Trixy3/requirements-ml.txt

# Trixy Voice Assistant - Machine Learning Requirements
# ====================================================
# Dependencies for ML training, model development, and advanced audio processing
# Includes the main requirements.txt as foundation

# Include base requirements
-r requirements.txt

# Advanced PyTorch ecosystem
torch>=1.11.0,<2.0.0      # Already in base, ensuring compatibility
torchaudio>=0.11.0,<1.0.0  # Already in base, ensuring compatibility
torchvision>=0.12.0,<1.0.0 # Computer vision utilities (may be used for spectrograms)

# ONNX support for model export/import
onnx>=1.12.0               # ONNX model format support
onnxruntime>=1.12.0        # ONNX runtime for inference
onnx-simplifier>=0.4.0     # ONNX model optimization

# Scientific computing and data analysis
numpy>=1.21.0,<2.0.0       # Already in base, ensuring compatibility
scipy>=1.7.0               # Scientific computing (signal processing, optimization)
pandas>=1.3.0              # Data manipulation and analysis

# Machine learning utilities
scikit-learn>=1.0.0        # Already in base, expanded ML utilities
joblib>=1.1.0              # Parallel processing and model persistence

# Audio processing and signal analysis
librosa>=0.9.0             # Advanced audio analysis library
soundfile>=0.10.3          # Already in base, audio file I/O
audioread>=2.1.0           # Audio file reading backend for librosa
resampy>=0.2.2             # High-quality audio resampling

# Voice Activity Detection (VAD)
webrtcvad>=2.0.0           # WebRTC VAD implementation
silero-vad>=4.0.0          # Silero VAD models

# Data visualization for training analysis
matplotlib>=3.5.0          # Plotting and visualization
seaborn>=0.11.0            # Statistical data visualization
plotly>=5.0.0              # Interactive plotting
tensorboard>=2.8.0         # Training visualization and monitoring

# Model training utilities
tqdm>=4.60.0               # Progress bars for training loops
wandb>=0.12.0              # Weights & Biases experiment tracking (optional)
mlflow>=1.24.0             # ML experiment tracking and model registry

# Data augmentation
audiomentations>=0.30.0    # Audio data augmentation
albumentations>=1.1.0      # Image augmentations (for spectrograms)

# Hyperparameter optimization
optuna>=3.0.0              # Hyperparameter optimization framework
ray[tune]>=2.0.0           # Distributed hyperparameter tuning

# Model compression and optimization
torch-pruning>=1.0.0       # Neural network pruning
torch-distillation>=0.1.0  # Knowledge distillation (if available)

# CUDA utilities (for GPU acceleration)
# These are typically installed with PyTorch, but listed for clarity
# cupy-cuda11x>=10.0.0     # CUDA-accelerated NumPy (optional, uncomment if needed)
# numba>=0.56.0            # JIT compilation for numerical functions

# Memory optimization
pympler>=0.9               # Memory profiling and analysis
memory-profiler>=0.60.0    # Already in base, memory monitoring

# Distributed training support
torch-distributed>=0.1.0   # Distributed training utilities (if available)

# Model serving and deployment
fastapi>=0.75.0            # Web API framework for model serving
uvicorn>=0.17.0            # ASGI server for FastAPI
pydantic>=1.8.0,<2.0.0     # Already in base, data validation

# Configuration management for ML experiments
hydra-core>=1.1.0          # Configuration management for ML experiments
omegaconf>=2.1.0           # Configuration system for Hydra

# Time series analysis (for audio sequences)
statsmodels>=0.13.0        # Statistical models and time series analysis

# Parallel processing
multiprocess>=0.70.0       # Better multiprocessing
concurrent.futures>=3.1.1; python_version<"3.9"  # Already in base

# Audio codec support
pyaudio>=0.2.11            # Real-time audio I/O (may need system dependencies)
portaudio19>=19.6.0        # PortAudio backend for PyAudio

# Voice recognition specific libraries
speechbrain>=0.5.0         # SpeechBrain toolkit for speech processing
transformers>=4.15.0       # Hugging Face transformers (for advanced models)
datasets>=2.0.0            # Hugging Face datasets

# Advanced signal processing
pyroomacoustics>=0.6.0     # Room acoustics simulation
pystoi>=0.3.0              # Short-time objective intelligibility measure
pesq>=0.0.3                # Perceptual evaluation of speech quality

# GPU memory management
gpustat>=1.0.0             # GPU monitoring
nvidia-ml-py3>=7.352.0     # NVIDIA GPU management

# Advanced optimization
torch-optimizer>=0.3.0     # Additional optimizers for PyTorch
ranger-fm>=1.9.0           # Ranger optimizer (if available)

# Feature engineering
featuretools>=1.0.0        # Automated feature engineering
category_encoders>=2.3.0   # Categorical data encoding

# Model interpretation and analysis
shap>=0.40.0               # Model explanation and interpretation
lime>=0.2.0                # Local interpretable model explanations
captum>=0.5.0              # Model interpretability for PyTorch

# Evaluation metrics
torchmetrics>=0.7.0        # PyTorch metrics collection
evaluate>=0.3.0            # Hugging Face evaluation library

# Data loading and preprocessing
webdataset>=0.2.0          # Efficient data loading for large datasets
ffcv>=1.0.0                # Fast computer vision data loading (if applicable)

# Audio format conversion
pydub>=0.25.0              # Audio manipulation and format conversion
ffmpeg-python>=0.2.0       # FFmpeg Python bindings

# Annotation and labeling tools
label-studio-sdk>=0.0.20   # Label Studio SDK for data annotation

# Advanced audio features
pyworld>=0.3.0             # WORLD vocoder for voice analysis
praat-parselmouth>=0.4.0   # Praat phonetics software interface

# Speaker diarization
pyannote.audio>=2.0.0      # Speaker diarization and voice activity detection
pyannote.core>=4.5.0       # Core utilities for pyannote
pyannote.database>=4.1.0   # Database utilities for pyannote
pyannote.metrics>=3.2.0    # Evaluation metrics for pyannote

# Optional: Advanced deep learning frameworks
# lightning>=1.5.0         # PyTorch Lightning for structured training
# timm>=0.6.0              # PyTorch image models (for vision-based features)

# Voice synthesis (TTS) - if needed for training data generation
espnet>=202207             # ESPnet speech processing toolkit (optional)

# Real-time processing
rtaudio>=0.1.0             # Real-time audio processing

# System dependencies note:
# Some packages may require system-level dependencies:
# - PortAudio (for pyaudio)
# - FFmpeg (for audio processing)
# - CUDA toolkit (for GPU acceleration)
# - Intel MKL or OpenBLAS (for optimized linear algebra)

# Installation command examples:
# pip install -r requirements-ml.txt
# pip install -r requirements-ml.txt --find-links https://download.pytorch.org/whl/torch_stable.html