Closed faku99 closed 1 year ago
The easiest way to do that is using our imperative DSL.
As we have released a new version of HeteroCL that entirely moves the compilation stack to MLIR, issues related to the original HalideIR flow will be closed. Please check out the latest HeteroCL and see if the issue still exists.
I can't find a way to implement
bias_add
operator.Let's say I have a
Tensor
with a shape of(1, 8, 28, 28)
and anotherTensor
(which contains biases) with a shape(8, 1, 1)
. How could I add the biases to the firstTensor
?TVM implements such an operator with C++ but I can't find a way to implement it with the
hcl.compute
API. Any thoughts?