|
|
--- |
|
|
datasets: |
|
|
- ylecun/mnist |
|
|
metrics: |
|
|
- accuracy |
|
|
pipeline_tag: image-classification |
|
|
library_name: torch |
|
|
--- |
|
|
|
|
|
# LeNet |
|
|
|
|
|
Toy model mainly used to showcase Aidge in various tutorial, for example: https://eclipse.dev/aidge/source/Tutorial/101_first_step.html |
|
|
|
|
|
## Aidge support |
|
|
|
|
|
> Note: We tested this network for the following features. If you encounter any error please open an [issue](https://gitlab.eclipse.org/groups/eclipse/aidge/-/issues). Features not tested in CI may not be functional. |
|
|
|
|
|
| Feature | Tested in CI | |
|
|
| :---------: | :----------: | |
|
|
| ONNX import | ✔ | |
|
|
| Backend CPU | ❌ | |
|
|
| Export CPP | ❌ | |
|
|
|
|
|
## MNIST |
|
|
|
|
|
* **Input** |
|
|
* size: [N, 1, 28, 28] |
|
|
* format: [N, C, H, W] |
|
|
* preprocessing: `None` |
|
|
* **Output** |
|
|
* size: [N, 10] |
|
|
|
|
|
### ONNX attributes |
|
|
|
|
|
* Opset: <opset> |
|
|
* Source: PyTorch |
|
|
* Operators: 22 (6 types) |
|
|
- Conv2D: 2 |
|
|
- FC: 3 |
|
|
- Flatten: 1 |
|
|
- MaxPooling2D: 2 |
|
|
- Producer: 10 |
|
|
- ReLU: 4 |
|
|
|
|
|
### Benchmark |
|
|
|
|
|
> Coming soon |
