From 3a382bc3c87b22054c74cbac381817eb8c4b820e Mon Sep 17 00:00:00 2001
From: goddamnVincent <1262780247@qq.com>
Date: Tue, 3 Dec 2024 20:02:43 +0800
Subject: [PATCH] 'update20241203'

---
 python-api-examples/streaming_server.py | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/python-api-examples/streaming_server.py b/python-api-examples/streaming_server.py
index 5691f67c3..ef6bac8e3 100755
--- a/python-api-examples/streaming_server.py
+++ b/python-api-examples/streaming_server.py
@@ -229,6 +229,28 @@ def add_hotwords_args(parser: argparse.ArgumentParser):
         --hotwords-file is given.
         """,
     )
+    parser.add_argument(
+        "--modeling-unit",
+        type=str,
+        default='cjkchar',
+        help="""
+        The modeling unit of the used model. Current supported units are:
+        - cjkchar(for Chinese)
+        - bpe(for English like languages)
+        - cjkchar+bpe(for multilingual models)
+        """,
+    )
+    parser.add_argument(
+        "--bpe-vocab",
+        type=str,
+        default='',
+        help="""
+        The bpe vocabulary generated by sentencepiece toolkit. 
+        It is only used when modeling-unit is bpe or cjkchar+bpe.
+        if you can’t find bpe.vocab in the model directory, please run:
+        python script/export_bpe_vocab.py --bpe-model exp/bpe.model
+        """,
+    )
 
 
 def add_modified_beam_search_args(parser: argparse.ArgumentParser):
@@ -409,6 +431,8 @@ def create_recognizer(args) -> sherpa_onnx.OnlineRecognizer:
             rule2_min_trailing_silence=args.rule2_min_trailing_silence,
             rule3_min_utterance_length=args.rule3_min_utterance_length,
             provider=args.provider,
+            modeling_unit=args.modeling_unit,
+            bpe_vocab=args.bpe_vocab
         )
     elif args.paraformer_encoder:
         recognizer = sherpa_onnx.OnlineRecognizer.from_paraformer(