https://github.com/wozeparrot/tinygrad-image-models
tinygrad image models, inference, and (soon) training
https://github.com/wozeparrot/tinygrad-image-models
cnn imagenet tinygrad vit
Last synced: 2 months ago
JSON representation
tinygrad image models, inference, and (soon) training
- Host: GitHub
- URL: https://github.com/wozeparrot/tinygrad-image-models
- Owner: wozeparrot
- License: apache-2.0
- Created: 2024-03-20T06:21:32.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-03-25T01:27:22.000Z (about 1 year ago)
- Last Synced: 2025-02-08T22:13:00.551Z (4 months ago)
- Topics: cnn, imagenet, tinygrad, vit
- Language: Python
- Homepage:
- Size: 545 KB
- Stars: 2
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# tinygrad Image Models
A collection of vision models implemented in tinygrad, in a similar vein to [timm](https://github.com/huggingface/pytorch-image-models).
Mostly targeting models trained on Imagenet-1k, and other models that are fast on resource-constrained devices.
## Models
- ShuffleNetV2 - [paper](https://arxiv.org/abs/1807.11164) [code](/tgim/models/shufflenetv2.py)
- GhostNetV2 - [paper](https://arxiv.org/abs/2211.12905) [code](/tgim/models/ghostnetv2.py)
- FocalNet - [paper](https://arxiv.org/abs/2203.11926) [code](/tgim/models/focalnet.py)
- FastViT - [paper](https://arxiv.org/abs/2303.14189) [code](/tgim/models/fastvit.py)
- RepViT - [paper](https://arxiv.org/abs/2307.09283) [code](/tgim/models/repvit.py)
## TODO
- [ ] For models that can be reparameterized, add that functionality
- [ ] Training
## License
See [LICENSE](/LICENSE).
Certain parts of the code are adapted from the original implementations, but they should all be under permissive licenses.