Fangjun Kuang"

- name: FireRedASR2 CTC (int8)
if: false
shell: bash
run: |
d=sherpa-onnx-fire-red-asr2-ctc-zh_en-int8-2026-02-25
mkdir $d

pushd $d

cat >README.md <<EOF
# Introduction
Model files are converted from
https://www.modelscope.cn/models/FireRedTeam/FireRedASR2-AED

We export only the encoder and the CTC branch. The attention decoder
is not used.
EOF

curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/ctc/model.int8.onnx
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/ctc/tokens.txt
mkdir test_wavs
cd test_wavs
for w in 0.wav 1.wav 2.wav 3-sichuan.wav 3.wav 4-tianjin.wav 5-henan.wav 8k.wav; do
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-fire-red-asr-large-zh_en-fp16-2025-02-16/resolve/main/test_wavs/$w
done

popd

ls -lh $d

tar cjvf $d.tar.bz2 $d

ls -lh *.tar.bz2

- name: FireRedASR2 CTC (fp32)
if: false
shell: bash
run: |
d=sherpa-onnx-fire-red-asr2-ctc-zh_en-2026-02-25
mkdir $d

pushd $d

cat >README.md <<EOF
# Introduction
Model files are converted from
https://www.modelscope.cn/models/FireRedTeam/FireRedASR2-AED

We export only the encoder and the CTC branch. The attention decoder
is not used.
EOF

curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/ctc/model.onnx
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/ctc/model.weights
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/ctc/tokens.txt
mkdir test_wavs
cd test_wavs
for w in 0.wav 1.wav 2.wav 3-sichuan.wav 3.wav 4-tianjin.wav 5-henan.wav 8k.wav; do
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-fire-red-asr-large-zh_en-fp16-2025-02-16/resolve/main/test_wavs/$w
done

popd

ls -lh $d

tar cjvf $d.tar.bz2 $d

ls -lh *.tar.bz2

- name: FireRedASR2 AED (int8)
if: false
shell: bash
run: |
d=sherpa-onnx-fire-red-asr2-zh_en-int8-2026-02-26
mkdir $d

pushd $d

cat >README.md <<EOF
# Introduction
Model files are converted from
https://www.modelscope.cn/models/FireRedTeam/FireRedASR2-AED
EOF

curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/aed/encoder.int8.onnx
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/aed/decoder.int8.onnx
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/aed/tokens.txt
mkdir test_wavs
cd test_wavs
for w in 0.wav 1.wav 2.wav 3-sichuan.wav 3.wav 4-tianjin.wav 5-henan.wav 8k.wav; do
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-fire-red-asr-large-zh_en-fp16-2025-02-16/resolve/main/test_wavs/$w
done

popd

ls -lh $d

tar cjvf $d.tar.bz2 $d

ls -lh *.tar.bz2

- name: FireRedASR2 AED (fp32)
if: false
shell: bash
run: |
d=sherpa-onnx-fire-red-asr2-zh_en-2026-02-26
mkdir $d

pushd $d

cat >README.md <<EOF
# Introduction
Model files are converted from
https://www.modelscope.cn/models/FireRedTeam/FireRedASR2-AED
EOF

curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/aed/encoder.onnx
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/aed/encoder.weights
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/aed/decoder.onnx
curl -SL -O https://www.modelscope.cn/models/csukuangfj/FireRedASR2-AED-onnx/resolve/master/aed/tokens.txt
mkdir test_wavs
cd test_wavs
for w in 0.wav 1.wav 2.wav 3-sichuan.wav 3.wav 4-tianjin.wav 5-henan.wav 8k.wav; do
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-fire-red-asr-large-zh_en-fp16-2025-02-16/resolve/main/test_wavs/$w
done

popd

ls -lh $d

tar cjvf $d.tar.bz2 $d

ls -lh *.tar.bz2

- name: Zipformer-30M-RNNT-6000h
if: false
shell: bash
Expand Down Expand Up @@ -925,6 +1061,10 @@ jobs:
sherpa-onnx-streaming-zipformer-bn-vosk-2026-02-09
sherpa-onnx-zipformer-vi-30M-int8-2026-02-09
sherpa-onnx-zipformer-vi-30M-2026-02-09
sherpa-onnx-fire-red-asr2-ctc-zh_en-int8-2026-02-25
sherpa-onnx-fire-red-asr2-ctc-zh_en-2026-02-25
sherpa-onnx-fire-red-asr2-zh_en-int8-2026-02-26
sherpa-onnx-fire-red-asr2-zh_en-2026-02-26
)
for d in ${models[@]}; do
if [ ! -d $d ]; then
Expand All @@ -944,6 +1084,7 @@ jobs:
pushd huggingface
git lfs track "*.onnx"
git lfs track "*.data"
git lfs track "*.weights"
git lfs track "bpe.model"
git lfs track "*.wav"
git lfs track "*.json"
Expand Down Expand Up @@ -989,11 +1130,6 @@ jobs:
popd
done

- uses: actions/upload-artifact@v4
if: false
with:
name: here
path: ./*.tar.bz2

- name: Release
if: true
Expand Down