RuntimeError:  INTERNAL ASSERT FAILED at "/opt/pytorch/nvfuser/csrc/transform_iter.cpp":546, please report a bug with repro script to NVFuser at https://github.com/NVIDIA/Fuser/issues. Error during replay, a transformation was called that conflicts with an rfactor call.
Exception raised from BestEffortReplay at /opt/pytorch/nvfuser/csrc/transform_iter.cpp:546 (most recent call first):
frame #0: nvfuser::nvfCheckFail(char const*, char const*, unsigned int, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x103 (0x7f67e51a55cf in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #1: nvfuser::nvfErrorFail(char const*, char const*, unsigned int, char const*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0x62 (0x7f67e5592302 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #2: <unknown function> + 0x9bd40c (0x7f67e5a3640c in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #3: nvfuser::BestEffortReplay::replayCasP(nvfuser::TensorView const*, nvfuser::TensorView const*, long, nvfuser::LogicalDomainMap const&, bool, bool, bool) + 0x683 (0x7f67e5a39eb3 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #4: nvfuser::TransformReplay::replayCasP(nvfuser::TensorView const*, nvfuser::TensorView const*, long, nvfuser::LogicalDomainMap const&, nvfuser::TransformReplayOptions) + 0x21a (0x7f67e5a424ea in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #5: nvfuser::TransformReplay::replayCasP(nvfuser::TensorView const*, nvfuser::TensorView const*, long, nvfuser::TransformReplayOptions) + 0x53 (0x7f67e5a44413 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #6: nvfuser::TransformPropagator::propagateP2C(nvfuser::TensorView*, nvfuser::TensorView*) + 0xe3 (0x7f67e5a44523 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #7: nvfuser::MaxInfoSpanningTree::traverse(nvfuser::MaxInfoSpanningTree::Propagator*) + 0xe2 (0x7f67e59c8ac2 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #8: <unknown function> + 0x91c630 (0x7f67e5995630 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #9: <unknown function> + 0x91d9af (0x7f67e59969af in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #10: <unknown function> + 0x882ef7 (0x7f67e58fbef7 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #11: nvfuser::FusionKernelRuntime::compileFusionParallel(nvfuser::KernelArgumentHolder) + 0x3e8 (0x7f67e58fd7b8 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #12: nvfuser::FusionExecutorCache::runFusionWithInputs(c10::ArrayRef<c10::IValue> const&, std::optional<nvfuser::PrimDataType>, std::optional<signed char>) + 0x1db (0x7f67e58f6dfb in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #13: nvfuser::python_frontend::FusionDefinition::execute(c10::ArrayRef<c10::IValue> const&, std::optional<signed char>, bool, bool, bool) const + 0x15a (0x7f67e5ab45ca in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)

import torch
from nvfuser import FusionDefinition, DataType

def nvfuser_fusion_id38(fd : FusionDefinition) -> None :
    T3 = fd.define_tensor(shape=[1, 2048, 512], contiguity=[None, True, True], dtype=DataType.BFloat16, is_cpu=False, stride_order=[2, 1, 0])
    T33 = fd.ops.reshape(T3, new_shape=[1, 2048, 8, 64])
    T34 = fd.ops.permute(T33, dims=[0, 2, 1, 3])
    T185 = fd.ops.broadcast_in_dim(T34, shape=[1, 8, 1, 2048, 64], broadcast_dims=[0, 1, 3, 4])
    T192 = fd.ops.broadcast_in_dim(T185, shape=[1, 8, 4, 2048, 64], broadcast_dims=[0, 1, 2, 3, 4])
    T198 = fd.ops.reshape(T192, new_shape=[1, 32, 2048, 64])
    fd.add_output(T34)
    fd.add_output(T198)

with FusionDefinition() as fd:
    nvfuser_fusion_id38(fd)

inputs = [
    torch.testing.make_tensor((1, 2048, 512), dtype=torch.bfloat16, device='cuda:0'),
]
fd.execute(inputs)