Refactor exporting NeMo models (#2362)

Refactors and extends model export support to include new NeMo Parakeet TDT int8 variants for English and Japanese, updating the Kotlin API, export scripts, test runners, and CI workflows.

- Added support for two new int8 model types in OfflineRecognizer.kt.
- Enhanced Python export scripts to perform dynamic quantization and metadata injection.
- Updated shell scripts and GitHub workflows to package, test, and publish int8 model artifacts.

.github/workflows/export-nemo-fast-conformer-hybrid-transducer-ctc-non-streaming.yaml

@@ -61,6 +61,11 @@ jobs:
               sherpa-onnx-nemo-fast-conformer-ctc-en-de-es-fr-14288
               sherpa-onnx-nemo-fast-conformer-ctc-be-de-en-es-fr-hr-it-pl-ru-uk-20k
               sherpa-onnx-nemo-parakeet_tdt_ctc_110m-en-36000
+              sherpa-onnx-nemo-fast-conformer-ctc-en-24500-int8
+              sherpa-onnx-nemo-fast-conformer-ctc-es-1424-int8
+              sherpa-onnx-nemo-fast-conformer-ctc-en-de-es-fr-14288-int8
+              sherpa-onnx-nemo-fast-conformer-ctc-be-de-en-es-fr-hr-it-pl-ru-uk-20k-int8
+              sherpa-onnx-nemo-parakeet_tdt_ctc_110m-en-36000-int8
             )
 
             for m in ${models[@]}; do
@@ -89,6 +94,11 @@ jobs:
             sherpa-onnx-nemo-fast-conformer-ctc-en-de-es-fr-14288
             sherpa-onnx-nemo-fast-conformer-ctc-be-de-en-es-fr-hr-it-pl-ru-uk-20k
             sherpa-onnx-nemo-parakeet_tdt_ctc_110m-en-36000
+            sherpa-onnx-nemo-fast-conformer-ctc-en-24500-int8
+            sherpa-onnx-nemo-fast-conformer-ctc-es-1424-int8
+            sherpa-onnx-nemo-fast-conformer-ctc-en-de-es-fr-14288-int8
+            sherpa-onnx-nemo-fast-conformer-ctc-be-de-en-es-fr-hr-it-pl-ru-uk-20k-int8
+            sherpa-onnx-nemo-parakeet_tdt_ctc_110m-en-36000-int8
           )
           for d in ${dirs[@]}; do
             tar cjvf ${d}.tar.bz2 ./$d

.github/workflows/export-nemo-fast-conformer-hybrid-transducer-ctc.yaml

@@ -54,13 +54,18 @@ jobs:
           curl -SL -O https://hf-mirror.com/csukuangfj/sherpa-onnx-nemo-ctc-en-conformer-small/resolve/main/test_wavs/trans.txt
           popd
 
-          cp -av test_wavs ./sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-80ms
-          cp -av test_wavs ./sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-480ms
-          cp -av test_wavs ./sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-1040ms
-
-          tar cjvf sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-80ms.tar.bz2 sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-80ms
-          tar cjvf sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-480ms.tar.bz2 sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-480ms
-          tar cjvf sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-1040ms.tar.bz2 sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-1040ms
+          names=(
+            sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-80ms
+            sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-480ms
+            sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-1040ms
+            sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-80ms-int8
+            sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-480ms-int8
+            sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-1040ms-int8
+          )
+          for d in ${names[@]}; do
+            cp -av test_wavs $d/
+            tar cjvf $d.tar.bz2 $d
+          done
 
       - name: Release
         uses: svenstaro/upload-release-action@v2
@@ -71,3 +76,41 @@ jobs:
           repo_name: k2-fsa/sherpa-onnx
           repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
           tag: asr-models
+
+      - name: Publish to huggingface
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 20
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
+            models=(
+              sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-80ms
+              sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-480ms
+              sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-1040ms
+              sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-80ms-int8
+              sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-480ms-int8
+              sherpa-onnx-nemo-streaming-fast-conformer-ctc-en-1040ms-int8
+            )
+
+            for m in ${models[@]}; do
+              rm -rf huggingface
+              export GIT_LFS_SKIP_SMUDGE=1
+              export GIT_CLONE_PROTECTION_ACTIVE=false
+              git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$m huggingface
+              cp -av $m/* huggingface
+              cd huggingface
+              git lfs track "*.onnx"
+              git lfs track "*.wav"
+              git status
+              git add .
+              git status
+              git commit -m "first commit"
+              git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$m main
+              cd ..
+            done

.github/workflows/export-nemo-fast-conformer-hybrid-transducer-transducer-non-streaming.yaml

