requirements-semantic-apple.txt

# Semantic Clustering Dependencies - Apple Silicon (M1/M2/M3)
#
# This file installs semantic clustering dependencies for Apple Silicon
# Provides 3-5x performance improvement over CPU-only processing on Apple Silicon
#
# IMPORTANT: Install PyTorch with MPS support FIRST, then this file!
#
# Step 1: Install PyTorch with MPS support
# pip install torch torchvision torchaudio
#
# Step 2: Install this requirements file
# pip install -r requirements-semantic-apple.txt
#
# Hardware Requirements:
# - Mac: Apple Silicon (M1, M1 Pro, M1 Max, M1 Ultra, M2, M2 Pro, M2 Max, M2 Ultra, M3, M3 Pro, M3 Max)
# - macOS: 12.3+ (for MPS support)
# - Memory: 8GB+ unified memory (16GB+ recommended)
# - Storage: ~800MB for dependencies + 2-10GB for models
#
# Performance:
# - Speed: 3-5x faster than CPU on Apple Silicon
# - Models: All models supported with appropriate memory
# - Compatibility: Apple Silicon Macs with macOS 12.3+

# =============================================================================
# Apple Silicon Compatibility
# =============================================================================
# 
# Supported Macs:
# ✅ MacBook Air M1/M2/M3
# ✅ MacBook Pro M1/M1 Pro/M1 Max/M2/M2 Pro/M2 Max/M3/M3 Pro/M3 Max
# ✅ iMac M1/M3
# ✅ Mac mini M1/M2/M2 Pro
# ✅ Mac Studio M1 Max/M1 Ultra/M2 Max/M2 Ultra
# ✅ Mac Pro M2 Ultra
#
# Requirements:
# - macOS 12.3+ for MPS acceleration
# - Older macOS versions will fall back to CPU

# =============================================================================
# Semantic Clustering Dependencies (PyTorch with MPS must be installed first!)
# =============================================================================

# Main semantic clustering library
sentence-transformers>=2.2.0

# Transformer models and tokenizers
transformers>=4.30.0

# Model acceleration and optimization
accelerate>=0.20.0

# Safe tensor format for faster model loading
safetensors>=0.3.0

# =============================================================================
# Optional Performance Enhancements
# =============================================================================

# Optimized model inference (uncomment if needed)
# optimum>=1.12.0

# =============================================================================
# Installation Notes
# =============================================================================

# Total Download Size: ~800MB (optimized for Apple Silicon)
# 
# Unified Memory Requirements by Model:
# ✅ gte-large: 8GB+ memory
# ✅ bge-large-en-v1.5: 8GB+ memory  
# ✅ sentence-t5-xl: 16GB+ memory
# ✅ Qwen-0.6B: 12GB+ memory
# ✅ Qwen-4B: 24GB+ memory
# ✅ Qwen-8B: 32GB+ memory
#
# Performance Tips:
# - Monitor memory usage with Activity Monitor
# - Close other memory-intensive applications
# - Apple Silicon shares memory between CPU and GPU
# - Use gte-large or bge-large for best compatibility
# - Larger models benefit more from MPS acceleration
#
# Troubleshooting:
# - If MPS not available: update to macOS 12.3+
# - If out of memory: use smaller model or close other apps
# - If slow performance: ensure MPS is being used (check logs)
# - For Intel Macs: use requirements-semantic-cpu.txt instead
#
# Checking MPS Support:
# Run this in Python to verify MPS is available:
# import torch
# print(f"MPS available: {torch.backends.mps.is_available()}")
# print(f"MPS built: {torch.backends.mps.is_built()}")
#
# Apple Silicon Advantages:
# - Unified memory architecture (shared CPU/GPU memory)
# - Excellent power efficiency
# - Good performance for most models
# - Native ARM64 optimization