TensorBoard logging of hyper and experiment params

[roybenhayun]

now that we are running multiple models with various configurations and hyperparams, we have to track the different parameters. doing this with TensorBoard HPARAMS tab seems the the right way.

when trying to log additional parameters, getting either the additional parameter, or the params logged usually by torchgeo e.g., model, backbone, weights etc:

tb_logger = TensorBoardLogger(save_dir=f"out/{experiment}/tb_logs/", name=f"{experiment}_{run_name}")
...
tb_logger.log_hyperparams({"num_channels:": config.num_channels, "input_image": ...})  <-- either these appear
...
trainer = pl.Trainer(logger=[tb_logger],  ... )  <-- or the torchgeo hparam

probably missing something. any suggestion how to make both set of parameters to appear in TensorBoard HPARAMS?