PyTorch
API: torch.cuda.is_available
Strategy: Direct Mapping
JAX (Core)
API: —
Strategy: Macro 'len(jax.devices('gpu')) > 0'
TensorFlow
API: —
Strategy: Macro 'len(tf.config.list_physical_devices('GPU')) > 0'
Apple MLX
API: mlx.core.is_available
Strategy: Direct Mapping