Add detailed ConvBERT model card with usage, architecture, and refere…

[Aesha19]

What does this PR do?

This PR adds a detailed and standardized model card for ConvBERT to improve Hugging Face Transformers documentation.

Includes:

• Model Overview and Architecture
• Training objective and dataset details
• Use cases and limitations
• Code usage examples via pipeline, AutoModel, and CLI
• Quantization and AttentionMaskVisualizer support
• Benchmarks and citation

File added:

• src/transformers/models/convbert/modelcard.md

This contribution helps improve model discoverability and provides users with accessible and actionable information about ConvBERT.

cc: @stevhliu (documentation reviewer)

---

Fixes: N/A

[Aesha19]

Hi! This is my first contribution — just checking in kindly for a review
cc: @stevhliu

[stevhliu]

Thanks, this is a nice start but I think you should revisit the template in the issue to make sure your model card follows the same format and language! As an example, take a look at the BERT docs :)

[stevhliu]

You don't have to remove this

Suggested change

	<!-- ConvBERT model card -->
	<!--Copyright 2020 The HuggingFace Team. All rights reserved.
	Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
	the License. You may obtain a copy of the License at
	http://www.apache.org/licenses/LICENSE-2.0
	Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
	an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
	specific language governing permissions and limitations under the License.
	⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
	rendered properly in your Markdown viewer.
	-->

[stevhliu]

Missing the TensorFlow badge and this should go above # ConvBERT

[stevhliu]

Suggested change

	---
	## Model Overview

[stevhliu]

Suggested change

	ConvBERT is a lightweight and efficient NLP transformer model introduced by YituTech. It improves on the classic BERT architecture by incorporating **span-based dynamic convolutions** into the self-attention mechanism. This hybrid approach enables ConvBERT to model both local and global dependencies more effectively while reducing the computational cost.
	[ConvBERT](https://huggingface.co/papers/2008.02496) incorporates a mixed attention block that makes it more efficient than [BERT](./bert). Attention is costly because it models global word relationships. This is inefficient because some heads only learn local word relationships. ConvBERT replaces some of the attention heads with a convolution head to handle this. The result of this new mixed attention design is a more lightweight model with lower training costs without compromising performance.
	Instead of using attention heads everywhere to model , ConvBERT also includes convolution heads to model local word relationships.
	is a lightweight and efficient NLP transformer model introduced by YituTech. It improves on the classic BERT architecture by incorporating **span-based dynamic convolutions** into the self-attention mechanism. This hybrid approach enables ConvBERT to model both local and global dependencies more effectively while reducing the computational cost.