在算能转换yolov26-pose模型bug

[vvwcap]

Cannot open libusercpu.so, disable user cpu layer.
in cmodel, enable profile.
in cmodel, enable profile.
in cmodel, enable profile.
%4 = "tpu.Cast"(%1) {round_mode = #tpu<round_mode HalfAwayFromZero>, with_scale = true} : (tensor<1x300x1xi32, 4467228672 : i64>) -> tensor<1x300x1xsi32, 4467224576 : i64> loc("/model.23/Expand_output_0_Unsqueeze/model.23/GatherElements_output_0_GatherElements_si32")
CastOp type error
UNREACHABLE executed at /__w/tpu-mlir/tpu-mlir/lib/Dialect/Tpu/Interfaces/BM1684/Cast.cpp:51!

[harrisonGPU]

Please describe the problem in English. Are you sure it is an LLVM issue?

[vvwcap]

Please describe the problem in English. Are you sure it is an LLVM issue?

This is a YOLO model conversion with post-processing on a Sophon device. When it reaches "model_deploy", it gives me the error message: "CastOp type error UNREACHABLE executed at /__w/tpu-mlir/tpu-mlir/lib/Dialect/Tpu/Interfaces/BM1684/Cast.cpp:51!" Then it tells me to "PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace." I apologize if the bug report I submitted at the link above is not related to this module.

[harrisonGPU]

I think you need to update the title in English and explain how to reproduce this issue in LLVM, or provide a small LLVM IR example that reproduces it. You should determine whether this is an LLVM issue before creating a report, rather than opening an issue as soon as you encounter a problem. You should spend some time analyzing the problem instead of wasting our time.

[vvwcap]

um, you're right, thank you.