[Bug] When there is a negative number in index, 'mindspore.ops.index_select' should throw an error message

[Redmept1on]

Environment

Hardware Environment(Ascend/GPU/CPU):

device gpu

Software Environment:

• MindSpore version (source or binary): v2.2.13
• Python version (e.g., Python 3.7.5): 3.8.18
• OS platform and distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 18.04

Describe the current behavior

[[7. 8. 9.]]

Describe the expected behavior

IndexError: index out of range in self

Steps to reproduce the issue

import mindspore as mind
import numpy as np

input_tensor = mind.tensor(np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]], dtype=np.float32))

index_tensor = mind.tensor([-1])

output_tensor = mind.ops.index_select(input_tensor, axis=-2, index=index_tensor)

print(output_tensor)

pytorch v2.0.0+cu117

import torch
import numpy as np

input_tensor = torch.tensor(np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]))

index_tensor = torch.tensor([-1])

output_tensor = torch.index_select(input_tensor, axis=-2, index=index_tensor)

print(output_tensor)

Special notes for this issue

The index parameter lacks a range check

[Ash-Lee233]

did you test the case on cpu? we test the case on gpu, it reports the same info as torch but the cpu gave [[7,8,9]], we will fix it in the next version

[Redmept1on]

Thanks for your quick response, sorry for wrong information, after verification I found that my code was running on the CPU.