LiSsHhUuAaIi opened a new issue, #18438:
URL: https://github.com/apache/tvm/issues/18438
### Description
When converting a PyTorch model containing index assignment operations
(`tensor[:, indices] = other_tensor`) using `torch.export`, TVM encounters a
segmentation fault during the `from_exported_program` conversion step.
### Expected behavior
The PyTorch model should be successfully converted to TVM Relax module
without segmentation faults, and the compiled TVM model should produce the same
output as the original PyTorch model.
### Actual behavior
A segmentation fault occurs during the `from_exported_program` call when
converting the exported PyTorch model to TVM Relax module. The crash happens in
TVM's internal Tuple object creation.
### Environment
* **OS:** Ubuntu 20.04.6 LTS
* **TVM version:** 0.23.dev0
* **Python version:** 3.11.14
### Steps to reproduce
```python
import torch
import torch.nn as nn
import tvm
from tvm import relax
class TestModel(nn.Module):
def __init__(self):
super().__init__()
def forward(self, indices: torch.Tensor, tensor_1: torch.Tensor,
tensor_2: torch.Tensor):
tensor_1[:, indices] = tensor_2
return tensor_1
model = TestModel()
model.eval()
indices = torch.tensor([0, 2, 4])
tensor_1 = torch.randn(5, 6)
tensor_2 = torch.randn(5, 3)
inputs = (indices, tensor_1, tensor_2)
exported_program = torch.export.export(model, inputs)
from tvm.relax.frontend.torch import from_exported_program
# Segmentation fault occurs here
mod = from_exported_program(exported_program)
```
### Error Log
```
!!!!!!! Segfault encountered !!!!!!!
...
...
Segmentation fault (core dumped)
```
### Triage
* needs-triage
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
