config-gpu.ini

# GPU Acceleration Configuration for Name Matching
# Copy this to config.ini and modify as needed

[database]
host = localhost
port = 3306
user = root
password = 
database = name_matching
pool_size = 5
max_overflow = 10
pool_timeout = 30
pool_recycle = 3600

[gpu]
# Enable or disable GPU acceleration
enabled = true

# Preferred GPU framework (auto, cupy, torch, numba)
# auto = automatically select best available framework
# cupy = Use CuPy (recommended for best performance)
# torch = Use PyTorch
# numba = Use Numba CUDA
framework = auto

# GPU device ID to use (0 for first GPU, 1 for second, etc.)
device_id = 0

# Batch size for GPU processing
# Larger values use more GPU memory but may be faster
# Adjust based on your GPU memory capacity
batch_size = 1000

# GPU memory limit in GB
# The system will try to stay under this limit
memory_limit_gb = 4.0

# Fallback threshold - use CPU for datasets smaller than this
# This avoids GPU overhead for small datasets
fallback_threshold = 10000

[matching]
# Matching thresholds
match_threshold = 0.75
non_match_threshold = 0.55

# Name component weights
first_name_weight = 0.4
middle_name_weight = 0.2
last_name_weight = 0.3
full_name_sorted_weight = 0.1

# Additional field weights
birthdate_weight = 0.3
geography_weight = 0.3

[logging]
level = INFO
format = %(asctime)s - %(name)s - %(levelname)s - %(message)s