Local PyTorch training runner with a small FastAPI web UI and import-path based extension points for datasets, models, and losses.
Run a deterministic synthetic dense-network example:
uv run python -m examples.train_synthetic_dense --epochs 1 --device cpuFine-tune a pretrained ResNet-18 on CIFAR-10:
uv run python -m examples.finetune_cifar10_resnet --epochs 1 --device cpu --train-limit 512 --val-limit 128