2e503d1a56
- Add app/diarize.py: merge faster-whisper segments with pyannote (A/B/C) - Wire /api/jobs and /api/transcribe; job API returns speaker_diarization, diarize_skip_reason - UI: meta line shows diarization applied/skipped; hint for models path - requirements.txt: pyannote.audio; README APP_DIARIZE / APP_PYANNOTE_MODEL_DIR - whisper_stt.py: validate config.yaml before loading pipeline - requirements-whisper-stt.txt: minor doc updates if any Made-with: Cursor
49 lines
2.7 KiB
Plaintext
49 lines
2.7 KiB
Plaintext
# 권장: conda activate stt 후 설치 (프로젝트 기본 환경)
|
|
# whisper_stt.py (OpenAI Whisper CLI) — 선택 설치
|
|
#
|
|
# ── Ubuntu/Linux 권장 순서 (torch 깨짐·OSError 방지) ─────────────────────────
|
|
# pyannote가 pip로 torch를 덮어쓰다가 기존 torch와 꼬이면
|
|
# "Could not install ... ATen.h" 같은 오류가 날 수 있습니다.
|
|
#
|
|
# 1) 새 환경에서 PyTorch를 conda로 먼저 깔고, 그다음 pip:
|
|
# conda create -n stt python=3.11 -y
|
|
# conda activate stt
|
|
# conda install pytorch torchaudio cpuonly -c pytorch -y
|
|
# pip install -r requirements-whisper-stt.txt
|
|
#
|
|
# 2) 이미 깨진 경우(위 오류가 난 뒤) 복구:
|
|
# conda activate stt # 또는 ncue 등 해당 env
|
|
# pip uninstall -y torch torchvision torchaudio functorch
|
|
# pip uninstall -y torch torchvision torchaudio functorch # Skipping만 나올 때까지 반복
|
|
# pip cache purge
|
|
# pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu
|
|
# pip install -r requirements-whisper-stt.txt
|
|
#
|
|
# 3) 2)를 해도 "ATen.h" / OSError 가 나면 — site-packages 에 깨진 torch 잔여 디렉터리가 남은 경우:
|
|
# conda activate <env>
|
|
# pip uninstall -y torch torchvision torchaudio functorch 2>/dev/null; true
|
|
# rm -rf "$CONDA_PREFIX/lib/python3.11/site-packages/torch" \
|
|
# "$CONDA_PREFIX/lib/python3.11/site-packages/torch-"*.dist-info \
|
|
# "$CONDA_PREFIX/lib/python3.11/site-packages/torchaudio" \
|
|
# "$CONDA_PREFIX/lib/python3.11/site-packages/torchaudio-"*.dist-info \
|
|
# "$CONDA_PREFIX/lib/python3.11/site-packages/torchgen" \
|
|
# "$CONDA_PREFIX/lib/python3.11/site-packages/functorch"
|
|
# (Python 버전이 3.12면 경로의 python3.11 을 python3.12 로 바꿀 것)
|
|
# pip cache purge
|
|
# pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu
|
|
# pip install -r requirements-whisper-stt.txt
|
|
#
|
|
# 4) 그래도 안 되면: conda create -n stt-whisper python=3.11 -y 후 1)부터 새 env에만 설치
|
|
#
|
|
# GPU(CUDA) 쓰는 경우: https://pytorch.org/get-started/locally/ 에서 맞는 pip/conda 명령 사용
|
|
#
|
|
# CLI: PATH에 hf — pip install huggingface_hub 후 hf auth login
|
|
#
|
|
# 모델: hf download pyannote/speaker-diarization-3.1 --local-dir ./models/pyannote-diarization-3.1
|
|
# (약관 동의·토큰 필요). whisper_stt.py 기본이 이 로컬 폴더 사용.
|
|
# ─────────────────────────────────────────────────────────────────────────────
|
|
openai-whisper
|
|
imageio-ffmpeg
|
|
huggingface_hub>=0.26.0
|
|
pyannote.audio>=3.1.0
|