From 8b74979bd38c70c731e04b32308f2c9edeabdc37 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Tue, 16 Jul 2024 13:57:51 +0800
Subject: [PATCH] sensevoice

---
 .../sense_voice/demo.py                       | 72 ++++++++++++++++---
 funasr/auto/auto_model.py                     |  7 +-
 2 files changed, 70 insertions(+), 9 deletions(-)

diff --git a/examples/industrial_data_pretraining/sense_voice/demo.py b/examples/industrial_data_pretraining/sense_voice/demo.py
index 1800d7591..f10e54b45 100644
--- a/examples/industrial_data_pretraining/sense_voice/demo.py
+++ b/examples/industrial_data_pretraining/sense_voice/demo.py
@@ -3,26 +3,82 @@
 # Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
 #  MIT License  (https://opensource.org/licenses/MIT)
 
-import sys
+
 from funasr import AutoModel
 from funasr.utils.postprocess_utils import rich_transcription_postprocess
 
 model_dir = "iic/SenseVoiceSmall"
-input_file = (
-    "https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav"
-)
+
 
 model = AutoModel(
     model=model_dir,
+    vad_model="fsmn-vad",
+    vad_kwargs={"max_single_segment_time": 30000},
+    device="cpu",
 )
 
+# en
 res = model.generate(
-    input=input_file,
+    input=f"{model.model_path}/example/en.mp3",
     cache={},
     language="auto",  # "zn", "en", "yue", "ja", "ko", "nospeech"
-    use_itn=False,
+    use_itn=True,
+    batch_size_s=60,
+    merge_vad=True,  #
+    merge_length_s=15,
+)
+text = rich_transcription_postprocess(res[0]["text"])
+print(text)
+
+# zh
+res = model.generate(
+    input=f"{model.model_path}/example/zh.mp3",
+    cache={},
+    language="auto",  # "zn", "en", "yue", "ja", "ko", "nospeech"
+    use_itn=True,
+    batch_size_s=60,
+    merge_vad=True,  #
+    merge_length_s=15,
+)
+text = rich_transcription_postprocess(res[0]["text"])
+print(text)
+
+# yue
+res = model.generate(
+    input=f"{model.model_path}/example/yue.mp3",
+    cache={},
+    language="auto",  # "zn", "en", "yue", "ja", "ko", "nospeech"
+    use_itn=True,
+    batch_size_s=60,
+    merge_vad=True,  #
+    merge_length_s=15,
+)
+text = rich_transcription_postprocess(res[0]["text"])
+print(text)
+
+# ja
+res = model.generate(
+    input=f"{model.model_path}/example/ja.mp3",
+    cache={},
+    language="auto",  # "zn", "en", "yue", "ja", "ko", "nospeech"
+    use_itn=True,
+    batch_size_s=60,
+    merge_vad=True,  #
+    merge_length_s=15,
+)
+text = rich_transcription_postprocess(res[0]["text"])
+print(text)
+
+
+# ko
+res = model.generate(
+    input=f"{model.model_path}/example/ko.mp3",
+    cache={},
+    language="auto",  # "zn", "en", "yue", "ja", "ko", "nospeech"
+    use_itn=True,
+    batch_size_s=60,
+    merge_vad=True,  #
+    merge_length_s=15,
 )
-
 text = rich_transcription_postprocess(res[0]["text"])
-
 print(text)
diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.py
index 1c2d2e48f..a82f6ed75 100644
--- a/funasr/auto/auto_model.py
+++ b/funasr/auto/auto_model.py
@@ -339,7 +339,9 @@ class AutoModel:
         #  FIX(gcf): concat the vad clips for sense vocie model for better aed
         if kwargs.get("merge_vad", False):
             for i in range(len(res)):
-                res[i]["value"] = merge_vad(res[i]["value"], kwargs.get("merge_length", 15000))
+                res[i]["value"] = merge_vad(
+                    res[i]["value"], kwargs.get("merge_length_s", 15) * 1000
+                )
 
         # step.2 compute asr model
         model = self.model
@@ -380,6 +382,9 @@ class AutoModel:
             if len(sorted_data) > 0 and len(sorted_data[0]) > 0:
                 batch_size = max(batch_size, sorted_data[0][0][1] - sorted_data[0][0][0])
 
+            if kwargs["device"] == "cpu":
+                batch_size = 0
+
             beg_idx = 0
             beg_asr_total = time.time()
             time_speech_total_per_sample = speech_lengths / 16000