This repository adapts the official pyannote/segmentation-3.0 checkpoint for a real Ascend NPU smoke run. The official Hugging Face page and Google search were rechecked on 2026-08-22. The model is MIT licensed, accepts 10 seconds of mono 16 kHz audio, and produces a 7-class segmentation matrix.
pyannote.audio 3.0.0 PyanNet architecture and checkpoint state dictionary.cpu, npu, and auto device selection./tmp.The official checkpoint loaded on Ascend910 npu:0 after loading the CANN environment:
| Metric | CPU | NPU (Ascend910 npu:0) |
|---|---|---|
| Parameters | 1,473,265 | 1,473,265 |
| Input shape | [1,1,160000] | [1,1,160000] |
| Output shape | [1,589,7] | [1,589,7] |
| Output dtype | float32 | float32 |
| Forward | 0.298 s | 3.725 s |
| Finite output | true | true |
CPU/NPU smoke comparison cosine is 0.9999998212, max absolute difference is 0.0020027161, and mean absolute difference is 0.0009162365. This is an NPU execution smoke result only, as requested for the remaining competition time. It is not a complete task-level precision gate.
Source the CANN environment before using --device npu:
source /usr/local/Ascend/ascend-toolkit/set_env.sh
python inference.py --model-path /tmp/model-agent/pyannote19/pytorch_model.bin --device npuOfficial model: https://huggingface.co/pyannote/segmentation-3.0
Official training/runtime version: pyannote.audio==3.0.0.


