k2-fsa · csukuangfj · Apr 20, 2025 · Apr 19, 2025 · Apr 19, 2025 · Apr 19, 2025
diff --git a/.github/scripts/test-offline-ctc.sh b/.github/scripts/test-offline-ctc.sh
@@ -47,9 +47,23 @@ for type in base small; do
   rm -rf sherpa-onnx-dolphin-$type-ctc-multi-lang-2025-04-02
 done
 
+log "------------------------------------------------------------"
+log "Run NeMo GigaAM Russian models v2"
+log "------------------------------------------------------------"
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19.tar.bz2
+tar xvf sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19.tar.bz2
+rm sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19.tar.bz2
+
+$EXE \
+  --nemo-ctc-model=./sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/model.int8.onnx \
+  --tokens=./sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/tokens.txt \
+  --debug=1 \
+  ./sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/test_wavs/example.wav
+
+rm -rf sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19
 
 log "------------------------------------------------------------"
-log "Run NeMo GigaAM Russian models"
+log "Run NeMo GigaAM Russian models v1"
 log "------------------------------------------------------------"
 curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24.tar.bz2
 tar xvf sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24.tar.bz2

diff --git a/.github/scripts/test-offline-transducer.sh b/.github/scripts/test-offline-transducer.sh
@@ -15,6 +15,24 @@ echo "PATH: $PATH"
 
 which $EXE
 
+log "------------------------------------------------------------"
+log "Run NeMo GigaAM Russian models v2"
+log "------------------------------------------------------------"
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19.tar.bz2
+tar xvf sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19.tar.bz2
+rm sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19.tar.bz2
+
+$EXE \
+  --encoder=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/encoder.int8.onnx \
+  --decoder=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/decoder.onnx \
+  --joiner=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/joiner.onnx \
+  --tokens=./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/tokens.txt \
+  --model-type=nemo_transducer \
+  ./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/test_wavs/example.wav
+
+rm sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19
+
+
 log "------------------------------------------------------------------------"
 log "Run zipformer transducer models (Russian)                              "
 log "------------------------------------------------------------------------"

diff --git a/.github/workflows/export-nemo-giga-am-to-onnx.yaml b/.github/workflows/export-nemo-giga-am-to-onnx.yaml
@@ -43,7 +43,8 @@ jobs:
           mv -v scripts/nemo/GigaAM/tokens.txt $d/
           mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
           mv -v scripts/nemo/GigaAM/run-ctc.sh $d/
-          mv -v scripts/nemo/GigaAM/*-ctc.py $d/
+          mv -v scripts/nemo/GigaAM/export-onnx-ctc.py $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-ctc.py $d/
 
           ls -lh scripts/nemo/GigaAM/
 
@@ -71,7 +72,8 @@ jobs:
           mv -v scripts/nemo/GigaAM/tokens.txt $d/
           mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
           mv -v scripts/nemo/GigaAM/run-rnnt.sh $d/
-          mv -v scripts/nemo/GigaAM/*-rnnt.py $d/
+          mv -v scripts/nemo/GigaAM/export-onnx-rnnt.py $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-rnnt.py $d/
 
           ls -lh scripts/nemo/GigaAM/
 
@@ -91,20 +93,49 @@ jobs:
           mkdir $d/test_wavs
           rm scripts/nemo/GigaAM/v2_ctc.onnx
           mv -v scripts/nemo/GigaAM/*.int8.onnx $d/
-          cp -v scripts/nemo/GigaAM/LICENCE $d/
+          cp -v scripts/nemo/GigaAM/LICENSE $d/
           mv -v scripts/nemo/GigaAM/tokens.txt $d/
           mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
-          mv -v scripts/nemo/GigaAM/run-ctc.sh $d/
+          mv -v scripts/nemo/GigaAM/run-ctc-v2.sh $d/
           mv -v scripts/nemo/GigaAM/*-ctc-v2.py $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-ctc.py $d/
 
           ls -lh scripts/nemo/GigaAM/
 
           ls -lh $d
 
           tar cjvf ${d}.tar.bz2 $d
 
+      - name: Run Transducer v2
+        shell: bash
+        run: |
+          pushd scripts/nemo/GigaAM
+          ./run-rnnt-v2.sh
+          popd
+
+          d=sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19
+          mkdir $d
+          mkdir $d/test_wavs
+
+          mv -v scripts/nemo/GigaAM/encoder.int8.onnx $d/
+          mv -v scripts/nemo/GigaAM/decoder.onnx $d/
+          mv -v scripts/nemo/GigaAM/joiner.onnx $d/
+
+          cp -v scripts/nemo/GigaAM/*.md $d/
+          cp -v scripts/nemo/GigaAM/LICENSE $d/
+          mv -v scripts/nemo/GigaAM/tokens.txt $d/
+          mv -v scripts/nemo/GigaAM/*.wav $d/test_wavs/
+          mv -v scripts/nemo/GigaAM/run-rnnt-v2.sh $d/
+          cp -v scripts/nemo/GigaAM/test-onnx-rnnt.py $d/
+
+          ls -lh scripts/nemo/GigaAM/
+
+          ls -lh $d
+
+          tar cjvf ${d}.tar.bz2 $d
 
