xujiashuai/timm-speech_bert_dialogue-detetction_speaker-diarization_chinese
Model Introduction
Files & versions
Pull Requests
Discussion
Analysis

BERT Dialogue Detection Model (speech_bert_dialogue-detetction_speaker-diarization_chinese)

BERT-based model for dialogue detection in speaker diarization scenarios.

  • Model ID: iic/speech_bert_dialogue-detetction_speaker-diarization_chinese
  • Task: Speaker Diarization - Dialogue Detection
  • Framework: PyTorch
  • Input: Chinese text (transcribed speech)
  • Output: Binary classification logits (2 classes)

Verification Environment

ItemValue
NPUAscend910B4
CANN8.5.1
PyTorch2.9.0+cpu
torch_npu2.9.0.post1
Python3.11.14

Accuracy Evaluation

CPU vs NPU cosine similarity:

MetricValue
Cosine Similarity1.000000

Inference Output

Logits shape: [1, 2]
Logits sample: [2.089, -2.814]
SUCCESS

Reproduction

python3 inference.py
python3 evaluate.py

Materials

  • inference.py - NPU inference script
  • evaluate.py - CPU vs NPU cosine similarity evaluation