Skip to content

Troubleshooting

float16 error on CPU

Requested float16 compute type, but the target device or backend do not support efficient float16 computation.

Fix: Add --compute-type float32:

speech-mine extract audio.wav out.csv \
  --hf-token TOKEN \
  --compute-type float32

Safest command for any system:

speech-mine extract audio.wav out.csv \
  --hf-token TOKEN \
  --device cpu \
  --compute-type float32 \
  --model base

pyannote model access error

Failed to load pyannote pipeline

Fix: Make sure you have:

  1. A valid HuggingFace token with read permissions
  2. Accepted the user agreement at pyannote/speaker-diarization-3.1

If speech-mine search returns no results, try lowering the similarity range:

speech-mine search "your query" output.csv --similarity-range 0.5 1.0

The default range is 0.0 1.0 (all matches), but if no results appear it may be that no windows score above 0.0 — try the query with simpler wording.