#148148th of 168 in ML frameworks and librariesFor science, health and research, research and papers
mish
Reference code and benchmarks for the Mish activation function
- Stars
- 1.3k
- License
- MIT
- Last commit
- Jul 2026
- Language
- Jupyter Notebook
1,295 on 21 Apr 20261,301 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Research repository for Mish, the activation f(x) = x·tanh(softplus(x)), from the BMVC 2020 paper. It collects ImageNet and MS-COCO results, loss landscape visualizations, DarkNet cfg files with pretrained weights, PyTorch benchmarks, and a list of frameworks that adopted Mish. It is a reference and benchmark collection, not an installable library.
Who it is for: Researchers and engineers evaluating activation functions in vision models
Strengths
- DarkNet cfg files and weights for ResNet-50, DarkNet-53, DenseNet-201 and ResNeXt-50 with Mish
- Ablation tables compare Mish with Swish, ReLU and Leaky ReLU on several CSP networks
- Links a PyTorch benchmarks folder, a CUDA version and a Weights & Biases integration
- Lists framework adoptions, including PyTorch, MXNet, TF.js and OpenVINO
Weaknesses
- Not a packaged library; Mish is now built into major frameworks
- Pretrained weights are hosted on Google Drive links, which can break
- Benchmarks date from 2019-2020 and use older architectures
- README is mostly changelog and media coverage, with little usage guidance
What it needs
- GPU optional
- Needs DarkNet (AlexeyAB), PyTorch
- Models: DarkNet cfg, PyTorch
Quick answers
Is mish open source?
Yes. Its code is on GitHub under the MIT license.
Does mish need a GPU?
No. It runs on a CPU and can use a GPU when you have one.
Also in ML frameworks and libraries
See all 168| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd | 93 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts | 93 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 92 out of 100 |