Dispatch from llama2 lowering error.

[bailuan]

What happened?

I applied a little old version of iree/compiler/src/iree/compiler/DispatchCreation/ConvertTensorToFlow.cpp and some other code modification to avoid dispatch fuse when lowering. One of dispatches can't be lowered. The mlir and the detailed log are uploaded as below files.dispatch_6_log.txt
dispatch_6.txt

Steps to reproduce your issue

iree-compile --mlir-disable-threading --iree-hal-target-device=__device_0=cuda --mlir-print-ir-after-all --iree-hal-dump-executable-intermediates-to=/workspace/test/ --iree-hal-dump-executable-configurations-to=/workspace/test/ /workspace/test/dispatch_6.mlir -o /workspace/test/dispatch_6.vmfb 2> /
workspace/test/dump.mlir

What component(s) does this issue relate to?

No response

Version information

iree version: f27feff
llvm version: ac8bb7353a7fe79cd99b3c041d5a153517c31abc

Additional context

No response