Practical overview
When it fits
Use Mmdnn when a model workflow crosses supported deep-learning frameworks or when you need to visualize a network locally. Check the specific framework direction and model guidance first, particularly for experimental or destination-only support.
Who it is for
- Developers and researchers working with deep-learning models in different frameworks.
- Teams considering model migration between supported frameworks.
- Users who want a local view of a model's network architecture.
Documented uses
- Convert a model from one supported framework to another.
- Inspect a deep-learning model's network architecture with the local visualizer.
- Use conversion-generated code snippets as a starting point for retraining or inference.
Requirements & limits
The public GitHub repository identifies an MIT license. The README documents installation through pip and a Docker image.
- ONNX is listed as a destination only.
- TensorFlow conversion is marked experimental; DarkNet is source-only and experimental.
- The README reports testing on some ImageNet models, not universal support for all models.
- The tool is distributed for local use through Python installation or Docker rather than presented as a hosted conversion service.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install MMdnn with pip
Install MMdnn using the stable package or the repository’s master branch.
Before you start
- Python installed.
Steps
For the stable version, run the command below.
pip install mmdnnAlternatively, to try the newest version, run the command below.
pip install -U git+https://github.com/Microsoft/MMdnn.git@masterFor Docker setup instead, follow the repository’s Docker instructions: install Docker Community Edition, pull
mmdnn/mmdnn:cpu.small, then start it interactively with the command below.docker run -it mmdnn/mmdnn:cpu.small
Expected outcome
MMdnn is installed through pip, or its CPU-small Docker image is running interactively.
Guide limitations
- The README does not specify Python version requirements.
- The documented Docker image is
cpu.small; no other image variants are specified here.
Convert the documented TensorFlow ResNet V2 152 model to PyTorch
Download the example TensorFlow ResNet V2 152 model and convert it to PyTorch format.
Before you start
- MMdnn installed and its command-line tools available.
- The example uses TensorFlow as the source framework and PyTorch as the destination.
Steps
Download the model into the current directory with
mmdownload -f tensorflow -n resnet_v2_152 -o ./.Convert it with
mmconvert -sf tensorflow -in imagenet_resnet_v2_152.ckpt.meta -iw imagenet_resnet_v2_152.ckpt --dstNodeName MMdnn_Output -df pytorch -om tf_resnet_to_pth.pth.
Expected outcome
The documented conversion targets the PyTorch output file `tf_resnet_to_pth.pth`.
Guide limitations
- The repository notes that TensorFlow support is experimental.
- This example covers a specific ResNet V2 152 model and does not establish compatibility for every model.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 12 October 2026
- Original website title
- microsoft/MMdnn · Official homepage title · retained in its source language