ttnn.prod_bw

ttnn.prod_bw(grad_tensor: ttnn.Tensor, input_tensor: ttnn.Tensor, *, dim: int = None, memory_config: ttnn.MemoryConfig = None) → List of ttnn.Tensor

Performs backward operations for prod on input_tensor with given grad_tensor along a particular dim. If no dim is provided, the prod is taken over all dimensions.

Parameters:

• grad_tensor (ttnn.Tensor) – the input gradient tensor.

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

Keyword Arguments:

• dim (int, optional) – dimension to perform prod backward. Defaults to None.

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

Returns:

List of ttnn.Tensor – the output tensor.

Note

Supported dtypes, layouts, and ranks:

Dtypes	Layouts	Ranks
BFLOAT16	TILE	4

For more details about BFLOAT8_B, refer to the BFLOAT8_B limitations.

Example

# Create sample tensors for backward product operation
grad_tensor = ttnn.from_torch(
    torch.rand([1, 1, 32, 32], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device
)
input_tensor = ttnn.from_torch(
    torch.rand([1, 1, 32, 32], dtype=torch.bfloat16, requires_grad=True), layout=ttnn.TILE_LAYOUT, device=device
)
# Define dimension for product operation
dim = 0

# Call the prod_bw function with specific dimension
output = ttnn.prod_bw(grad_tensor, input_tensor, dim=dim)
logger.info(f"Prod Backward (dim={dim}): {output}")

# Call the prod_bw function for all dimensions
all_dims_output = ttnn.prod_bw(grad_tensor, input_tensor)
logger.info(f"Prod Backward (all dims): {all_dims_output}")