Funasr1.0 (#1284)

* funasr1.0 update * funasr1.0 paraformer-en * update with main (#1281) * Funasr1.0 (#1279) * funasr1.0 update * funasr1.0 paraformer-en * update speaker infer * update device * update device * update raw_text * update infer * update * update infer * bug fix --------- Co-authored-by: shixian.shi <[email protected]> Co-authored-by: Shi Xian <[email protected]> * punc bugfix * reduce_channels --------- Co-authored-by: shixian.shi <[email protected]> Co-authored-by: Shi Xian <[email protected]>
modelscope · Jan 23, 2024 · 2c3183b · 2c3183b
1 parent d8ef0e9
commit 2c3183b
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 2 deletions.
diff --git a/funasr/models/ct_transformer/model.py b/funasr/models/ct_transformer/model.py
@@ -333,12 +333,13 @@ def inference(self,
                 elif new_mini_sentence[-1] == ",":
                     new_mini_sentence_out = new_mini_sentence[:-1] + "."
                     new_mini_sentence_punc_out = new_mini_sentence_punc[:-1] + [self.sentence_end_id]
-                elif new_mini_sentence[-1] != "。" and new_mini_sentence[-1] != "？" and len(new_mini_sentence[-1].encode())==0:
+                elif new_mini_sentence[-1] != "。" and new_mini_sentence[-1] != "？" and len(new_mini_sentence[-1].encode())!=1:
                     new_mini_sentence_out = new_mini_sentence + "。"
                     new_mini_sentence_punc_out = new_mini_sentence_punc[:-1] + [self.sentence_end_id]
                 elif new_mini_sentence[-1] != "." and new_mini_sentence[-1] != "?" and len(new_mini_sentence[-1].encode())==1:
                     new_mini_sentence_out = new_mini_sentence + "."
                     new_mini_sentence_punc_out = new_mini_sentence_punc[:-1] + [self.sentence_end_id]
+
             # keep a punctuations array for punc segment
             if punc_array is None:
                 punc_array = punctuations

diff --git a/funasr/utils/load_utils.py b/funasr/utils/load_utils.py
@@ -39,7 +39,8 @@ def load_audio_text_image_video(data_or_path_or_list, fs: int = 16000, audio_fs:
     if isinstance(data_or_path_or_list, str) and os.path.exists(data_or_path_or_list): # local file
         if data_type is None or data_type == "sound":
             data_or_path_or_list, audio_fs = torchaudio.load(data_or_path_or_list)
-            data_or_path_or_list = data_or_path_or_list[0, :]
+            if kwargs.get("reduce_channels", True):
+                data_or_path_or_list = data_or_path_or_list.mean(0)
         elif data_type == "text" and tokenizer is not None:
             data_or_path_or_list = tokenizer.encode(data_or_path_or_list)
         elif data_type == "image": # undo