Fatal Abort with Core Dump on Index Out of Bounds with randperm and Tensor Reshaping

[LiSsHhUuAaIi]

Summary

I encountered a fatal program termination with core dump when running a model that should normally throw a catchable IndexError for out-of-bounds indexing. The issue occurs in eager mode when using flow.randperm with tensor reshaping, but instead of raising a proper Python exception that can be handled, OneFlow terminates the entire process with Aborted (core dumped).

Code to reproduce bug

import oneflow as flow
import oneflow.nn as nn

class TestModel(flow.nn.Module):
    def __init__(self):
        super(TestModel, self).__init__()

    def forward(self, x):
        perm = flow.randperm(x.numel())
        x = x.view(-1, 128)
        return x[perm[:64]]

x = flow.randn(32, 128)
model = TestModel()

try:
    # This should raise a catchable IndexError but instead aborts
    result = model(x)
    print("Success:", result)
except Exception as e:
    print("Caught exception:", e)
    print("Program continues normally...")

print("This line should be reachable but isn't due to abort")

Error Logs

terminate called after throwing an instance of 'oneflow::Exception'
  what():  Check failed: (nd_index[i] < dense_shape[i] && nd_index[i] >= -dense_shape[i]) IndexError: index 1495 is out of bounds for dimension 0 with size 32
  ...
  ...
Aborted (core dumped)

System Information

• OS: Ubuntu 20.04.6 LTS
• OneFlow version: '1.0.0.dev20251101+cpu'
• Python version: Python 3.10.19