I've been setting up tensorboard with PyTorch. Originally I was using lightning, and ran into an issue where I would log my hyperparameter, but the metrics in the HPARAMS tab would be empty. So, to pin point the issue I moved over to the simplest possible example and the issue is still happening from torch.utils.tensorboard import SummaryWriter writer = SummaryWriter(log_dir="tb_logs/test") writer.add_hparams({"a": 0, "b": 0}, {"hparam/test_accuracy": 0.5}) writer.add_hparams({"a": 0, "b": 1}, {"hparam/test_accuracy": 0.6}) writer.add_hparams({"a": 1, "b": 0}, {"hparam/test_accuracy": 0.61}) writer.add_hparams({"a": 1, "b": 1}, {"hparam/test_accuracy": 0.4}) writer.add_hparams({"a": 2.0, "b": 1.5}, {"hparam/test_accuracy": 0.7}) Per the PyTorch docs I'd expect to this exact table in the HPARAMS tab, but when I go I am met with this instead (https://imgur.com/a/zb1yTPA) I am using lightning 2.4.0 torch 2.4.0 tensorboard 2.17.0 Continue reading...
