Skip to content

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

auto frontend

游雁ddd4ad419bd1
diff --git a/funasr/auto/auto_frontend.py b/funasr/auto/auto_frontend.py
index 3d6db8d8c..696a51e08 100644
--- a/funasr/auto/auto_frontend.py
+++ b/funasr/auto/auto_frontend.py
@@ -87,8 +87,16 @@ class AutoFrontend:
speech_lengths.sum().item() * self.frontend.frame_shift * self.frontend.lfr_n / 1000
)
- speech.to(device=device), speech_lengths.to(device=device)
- batch = {"input": speech, "input_len": speech_lengths, "key": key_batch}
+ if kwargs.get("return_pt", True):
+ speech, speech_lengths = speech.to(device=device), speech_lengths.to(device=device)
+ else:
+ speech, speech_lengths = speech.numpy(), speech_lengths.numpy()
+ batch = {
+ "input": speech,
+ "input_len": speech_lengths,
+ "key": key_batch,
+ data_type: "fbank",
+ }
result_list.append(batch)
pbar.update(1)