Welcome to EggNNAS Documentation
Getting Started
Installation
pip install eggnnas
Install via PyPI for access to core libraries and models.
Basic Usage
from eggnnas import AutoModel
model = AutoModel.from_pretrained("eggnnas-base")
Quickly create pre-trained model instances for inference or training.
Training
from eggnnas import Trainer
trainer = Trainer()
trainer.train_model(dataset="cifar10", epochs=100)
Fine-tune models with custom datasets and optimizer configurations.
API Reference
Model Architectures
- EggNet-18: Lightweight CNN for image classification (80.2% ImageNet)
- EggResNet-101: Deep residual network for complex tasks
- EggTransformer: Vision Transformer with multi-head attention
Training Parameters
batch_size: 128 (default), supports dynamic adjustmentlearning_rate: 1e-3 with warmup scheduleprecision: 16-bit AMP supported
Available Datasets
- cifar-10
- imagenet-1k
- mnist
- fashion-mnist
- custom
Frequently Asked Questions
How do I switch model versions?
Use the set_version() method on your model instance: model.set_version("v0.4.2")
Can I run models without a GPU?
Yes, the framework supports CPU execution but performance will be significantly slower - at least 8x slower.
How do I track training progress?
Training metrics are accessible via the metrics() method, and can be visualized using TensorBoard with our built-in exporter.
Is there a local cache for models?
All downloaded models are cached in ~/.cache/eggnnas for reuse between runs.
Need Help?
Can't find what you're looking for? Our community is active on Discord and we provide enterprise support for professional users.