iAura_1 / requirements.txt
akpande2's picture
Update requirements.txt
7bc0b42 verified
# ==============================================================================
# CALL CENTER ANALYTICS - UPDATED REQUIREMENTS
# ==============================================================================
# Core ML/DL Frameworks
torch==2.5.1
torchaudio==2.5.1
numpy
scipy
# Audio Processing
librosa
praat-parselmouth
# Speech-to-Text
openai-whisper
# Diarization & Voice Analysis
pyannote.audio
# Transformers & NLP
transformers==4.46.1
accelerate
# NEW: Keyword Extraction
keybert
sentence-transformers
# NEW: Additional ML utilities (required by KeyBERT)
scikit-learn
# Web Framework & API
gradio
uvicorn
fastapi
python-multipart
# ==============================================================================
# INSTALLATION NOTES:
# ==============================================================================
# 1. Basic installation:
# pip install -r requirements.txt
#
# 2. For GPU support (recommended):
# - Install CUDA 11.8 or 12.1
# - PyTorch should auto-detect CUDA
#
# 3. System dependencies (Ubuntu/Debian):
# sudo apt-get update
# sudo apt-get install -y ffmpeg
#
# 4. Hugging Face Token:
# - Required for pyannote.audio diarization
# - Set as environment variable: HF_TOKEN=your_token_here
# - Get token from: https://huggingface.co/settings/tokens
# - Accept pyannote terms: https://huggingface.co/pyannote/speaker-diarization-3.1
#
# ==============================================================================
# ESTIMATED SIZES:
# ==============================================================================
# Total download size: ~6-7 GB
# - Whisper Large-v3: 2.9 GB
# - Pyannote 3.1: 500 MB
# - BART-MNLI: 1.6 GB
# - Other models: ~2 GB
#
# GPU VRAM Requirements:
# - Minimum: 8 GB (will work but slow)
# - Recommended: 10-12 GB (smooth operation)
# - Optimal: 16 GB+ (batch processing)
#
# ==============================================================================