File I/O & Data Persistence

Binary Formats (.npy & .npz)

Fast, efficient storage designed specifically for NumPy:

# Save single array
np.save('data.npy', arr)
arr = np.load('data.npy')

# Save multiple arrays (compressed)
np.savez_compressed('data.npz', 
                     features=X, 
                     labels=y)

# Load from .npz
loaded = np.load('data.npz')
X = loaded['features']
y = loaded['labels']

Advantages of .npy/.npz

• Fast (binary format)
• Preserves dtype (int32 stays int32)
• Preserves shape (no ambiguity)
• Compressed option available

Text Formats (CSV, TSV)

Human-readable, shareable:

# Save as CSV
np.savetxt('data.csv', arr, delimiter=',')

# Load from CSV
arr = np.loadtxt('data.csv', delimiter=',')

When to Use Each

Production tip: Use .npz for internal storage, CSV for sharing with non-Python tools.