@@ -61,6 +61,11 @@ jobs:
               sherpa-onnx-nemo-fast-conformer-transducer-en-de-es-fr-14288
               sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k
               sherpa-onnx-nemo-parakeet_tdt_transducer_110m-en-36000
+              sherpa-onnx-nemo-fast-conformer-transducer-en-24500-int8
+              sherpa-onnx-nemo-fast-conformer-transducer-es-1424-int8
+              sherpa-onnx-nemo-fast-conformer-transducer-en-de-es-fr-14288-int8
+              sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k-int8
+              sherpa-onnx-nemo-parakeet_tdt_transducer_110m-en-36000-int8
             )
 
             for m in ${models[@]}; do
@@ -88,6 +93,11 @@ jobs:
             sherpa-onnx-nemo-fast-conformer-transducer-en-de-es-fr-14288
             sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k
             sherpa-onnx-nemo-parakeet_tdt_transducer_110m-en-36000
+            sherpa-onnx-nemo-fast-conformer-transducer-en-24500-int8
+            sherpa-onnx-nemo-fast-conformer-transducer-es-1424-int8
+            sherpa-onnx-nemo-fast-conformer-transducer-en-de-es-fr-14288-int8
+            sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k-int8
+            sherpa-onnx-nemo-parakeet_tdt_transducer_110m-en-36000-int8
           )
           for d in ${dirs[@]}; do
             tar cjvf ${d}.tar.bz2 ./$d

.github/workflows/export-nemo-fast-conformer-hybrid-transducer-transducer.yaml

@@ -54,13 +54,18 @@ jobs:
           curl -SL -O https://hf-mirror.com/csukuangfj/sherpa-onnx-nemo-ctc-en-conformer-small/resolve/main/test_wavs/trans.txt
           popd
 
-          cp -av test_wavs ./sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms
-          cp -av test_wavs ./sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-480ms
-          cp -av test_wavs ./sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-1040ms
-
-          tar cjvf sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms.tar.bz2 sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms
-          tar cjvf sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-480ms.tar.bz2 sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-480ms
-          tar cjvf sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-1040ms.tar.bz2 sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-1040ms
+          models=(
+            sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms
+            sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-480ms
+            sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-1040ms
+            sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms-int8
+            sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-480ms-int8
+            sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-1040ms-int8
+          )
+          for m in ${models[@]}; do
+            cp -av test_wavs $m
+            tar cjvf $m.tar.bz2 $m
+          done
 
       - name: Release
         uses: svenstaro/upload-release-action@v2
@@ -71,3 +76,41 @@ jobs:
           repo_name: k2-fsa/sherpa-onnx
           repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
           tag: asr-models
+
+      - name: Publish to huggingface
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 20
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
+            models=(
+              sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms
+              sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-480ms
+              sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-1040ms
+              sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms-int8
+              sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-480ms-int8
+              sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-1040ms-int8
+            )
+
+            for m in ${models[@]}; do
+              rm -rf huggingface
+              export GIT_LFS_SKIP_SMUDGE=1
+              export GIT_CLONE_PROTECTION_ACTIVE=false
+              git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$m huggingface
+              cp -av $m/* huggingface
+              cd huggingface
+              git lfs track "*.onnx"
+              git lfs track "*.wav"
+              git status
+              git add .
+              git status
+              git commit -m "first commit"
+              git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$m main
+              cd ..
+            done

.github/workflows/export-nemo-parakeet-tdt.yaml

@@ -0,0 +1,105 @@
+name: export-nemo-parakeet-tdt
+
+on:
+  push:
+    branches:
+      - refactor-export-nemo
+  workflow_dispatch:
+
+concurrency:
+  group: export-nemo-parakeet-tdt-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  export-nemo-parakeet-tdt-0_6b-v2:
+    if: github.repository_owner == 'k2-fsa' || github.repository_owner == 'csukuangfj'
+    name: parakeet tdt
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [macos-latest]
+        python-version: ["3.10"]
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install python dependencies
+        shell: bash
+        run: |
+          pip install \
+            nemo_toolkit['asr'] \
+            "numpy<2" \
+            ipython \
+            kaldi-native-fbank \
+            librosa \
+            onnx==1.17.0 \
+            onnxmltools==1.13.0 \
+            onnxruntime==1.17.1 \
+            soundfile
+
+      - name: Run
+        shell: bash
+        run: |
+          cd scripts/nemo/parakeet-tdt_ctc-0.6b-ja
+          ./run-ctc.sh
+
+      - name: Collect files
+        shell: bash
+        run: |
+          models=(
+            sherpa-onnx-nemo-parakeet-tdt_ctc-0.6b-ja-35000-int8
+          )
+          for m in ${models[@]}; do
+            mv -v scripts/nemo/parakeet-tdt_ctc-0.6b-ja/$m .
+            tar cjfv $m.tar.bz2 $m
+          done
+
+
+      - name: Publish to huggingface
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 20
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
+            models=(
+              sherpa-onnx-nemo-parakeet-tdt_ctc-0.6b-ja-35000-int8
+            )
+
+            for m in ${models[@]}; do
+              rm -rf huggingface
+              export GIT_LFS_SKIP_SMUDGE=1
+              export GIT_CLONE_PROTECTION_ACTIVE=false
+              git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$m huggingface
+              cp -av $m/* huggingface
+              cd huggingface
+              git lfs track "*.onnx"
+              git lfs track "*.wav"
+              git status
+              git add .
+              git status
+              git commit -m "first commit"
+              git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$m main
+              cd ..
+            done
+
+      - name: Release
+        uses: svenstaro/upload-release-action@v2
+        with:
+          file_glob: true
+          file: ./*.tar.bz2
+          overwrite: true
+          repo_name: k2-fsa/sherpa-onnx
+          repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
+          tag: asr-models