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

Msdnet

Msdnet is an open-source Lua Torch implementation of a convolutional image-classification network with classifiers at intermediate depths. It supports progressively updating predictions as computation continues and allocating a fixed computation budget unevenly across examples in a batch.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Msdnet — 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 the repository when you need research code and documented recipes for exploring image classification with intermediate predictions or batch computation budgets.

Who it is for

  • Researchers exploring resource-efficient image classification.
  • Developers evaluating progressive prediction or adaptive computation strategies.
  • Practitioners benchmarking image-classification accuracy under constrained computation.

Documented uses

  • Train a CIFAR-10 MSDNet configured for progressive anytime prediction.
  • Train a CIFAR-10 model configured for uneven computation across a batch.
  • Evaluate pretrained MSDNet checkpoints using ImageNet data and validation indices.
  • Study classification results under limited or fixed computation budgets.
Before you choose

Requirements & limits

The repository is public and identifies its license as MIT.

  • The documented implementation uses Lua Torch and is written under the Torch ResNet framework.
  • The repository provides CIFAR-10 training recipes and references pretrained ImageNet checkpoints.
  • The README describes a separate PyTorch implementation in another repository.

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

Train an MSDNet for anytime prediction on CIFAR-10

Use the repository’s documented configuration for a network with classifiers attached to every other layer.

Before you start

  • The MSDNet repository and its Lua Torch ResNet framework.
  • CIFAR-10 training data.

Steps

  1. Run th main.lua --help to list the training script options.

  2. Use the documented recipe: th main.lua -netType msdnet -dataset cifar10 -batchSize 64 -nEpochs 300 -nBlocks 10 -stepmode even -step 2 -base 4.

Expected outcome

The documented configuration trains an MSDNet with 10 classifiers for anytime prediction.

Guide limitations

  • The repository recommends using a validation set for model selection and describes holding out 5,000 CIFAR training images.
  • The guide reflects the Lua Torch repository’s recipe; it does not cover the separate PyTorch implementation.
Documentation 1 ↗
Source-based practical guide

Test a pretrained MSDNet checkpoint on ImageNet

Follow the documented checkpoint-based ImageNet testing workflow.

Before you start

  • A downloaded pretrained model checkpoint and validation-set indices.
  • ImageNet data and the repository’s Lua Torch testing script.

Steps

  1. Download the model checkpoints and validation-set indices referenced in the README.

  2. Run the documented test command with the paths to the model, ImageNet data and validation indices: th main.lua -dataset imagenet -testOnly true -resume -data -gen .

Expected outcome

The repository’s testing script evaluates the selected checkpoint using the supplied ImageNet data and validation indices.

Guide limitations

  • The README does not specify checkpoint performance or provide a result for a particular local run.
  • The command requires substituting paths for the checkpoint, dataset and validation indices.
Documentation 1 ↗
Evidence and transparency

Sources & review

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