Add Python APIs for WeNet CTC models (#428)

python-api-examples/non_streaming_server.py

@@ -58,7 +58,19 @@ python3 ./python-api-examples/non_streaming_server.py \
   --nemo-ctc ./sherpa-onnx-nemo-ctc-en-conformer-medium/model.onnx \
   --tokens ./sherpa-onnx-nemo-ctc-en-conformer-medium/tokens.txt
 
-(4) Use a Whisper model
+(4) Use a non-streaming CTC model from WeNet
+
+cd /path/to/sherpa-onnx
+GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/csukuangfj/sherpa-onnx-zh-wenet-wenetspeech
+cd sherpa-onnx-zh-wenet-wenetspeech
+git lfs pull --include "*.onnx"
+cd ..
+
+python3 ./python-api-examples/non_streaming_server.py \
+  --wenet-ctc ./sherpa-onnx-zh-wenet-wenetspeech/model.onnx \
+  --tokens ./sherpa-onnx-zh-wenet-wenetspeech/tokens.txt
+
+(5) Use a Whisper model
 
 cd /path/to/sherpa-onnx
 GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/csukuangfj/sherpa-onnx-whisper-tiny.en
@@ -210,6 +222,15 @@ def add_nemo_ctc_model_args(parser: argparse.ArgumentParser):
     )
 
 
+def add_wenet_ctc_model_args(parser: argparse.ArgumentParser):
+    parser.add_argument(
+        "--wenet-ctc",
+        default="",
+        type=str,
+        help="Path to the model.onnx from WeNet CTC",
+    )
+
+
 def add_tdnn_ctc_model_args(parser: argparse.ArgumentParser):
     parser.add_argument(
         "--tdnn-model",
@@ -261,6 +282,7 @@ def add_model_args(parser: argparse.ArgumentParser):
     add_transducer_model_args(parser)
     add_paraformer_model_args(parser)
     add_nemo_ctc_model_args(parser)
+    add_wenet_ctc_model_args(parser)
     add_tdnn_ctc_model_args(parser)
     add_whisper_model_args(parser)
 
@@ -804,6 +826,7 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
     if args.encoder:
         assert len(args.paraformer) == 0, args.paraformer
         assert len(args.nemo_ctc) == 0, args.nemo_ctc
+        assert len(args.wenet_ctc) == 0, args.wenet_ctc
         assert len(args.whisper_encoder) == 0, args.whisper_encoder
         assert len(args.whisper_decoder) == 0, args.whisper_decoder
         assert len(args.tdnn_model) == 0, args.tdnn_model
@@ -827,6 +850,7 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
         )
     elif args.paraformer:
         assert len(args.nemo_ctc) == 0, args.nemo_ctc
+        assert len(args.wenet_ctc) == 0, args.wenet_ctc
         assert len(args.whisper_encoder) == 0, args.whisper_encoder
         assert len(args.whisper_decoder) == 0, args.whisper_decoder
         assert len(args.tdnn_model) == 0, args.tdnn_model
@@ -842,6 +866,7 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
             decoding_method=args.decoding_method,
         )
     elif args.nemo_ctc:
+        assert len(args.wenet_ctc) == 0, args.wenet_ctc
         assert len(args.whisper_encoder) == 0, args.whisper_encoder
         assert len(args.whisper_decoder) == 0, args.whisper_decoder
         assert len(args.tdnn_model) == 0, args.tdnn_model
@@ -856,6 +881,21 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
             feature_dim=args.feat_dim,
             decoding_method=args.decoding_method,
         )
+    elif args.wenet_ctc:
+        assert len(args.whisper_encoder) == 0, args.whisper_encoder
+        assert len(args.whisper_decoder) == 0, args.whisper_decoder
+        assert len(args.tdnn_model) == 0, args.tdnn_model
+
+        assert_file_exists(args.wenet_ctc)
+
+        recognizer = sherpa_onnx.OfflineRecognizer.from_wenet_ctc(
+            model=args.wenet_ctc,
+            tokens=args.tokens,
+            num_threads=args.num_threads,
+            sample_rate=args.sample_rate,
+            feature_dim=args.feat_dim,
+            decoding_method=args.decoding_method,
+        )
     elif args.whisper_encoder:
         assert len(args.tdnn_model) == 0, args.tdnn_model
         assert_file_exists(args.whisper_encoder)