ttnn.from_device

ttnn.from_device(tensor: ttnn.Tensor, blocking: bool = True, *, queue_id: ttnn.QueueId = None) → ttnn.Tensor

Copy tensor from device to host.

Parameters:

  • tensor (ttnn.Tensor)the tensor to be copied from device to host.

  • blocking (bool, optional)whether the operation should be blocked until the copy is complete. Defaults to True.

Keyword Arguments:

queue_id (ttnn.QueueId, optional)The queue id to use. Defaults to None.

Returns:

ttnn.Tensorthe host tensor copy.

Example

# Open the device
# device_id = 0
# device = ttnn.open_device(device_id=device_id)

# Create a TT-NN tensor on the device and move it back to the host
ttnn_tensor_on_device = ttnn.rand((2, 3), dtype=ttnn.bfloat16, device=device)
ttnn_tensor_on_host = ttnn.from_device(ttnn_tensor_on_device)

logger.info(f"TT-NN tensor shape after moving from device to host: {ttnn_tensor_on_host.shape}")