Purfview comments

Results 268 comments of


                                            Purfview

Faster Whisper large-v2 model is repeating the segments

Do you get repeats with settings I used? Only differences from yours were `device=cpu` and `vad_filter=False`.

Empty sequence when using faster-whisper's transcribe on fine-tuned model

```python segments, info = model.transcribe('foo.wav', condition_on_previous_text=False) ```

Empty sequence when using faster-whisper's transcribe on fine-tuned model

Try this: ```python segments, info = model.transcribe('foo.wav', condition_on_previous_text=False, without_timestamps=True) ```

Update to CUDA 12.8

Completely forgot, the PR includes https://github.com/OpenNMT/CTranslate2/pull/1920 , I think, merge his PR first to keep attribution.

Update to CUDA 12.8

> Have you tried the new wheels under Linux and Windows on a machine with a CUDA GPU? Yes, on both, Linux and Windows with the CUDA GPUs.

Update to CUDA 12.8

Removed it (Note: CI will fail without it)

>I get this error when I use CUDA: >Could not load symbol cudnnGetLibConfig. Error: /lib/x86_64-linux-gnu/libcudnn_graph.so.9: undefined symbol: cudnnGetLibConfig Check your Nvidia driver, CUDA/CUDA toolkit versions. Maybe needs to be updated.

itertracks() doesn't work properly in 0.1.1

> You could try sherpa-onnx with pyannote-audio. I tried it but it doesn't have "min_duration_on"/"min_duration_off" implemented too, btw, it works very well, almost same output as original pyannote segmentation. Another...

itertracks() doesn't work properly in 0.1.1

> step = self.duration // 2 Tested `step = self.duration // 10` it got slower only ~10%

Feature request - Support CUDA 12.x with onnxruntime-gpi 1.21.x

Why do you need to run it on GPU? Silero ran much slower on GPU [RTX 4070] than on CPU when I tested it.