ops.equal或==异步算子导致判断语句出错

| name       | about	                             | labels   |
| ---------- | ------------------------------------- | -------- |
| Bug Report | ops.equal或==异步算子导致判断语句出错 |  mindspore-assistant |

### 问题描述
执行语句：
```python
special_audio_token_mask = input_ids == self.config.audio_token_index
# 或者
special_audio_token_mask = ops.equal(input_ids, self.config.audio_token_index)
```
如果在该行加上断点时，`input_ids == self.config.audio_token_index`判断正确，正确赋值给special_audio_token_mask，但是当不加断点时，判断出现异常，具体如图所示:

![Issue](https://foruda.gitee.com/images/1712841806308896219/1a20cfa2_14290977.png "issue.png")

此时不加断点的话，special_audio_token_mask赋值异常，即`input_ids == self.config.audio_token_index`语句出错
可能是由于异步算子问题导致
 
### Environment / 环境信息 (Mandatory / 必填)
- **硬件环境**:  CPU

- **Software Environment**:
-- MindSpore version: 2.2.11
-- Python version: 3.9.0
-- Windows 11
-- Python Intepreter

- **Excute Mode**: `Pynative`
> /mode pynative
 
### Related log
不加断点时错误日志:
```
Traceback (most recent call last):
  File "c:\Users\lenovo\Desktop\DialectCLIP\src\model.py", line 801, in <module>
    outputs = model.generate(**inputs, generation_config=generation_config)
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindnlp\utils\generic.py", line 341, in wrapper
    outputs = func(*args, **kwargs)
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindnlp\transformers\generation\utils.py", line 1509, in generate
    return self.greedy_search(
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindnlp\transformers\generation\utils.py", line 2331, in greedy_search
    outputs = self(
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindspore\nn\cell.py", line 705, in __call__
    raise err
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindspore\nn\cell.py", line 701, in __call__
    output = self._run_construct(args, kwargs)
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindspore\nn\cell.py", line 482, in _run_construct
    output = self.construct(*cast_inputs, **kwargs)
  File "c:\Users\lenovo\Desktop\DialectCLIP\src\model.py", line 693, in construct
    = self._merge_input_ids_with_audio_features(audio_features, inputs_embeds=inputs_embeds, input_ids=input_ids, attention_mask=attention_mask, labels=labels)
  File "c:\Users\lenovo\Desktop\DialectCLIP\src\model.py", line 406, in _merge_input_ids_with_audio_features
    max_seq_length = num_special_audio_tokens_per_seq.max().item() * (audio_seq_length - 1) + text_seq_length
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindspore\common\tensor.py", line 880, in item
    output = tensor_operator_registry.get('item')(self, index)
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindspore\ops\composite\multitype_ops\_compile_utils.py", line 320, in tensor_item   
    return data.asnumpy().item()
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindspore\common\_stub_tensor.py", line 49, in fun
    return method(*arg, **kwargs)
  File "D:\Program\Anaconda3\envs\mindspore\lib\site-packages\mindspore\common\tensor.py", line 964, in asnumpy
    return Tensor_.asnumpy(self)
RuntimeError: For 'Gather', the 'input_indices' should be in the range [0, 151646), but got 151646, error_code[58]
```

截图：
![Issue](https://foruda.gitee.com/images/1712841806308896219/1a20cfa2_14290977.png "issue.png")

MindSpore Lab/mindnlp

内容风险标识

评论 (1)

MindSpore Lab/mindnlp .gitee-modal { width: 500px !important; }

内容风险标识