Test failed with Error replaying transforms in contiguous ID checker, expected iS10{9} to be in the active ID set.

[xwang233]

Repro:

import torch
from nvfuser import FusionDefinition, DataType

def nvfuser_fusion_id5(fd : FusionDefinition) -> None :
    T0 = fd.define_tensor(shape=[1, 1], contiguity=[None, None], dtype=DataType.Int, is_cpu=False, stride_order=[1, 0])
    S1 = fd.define_scalar(-2, dtype=DataType.Int)
    T7 = fd.ops.pad(T0, [0, 2, 0, 2], S1)
    fd.add_output(T7)

with FusionDefinition() as fd:
    nvfuser_fusion_id5(fd)

inputs = [
    torch.testing.make_tensor((1, 1), dtype=torch.int64, device='cuda:0'),
]
fd.execute(inputs)

Stacktrace:

Error replaying transforms in contiguous ID checker, expected iS10{9} to be in the active ID set.
Exception raised from checkExclusivelyConsumesAllocs at /opt/pytorch/nvfuser/csrc/contiguity.cpp:51 (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 (0x7f3d6bb6010f 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 (0x7f3d6bf9fea2 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)
frame #2: <unknown function> + 0x3dd516 (0x7f3d6bdc0516 in /opt/pytorch/nvfuser/nvfuser/_C.cpython-312-x86_64-linux-gnu.so)

Env:

• pjnl-20250218
• H100

This also caused thunder test to fail

pytest -vsx thunder/tests/test_ops.py -k test_core_vs_jax_consistency_pad_nvfuser_cuda_thunder

[naoyam]

Thanks @xwang233. Do you know if it's a new test that had been working before or is it a new test?

[xwang233]

The thunder test failure looks new and we don't have enough data for it.