Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

Dev gzf (#1435)

zhifu gao294c1162df01
diff --git a/examples/industrial_data_pretraining/qwen_audio/demo.py b/examples/industrial_data_pretraining/qwen_audio/demo.py
index 41bbb9977..04a8e58e9 100644
--- a/examples/industrial_data_pretraining/qwen_audio/demo.py
+++ b/examples/industrial_data_pretraining/qwen_audio/demo.py
@@ -9,5 +9,8 @@ from funasr import AutoModel
model = AutoModel(model="Qwen-Audio")
-res = model.generate(input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav", language=None)
+audio_in = "https://github.com/QwenLM/Qwen-Audio/raw/main/assets/audio/1272-128104-0000.flac"
+prompt = "<|startoftranscription|><|en|><|transcribe|><|en|><|notimestamps|><|wo_itn|>"
+
+res = model.generate(input=audio_in, prompt=prompt)
print(res)
diff --git a/examples/industrial_data_pretraining/qwen_audio/demo_from_local.py b/examples/industrial_data_pretraining/qwen_audio/demo_from_local.py
index 54ae1f023..40cf4b941 100644
--- a/examples/industrial_data_pretraining/qwen_audio/demo_from_local.py
+++ b/examples/industrial_data_pretraining/qwen_audio/demo_from_local.py
@@ -11,5 +11,8 @@ model = AutoModel(model="Qwen-Audio",
model_path="/nfs/zhifu.gzf/init_model/qwen/Qwen-Audio",
)
-res = model.generate(input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav", language=None)
+audio_in = "https://github.com/QwenLM/Qwen-Audio/raw/main/assets/audio/1272-128104-0000.flac"
+prompt = "<|startoftranscription|><|en|><|transcribe|><|en|><|notimestamps|><|wo_itn|>"
+
+res = model.generate(input=audio_in, prompt=prompt)
print(res)
diff --git a/funasr/models/qwen_audio/model.py b/funasr/models/qwen_audio/model.py
index e419b1ef0..86b410520 100644
--- a/funasr/models/qwen_audio/model.py
+++ b/funasr/models/qwen_audio/model.py
@@ -54,9 +54,8 @@ class QwenAudioWarp(nn.Module):
meta_data = {}
# meta_data["batch_data_time"] = -1
-
- sp_prompt = "<|startoftranscription|><|en|><|transcribe|><|en|><|notimestamps|><|wo_itn|>"
- query = f"<audio>{data_in[0]}</audio>{sp_prompt}"
+ prompt = kwargs.get("prompt", "<|startoftranscription|><|en|><|transcribe|><|en|><|notimestamps|><|wo_itn|>")
+ query = f"<audio>{data_in[0]}</audio>{prompt}"
audio_info = self.tokenizer.process_audio(query)
inputs = self.tokenizer(query, return_tensors='pt', audio_info=audio_info)
inputs = inputs.to(self.model.device)