[Bug] wrong order of arguments in scaled_dot_product_attention__default

[lilanxiao]

Checklist

• I have searched related issues but cannot get the expected help.
• 2. I have read the FAQ documentation but cannot get the expected help.
• 3. The bug has not been fixed in the latest version.

Describe the bug

The order of arguments scale and is_causal in the multi_head_attention__default function is different from the original PyTorch API in the 2.1 and 2.2 versions.

The mmdeploy's version:

def scaled_dot_product_attention__default(query,
                                          key,
                                          value,
                                          attn_mask=None,
                                          dropout_p=0.,
                                          scale=None,
                                          is_causal=False):

The API in PyTorch 2.1 and 2.2:

def scaled_dot_product_attention(query, key, value, attn_mask=None, dropout_p=0.0, is_causal=False, scale=None) -> torch.Tensor:

This can cause an error at this line since the variable scale becomes True in some cases. Dividing by a boolean value is not allowed.

Reproduction

1. Unable to share my script due to company constraints. But it is not needed anyway because it's so obviously in the source code.
2. No special modifications.

Environment

I don't think it's necessary.

Error traceback

No response