ttnn.unary_chain

ttnn.unary_chain(input_tensor: ttnn.Tensor, ops_chain: list[ttnn.UnaryWithParam], *, memory_config: ttnn.MemoryConfig = None, output_tensor: ttnn.Tensor = None) → ttnn.Tensor

Applies unary_chain to input_tensor element-wise.

\[\mathrm{output\_tensor}_i = \verb|unary_chain|(\mathrm{input\_tensor}_i)\]
Parameters:

  • input_tensor (ttnn.Tensor) – the input tensor.

  • ops_chain (list[ttnn.UnaryWithParam]) – list of unary ops to be chained.

Keyword Arguments:

  • memory_config (ttnn.MemoryConfig, optional) – memory configuration for the operation. Defaults to None.

  • output_tensor (ttnn.Tensor, optional) – preallocated output tensor. Defaults to None.

Returns:

ttnn.Tensor – the output tensor.

Note

Supported dtypes, layouts, and ranks:

Dtypes

Layouts

Ranks

BFLOAT16, BFLOAT8_B

TILE

2, 3, 4

Example

# Create a tensor with random normal values
tensor = ttnn.from_torch(torch.randn([32, 32], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device)
ops_chain = [
    ttnn.UnaryWithParam(ttnn.UnaryOpType.RELU),
    ttnn.UnaryWithParam(ttnn.UnaryOpType.EXP, False),
    ttnn.UnaryWithParam(ttnn.UnaryOpType.POWER, 2),
]

# Apply a chain of unary operations
output = ttnn.unary_chain(tensor, ops_chain)
logger.info(f"Unary chain: {output}")