Upload WenetSpeech-Wu u2pp ASR models. (k2-fsa#3125)

Author: csukuangfj

.github/workflows/upload-models.yaml

@@ -41,6 +41,29 @@ jobs:
           git config --global user.email "csukuangfj@gmail.com"
           git config --global user.name "Fangjun Kuang"
 
+      - name: WenetSpeech Wu
+        if: true
+        shell: bash
+        run: |
+          git lfs install
+          git clone https://huggingface.co/csukuangfj2/sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-int8-2026-02-03
+          git clone https://huggingface.co/csukuangfj2/sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-2026-02-03
+
+          d=sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-int8-2026-02-03
+          rm -rf $d/.git*
+
+          tar cjfv $d.tar.bz2 $d
+
+          rm -rf $d
+
+          d=sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-2026-02-03
+          rm -rf $d/.git*
+
+          tar cjfv $d.tar.bz2 $d
+          rm -rf $d
+
+          ls -lh *.tar.bz2
+
       - name: Setup tmate session
         if: false
         uses: mxschmitt/action-tmate@v3
@@ -172,7 +195,7 @@ jobs:
           tar cjvf $d.tar.bz2 $d
 
       - name: Collect funasr-nano with LLM float32
-        if: true
+        if: false
         shell: bash
         run: |
           d=sherpa-onnx-funasr-nano-2025-12-30
@@ -804,7 +827,7 @@ jobs:
           mv models/* .
 
       - name: Publish to huggingface
-        if: true
+        if: false
         env:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
         uses: nick-fields/retry@v3
@@ -864,16 +887,7 @@ jobs:
           command: |
             git config --global user.email "csukuangfj@gmail.com"
             git config --global user.name "Fangjun Kuang"
-            models=(
-              sherpa-onnx-funasr-nano-int8-2025-12-30.tar.bz2
-              sherpa-onnx-funasr-nano-fp16-2025-12-30.tar.bz2
-              sherpa-onnx-funasr-nano-2025-12-30.tar.bz2
-            )
-            for m in ${models[@]}; do
-              if [ ! -f $m ]; then
-                continue;
-              fi
-
+            for m in *.tar.bz2; do
               export GIT_LFS_SKIP_SMUDGE=1
               export GIT_CLONE_PROTECTION_ACTIVE=false
 

scripts/apk/generate-vad-asr-apk-script.py

@@ -790,6 +790,38 @@ def get_models():
 
             ls -lh
 
+            popd
+            """,
+        ),
+        Model(
+            model_name="sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-int8-2026-02-03",
+            idx=47,
+            lang="wu",
+            lang2="吴语",
+            short_name="wenetspeech_wu_u2pconformer_ctc_2026_02_03_int8",
+            cmd="""
+            pushd $model_name
+
+            rm -rfv test_wavs
+
+            ls -lh
+
+            popd
+            """,
+        ),
+        Model(
+            model_name="sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-2026-02-03",
+            idx=48,
+            lang="wu",
+            lang2="吴语",
+            short_name="wenetspeech_wu_u2pconformer_ctc_2026_02_03",
+            cmd="""
+            pushd $model_name
+
+            rm -rfv test_wavs
+
+            ls -lh
+
             popd
             """,
         ),

sherpa-onnx/csrc/offline-recognizer-ctc-impl.h

@@ -151,6 +151,7 @@ class OfflineRecognizerCtcImpl : public OfflineRecognizerImpl {
       // WeNet CTC models assume input samples are in the range
       // [-32768, 32767], so we set normalize_samples to false
       config_.feat_config.normalize_samples = false;
+      config_.feat_config.dither = 1;
     }
 
     if (!config_.model_config.medasr.model.empty()) {

sherpa-onnx/kotlin-api/OfflineRecognizer.kt

@@ -779,6 +779,26 @@ fun getOfflineModelConfig(type: Int): OfflineModelConfig? {
             )
         }
 
+        47 -> {
+            val modelDir = "sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-int8-2026-02-03"
+            return OfflineModelConfig(
+                wenetCtc = OfflineWenetCtcModelConfig(
+                    model = "$modelDir/model.int8.onnx",
+                ),
+                tokens = "$modelDir/tokens.txt",
+            )
+        }
+
+        48 -> {
+            val modelDir = "sherpa-onnx-wenetspeech-wu-u2pp-conformer-ctc-zh-2026-02-03"
+            return OfflineModelConfig(
+                wenetCtc = OfflineWenetCtcModelConfig(
+                    model = "$modelDir/model.onnx",
+                ),
+                tokens = "$modelDir/tokens.txt",
+            )
+        }
+
         9000 -> {
             val modelDir =
                 "sherpa-onnx-qnn-5-seconds-sense-voice-zh-en-ja-ko-yue-2024-07-17-int8-android-aarch64"