Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
auto frontend
游雁8ba12df6f7f4
diff --git a/funasr/models/llm_asr/model.py b/funasr/models/llm_asr/model.pyindex e2880c831..66a69eb2c 100644--- a/funasr/models/llm_asr/model.py+++ b/funasr/models/llm_asr/model.py@@ -483,8 +483,9 @@ class LLMASR2(nn.Module):_, l, _ = encoder_out.shapefor batch_idx in range(batch_size):fbank_beg_idx = fbank_beg[batch_idx, 0].item()- inputs_embeds[batch_idx, fbank_beg_idx : fbank_beg_idx + l, :] = encoder_out[- batch_idx, :l, :+ min_len = min(l, inputs_embeds.shape[1] - fbank_beg_idx)+ inputs_embeds[batch_idx, fbank_beg_idx : fbank_beg_idx + min_len, :] = encoder_out[+ batch_idx, :min_len, :]labels_ids[labels_ids == -1] = -100model_outputs = self.llm(