|
--- |
|
datasets: |
|
- imagenet-1k |
|
pipeline_tag: image-classification |
|
--- |
|
|
|
## Model Architecture Details |
|
|
|
### Architecture Overview |
|
|
|
- **Architecture**: ViT Base |
|
|
|
### Configuration |
|
|
|
| Attribute | Value | |
|
|----------------------|----------------| |
|
| Patch Size | 16 | |
|
| Image Size | 224 | |
|
| Num Layers | 2 | |
|
| Attention Heads | 4 | |
|
| Objective Function | CrossEntropy | |
|
|
|
### Performance |
|
|
|
- **Validation Accuracy (Top 5)**: 0.34 |
|
- **Validation Accuracy**: 0.16 |
|
|
|
### Additional Resources |
|
|
|
The model was trained using the library: [ViT-Prisma](https://github.com/soniajoseph/ViT-Prisma).\ |
|
For detailed metrics, plots, and further analysis of the model's training process, refer to the [training report](https://wandb.ai/perceptual-alignment/Imagenet/reports/ViT-Small-Imagenet-training-report--Vmlldzo3MDk3MTM5). |
|
|
