index_put : INTERNAL ASSERT FAILED

[kshitij12345]

🐛 Describe the bug

Previous related issues: #39227 #66495

import torch

# device = 'cpu'  # Works

# RuntimeError: linearIndex.numel()*sliceSize*nElemBefore == expandedValue.numel()
# INTERNAL ASSERT FAILED at
# "../aten/src/ATen/native/cuda/Indexing.cu":265, please report a bug to PyTorch.
# number of flattened indices did not match number of elements in the value tensor: 10 vs 5
device = 'cuda'  # Errors
t = torch.zeros(5, 5, device=device)
idx = torch.tensor([0, 1], device=device)
v = torch.ones((5,), device=device)

print(torch.index_put(t, (idx, ), v, accumulate=True))

Versions

master

cc @ezyang @gchanan @zou3519 @ngimel

[VitalyFedyunin]

high pri because similar to #66495

[bkmi]

It seems that this issue is not fixed yet. Any possibility to pull in that fix?

FYI for those who find themselves here, take a look at this workaround:
#104217 (comment)