       - name: Release
+        if: github.repository_owner == 'csukuangfj'
         uses: svenstaro/upload-release-action@v2
         with:
           file_glob: true
@@ -114,7 +145,16 @@ jobs:
           repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
           tag: asr-models
 
-      - name: Publish to huggingface (Transducer)
+      - name: Release
+        if: github.repository_owner == 'k2-fsa'
+        uses: svenstaro/upload-release-action@v2
+        with:
+          file_glob: true
+          file: ./*.tar.bz2
+          overwrite: true
+          tag: asr-models
+
+      - name: Publish to huggingface (CTC)
         env:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
         uses: nick-fields/retry@v3
@@ -126,11 +166,66 @@ jobs:
             git config --global user.email "csukuangfj@gmail.com"
             git config --global user.name "Fangjun Kuang"
 
+            d=sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24/
+            export GIT_LFS_SKIP_SMUDGE=1
+            export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
+            git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
+            cp -av $d/* ./huggingface
+            cd huggingface
+            git lfs track "*.onnx"
+            git lfs track "*.wav"
+            git status
+            git add .
+            git status
+            git commit -m "add models"
+            git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d main
+
+      - name: Publish to huggingface (Transducer)
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 5
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
             d=sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24/
             export GIT_LFS_SKIP_SMUDGE=1
             export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
+            git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
+            cp -av $d/* ./huggingface
+            cd huggingface
+            git lfs track "*.onnx"
+            git lfs track "*.wav"
+            git status
+            git add .
+            git status
+            git commit -m "add models"
+            git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d main
+
+      - name: Publish v2 to huggingface (CTC)
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 5
+          timeout_seconds: 200
+          shell: bash
+          command: |
+            git config --global user.email "csukuangfj@gmail.com"
+            git config --global user.name "Fangjun Kuang"
+
+            d=sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19/
+            export GIT_LFS_SKIP_SMUDGE=1
+            export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
             git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
-            mv -v $d/* ./huggingface
+            cp -av $d/* ./huggingface
             cd huggingface
             git lfs track "*.onnx"
             git lfs track "*.wav"
@@ -145,7 +240,7 @@ jobs:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
         uses: nick-fields/retry@v3
         with:
-          max_attempts: 20
+          max_attempts: 5
           timeout_seconds: 200
           shell: bash
           command: |
@@ -155,8 +250,9 @@ jobs:
             d=sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/
             export GIT_LFS_SKIP_SMUDGE=1
             export GIT_CLONE_PROTECTION_ACTIVE=false
+            rm -rf huggingface
             git clone https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$d huggingface
-            mv -v $d/* ./huggingface
+            cp -av $d/* ./huggingface
             cd huggingface
             git lfs track "*.onnx"
             git lfs track "*.wav"

diff --git a/scripts/nemo/GigaAM/README.md b/scripts/nemo/GigaAM/README.md
@@ -7,4 +7,4 @@ to sherpa-onnx.
 The ASR models are for Russian speech recognition in this folder.
 
 Please see the license of the models at
-https://github.com/salute-developers/GigaAM/blob/main/GigaAM%20License_NC.pdf
+https://github.com/salute-developers/GigaAM/blob/main/LICENSE
diff --git a/scripts/nemo/GigaAM/export-onnx-ctc-v2.py b/scripts/nemo/GigaAM/export-onnx-ctc-v2.py
@@ -1,3 +1,4 @@
+#!/usr/bin/env python3
 import gigaam
 import onnx
 import torch
@@ -27,7 +28,13 @@ def add_meta_data(filename: str, meta_data: dict[str, str]):
 
 def main() -> None:
     model_name = "v2_ctc"
-    model = gigaam.load_model(model_name, fp16_encoder=False, use_flash=False, download_root=".")
+    model = gigaam.load_model(
+        model_name, fp16_encoder=False, use_flash=False, download_root="."
+    )
+
+    # use characters
+    # space is 0
+    # <blk> is the last token
     with open("./tokens.txt", "w", encoding="utf-8") as f:
         for i, s in enumerate(model.cfg["labels"]):
             f.write(f"{s} {i}\n")
@@ -53,5 +60,5 @@ def main() -> None:
     )
 
 
-if __name__ == '__main__':
+if __name__ == "__main__":
     main()
diff --git a/scripts/nemo/GigaAM/export-onnx-ctc.py b/scripts/nemo/GigaAM/export-onnx-ctc.py
@@ -82,6 +82,9 @@ def main():
     model.load_state_dict(ckpt, strict=False)
     model.eval()
 
+    # use characters
+    # space is 0
+    # <blk> is the last token
     with open("tokens.txt", "w", encoding="utf-8") as f:
         for i, t in enumerate(model.cfg.labels):
             f.write(f"{t} {i}\n")