search

KoljaB / RealtimeSTT

A robust, efficient, low-latency speech-to-text library with advanced voice activity detection, wake word activation and instant transcription.
MIT License
2.09k stars 190 forks source link

Question: verify installation #38

Open jgilbert2017 opened 7 months ago

jgilbert2017 commented 7 months ago

I just installed RealtimeSTT on windows.

I downloaded and installed cudnn and ffmpeg but they are just archives with binaries.

I didn't add anything to my PATH but the realtimestt_test.py seems to work fine.

How do I verify that my installation is correct (what are cuddn/ffmpeg required by)?

Also, is there any direct way to verify that my GPU is being used?

Thanks

KoljaB commented 7 months ago

what are cuddn/ffmpeg required by cudnn coming from faster_whisper, ffmpeg may actually not be needed unless you try some of the examples.

How do I verify that my installation is correct Check if it works and throws no errors.

any direct way to verify that my GPU is being used?

  1. Open Python console: Type python in your command line and press Enter.
  2. Import PyTorch: Type import torch and press Enter.
  3. Check CUDA: Type print(torch.cuda.is_available()) and press Enter.

How it might look in your CLI:

$ python
>>> import torch
>>> print(torch.cuda.is_available())
False
>>> exit()

False = bad

KoljaB commented 7 months ago

If it is False and you have a GPU with CUDA support:

pip install torch==2.2.2+cu118 torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cu118

To install torch support for CUDA 11.8 and

pip install torch==2.2.2+cu121 torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cu121

for CUDA 12.1