ttnn

External Resources

• TTNN Documentation
• TT-Metal Repository
• TTNN API Reference -- full operation catalog (~400+ ops organized by category)
• Tensor Reference -- shapes, layouts, memory configs, data types
• Converting PyTorch Models to TTNN -- step-by-step conversion guide
• Multi-Device Programming -- MeshDevice, tensor parallelism, data parallelism, CCL ops
• Tensor Sharding -- height, width, and block sharding strategies

Multi-Device

TTNN natively supports multi-chip execution via the MeshDevice abstraction. See multi_device.md for full details.

# Single device
device = ttnn.open_device(device_id=0, trace_region_size=100000000)