ttnn.reallocate

ttnn.reallocate(tensor: ttnn.Tensor, memory_config: ttnn.MemoryConfig = None) → ttnn.Tensor

reallocate(tensor: ttnn._ttnn.tensor.Tensor, memory_config: ttnn._ttnn.tensor.MemoryConfig | None = None) -> ttnn._ttnn.tensor.Tensor

Deallocates device tensor and returns a reallocated tensor.

Parameters:

tensor (ttnn.Tensor) – the input tensor.
memory_config (ttnn.MemoryConfig, optional) – memory configuration for the reallocated tensor. Defaults to None.

Returns:

ttnn.Tensor – the reallocated tensor.

Example

# Create a TT-NN tensor, deallocate it, and then reallocate it
ttnn_tensor = ttnn.rand((2, 3), dtype=ttnn.bfloat16, device=device)
ttnn_tensor_reallocated = ttnn.reallocate(ttnn_tensor)

logger.info("Reallocated TT-NN tensor shape", ttnn_tensor_reallocated.shape)