OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Bottlenecktransformers

Bottlenecktransformers is a Python and PyTorch implementation of Bottleneck Transformers for visual recognition, including ResNet-style model examples and a multi-head self-attention module.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Bottlenecktransformers — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when examining or adapting a PyTorch Bottleneck Transformer for visual-recognition work, or when consulting the included image-classification experiment table.

Who it is for

  • Computer-vision developers exploring PyTorch model architectures.
  • Researchers comparing attention-head configurations in image-classification experiments.

Documented uses

  • Inspect a ResNet-style Bottleneck Transformer model interface.
  • Use an example multi-head self-attention module in a visual-recognition model.
  • Review the repository's reported classification results across model variants.
Before you choose

Requirements & limits

The repository is public and specifies the MIT license.

  • The examples show model construction and a forward-pass shape, but not a complete training or deployment workflow.
  • The experiment table reports specific results and includes a work-in-progress entry; it is not a general performance guarantee.
  • The implementation uses Python and PyTorch.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Use the ResNet50 model example

Instantiate the documented ResNet50 model and pass it a sample tensor.

Before you start

  • The project’s Python model code is available.
  • PyTorch and the names used in the example, including torch, are available in the Python environment.

Steps

  1. Import Model from model, as shown in the README.

  2. Instantiate ResNet50(num_classes=1000, resolution=(224, 224)).

  3. Create a sample tensor with torch.randn([2, 3, 224, 224]).

  4. Pass the tensor to the model and print model(x).size().

Expected outcome

The example prints the size of the model’s output tensor.

Guide limitations

  • The README’s import line names Model, while its instantiation line uses ResNet50; it does not explain this discrepancy.
  • The example does not provide installation instructions, dataset handling, training, or evaluation steps.
Documentation 1 ↗
Source-based practical guide

Instantiate the MHSA module example

Create the documented multi-head self-attention module with a resolution of 14.

Before you start

  • The project’s Python model code is available.
  • The planes value required by the constructor is defined.

Steps

  1. Import MHSA from model.

  2. Set resolution = 14.

  3. Instantiate MHSA(planes, width=resolution, height=resolution).

Expected outcome

An `MHSA` module object is instantiated using the supplied channel and spatial dimensions.

Guide limitations

  • The README does not define planes or show how to integrate the module into another model.
  • It does not document training or evaluation steps.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
leaderj1001/BottleneckTransformers · Official homepage title · retained in its source language
How we review sourcesSuggest a correction