nvfuser_bench is broken

[zasdfgbnm]

🐛 Describe the bug

./build/bin/nvfuser_bench --benchmark_filter=.*Transpose_Random_fp16_Inner_2D_01_Axis.*

terminate called after throwing an instance of 'c10::Error'
  what():  !mismatch INTERNAL ASSERT FAILED at "/home/gaoxiang/nvfuser/torch/csrc/jit/codegen/cuda/executor_utils.cpp":354, please report a bug to PyTorch. Found one or more invalid arguments: Argument element type is Half, but the parameter is float
Argument element type is Half, but the parameter is float

Exception raised from validateKernelInputs at /home/gaoxiang/nvfuser/torch/csrc/jit/codegen/cuda/executor_utils.cpp:354 (most recent call first):
frame #0: <unknown function> + 0x85b50 (0x7f3dd037ab50 in /home/gaoxiang/nvfuser/build/lib/libc10.so)
frame #1: <unknown function> + 0x85ae0 (0x7f3dd037aae0 in /home/gaoxiang/nvfuser/build/lib/libc10.so)
frame #2: <unknown function> + 0x859e0 (0x7f3dd037a9e0 in /home/gaoxiang/nvfuser/build/lib/libc10.so)
frame #3: <unknown function> + 0x87c48 (0x7f3dd037cc48 in /home/gaoxiang/nvfuser/build/lib/libc10.so)
frame #4: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x65 (0x7f3dd037b265 in /home/gaoxiang/nvfuser/build/lib/libc10.so)
frame #5: c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x7a (0x7f3dd0378c0a in /home/gaoxiang/nvfuser/build/lib/libc10.so)
frame #6: c10::detail::torchInternalAssertFail(char const*, char const*, unsigned int, char const*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x5d (0x7f3dd0378e7d in /home/gaoxiang/nvfuser/build/lib/libc10.so)
frame #7: <unknown function> + 0x6604a15 (0x7f3df0426a15 in /home/gaoxiang/nvfuser/build/lib/libtorch_cuda.so)
frame #8: torch::jit::fuser::cuda::FusionExecutor::runFusion(c10::ArrayRef<c10::IValue> const&, std::vector<at::Tensor, std::allocator<at::Tensor> > const&, torch::jit::fuser::cuda::LaunchParams const&, c10::optional<unsigned long> const&) + 0x1385 (0x7f3df03967b5 in /home/gaoxiang/nvfuser/build/lib/libtorch_cuda.so)
frame #9: <unknown function> + 0x6772350 (0x7f3df0594350 in /home/gaoxiang/nvfuser/build/lib/libtorch_cuda.so)
frame #10: torch::jit::fuser::cuda::FusionKernelRuntime::runKernelWithInput(c10::ArrayRef<c10::IValue> const&, unsigned long, torch::jit::fuser::cuda::SegmentedGroup*) + 0x7a8 (0x7f3df0584af8 in /home/gaoxiang/nvfuser/build/lib/libtorch_cuda.so)
frame #11: torch::jit::fuser::cuda::FusionKernelRuntime::runWithInput(c10::ArrayRef<c10::IValue> const&, unsigned long) + 0x747 (0x7f3df0581ca7 in /home/gaoxiang/nvfuser/build/lib/libtorch_cuda.so)
frame #12: torch::jit::fuser::cuda::FusionExecutorCache::runFusionWithInputs(c10::ArrayRef<c10::IValue> const&) + 0x5a3 (0x7f3df0580db3 in /home/gaoxiang/nvfuser/build/lib/libtorch_cuda.so)
frame #13: runBenchmarkIterations(benchmark::State&, torch::jit::fuser::cuda::FusionExecutorCache*, std::vector<c10::IValue, std::allocator<c10::IValue> >&) + 0x68 (0x55623e2b1c88 in ./build/bin/nvfuser_bench)
frame #14: <unknown function> + 0x15e080 (0x55623e299080 in ./build/bin/nvfuser_bench)
frame #15: NF_Transpose_Random_fp16_Inner_2D_01_Axis___GRAPH_NF_Transpose_Random_fp16_Inner_2D_01_Axis_Benchmark::BenchmarkCase(benchmark::State&) + 0x7d (0x55623e2975ed in ./build/bin/nvfuser_bench)
frame #16: <unknown function> + 0xdbb60 (0x55623e216b60 in ./build/bin/nvfuser_bench)
frame #17: benchmark::internal::BenchmarkInstance::Run(unsigned long, int, benchmark::internal::ThreadTimer*, benchmark::internal::ThreadManager*, benchmark::internal::PerfCountersMeasurement*) const + 0x82 (0x55623e32b6c2 in ./build/bin/nvfuser_bench)
frame #18: <unknown function> + 0x1cbda9 (0x55623e306da9 in ./build/bin/nvfuser_bench)
frame #19: benchmark::internal::BenchmarkRunner::DoNIterations() + 0x351 (0x55623e306851 in ./build/bin/nvfuser_bench)
frame #20: benchmark::internal::BenchmarkRunner::DoOneRepetition() + 0xd8 (0x55623e307528 in ./build/bin/nvfuser_bench)
frame #21: <unknown function> + 0x17ea7c (0x55623e2b9a7c in ./build/bin/nvfuser_bench)
frame #22: benchmark::RunSpecifiedBenchmarks(benchmark::BenchmarkReporter*, benchmark::BenchmarkReporter*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x5a6 (0x55623e2b8ca6 in ./build/bin/nvfuser_bench)
frame #23: benchmark::RunSpecifiedBenchmarks() + 0x39 (0x55623e2b86a9 in ./build/bin/nvfuser_bench)
frame #24: main + 0x5a (0x55623e2b714a in ./build/bin/nvfuser_bench)
frame #25: <unknown function> + 0x29290 (0x7f3db6de6290 in /usr/lib/libc.so.6)
frame #26: __libc_start_main + 0x8a (0x7f3db6de634a in /usr/lib/libc.so.6)
frame #27: _start + 0x25 (0x55623e213f75 in ./build/bin/nvfuser_bench)

Versions

devel