TorchJD: Training with multiple losses in PyTorch

TorchJD is a PyTorch library for handling multiple losses in multi-task learning and constrained optimization. It provides automatic gradient balancing and is available as an open-source project on GitHub.