Reduce warning noise caused by Tensor.new_tensor

[st81]

What does this PR do?

This PR replaces Tensor.new_tensor() calls to suppress warning that users cannot address. This allows users to avoid being confused by warnings they cannot act on, making it easier for them to focus on meaningful warnings.

More specifically, when calling the EncoderDecoderModel forward method, users see warning like:

from transformers import EncoderDecoderModel, AutoTokenizer

tok = AutoTokenizer.from_pretrained("bert-base-uncased")
model = EncoderDecoderModel.from_encoder_decoder_pretrained("bert-base-uncased", "bert-base-uncased")
model.config.decoder_start_token_id = tok.cls_token_id
model.config.pad_token_id = tok.pad_token_id

src = tok("hello world", return_tensors="pt")
tgt = tok("hi there", return_tensors="pt").input_ids
labels = tgt.clone()
labels[labels == tok.pad_token_id] = -100

out = model(
    input_ids=src["input_ids"],
    attention_mask=src["attention_mask"],
    labels=labels,
)

/home/shutotakahashi/projects/transformers-uv/transformers/src/transformers/models/encoder_decoder/modeling_encoder_decoder.py:453:
UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than tensor.new_tensor(sourceTensor).
  decoder_attention_mask = decoder_input_ids.new_tensor(decoder_input_ids != self.config.pad_token_id)

This warning is triggered by the use of Tensor.new_tensor() in the internal code and cannot be resolved by users.

The fix is functionally equivalent because new_tensor() creates a new tensor with the same dtype as the original tensor. Additionally, I think the fixed version is more semantically clear as it explicitly shows that decoder_attention_mask is created through tensor comparison followed by type casting.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

• text models: @ArthurZucker @Cyrilvallez

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: encoder_decoder

[Rocketknight1]

Yeah, this is a more idiomatic way to do it, too. Thank you!

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.