As per title.
I assume that when the second kernel is programmed onto the npu (using NPU1Col1() as device) the first one gets overwritten so that when it's called something does not match and the error is thrown.
Tried using AIEDevice.npu1_1col but does not work ("Not Resolvable" something error). According to this file NPUXColY it's just a class with a resolve method that returns the corresponding npuX_Ycol.
Maybe related to Xilinx/mlir-aie#2302. According to a comment under that issue, this PR should fix the problem. According to another comment under the same issue, the fix has already been pushed bu I cannot test it until the toolchain on aupcloud.io is updated (currenttly is about one month old).