Hey all, very new to clearML! I am trying to design an HPO setup using the optuna configuration, and I'm working on getting my template trainer set up. The issue I'm having is it's unclear to me how to define one of my hyperparameters whose size is dynamic. As a simple example

hidden_layers_1 = []
for i in range(num_layers1):
a = trial.suggest_int(f"a1_{i}", 0, 4)
m = trial.suggest_int(f"m1_{i}", 1, 4)
hidden_layers_1.append([a, m])

in this example num_layers is a hyperparameter, but a1_{i} and m1_{i} are also hyperparameters. For optimization, num_layers can be either 1, 2 or 3 - but this means that there can be either 1, 2, or 3 values for a1_{i} and m1_{i} , which themselves can be 0 to 4 or 1 to 4, respectively. For my template trainer, though, I obviously won't be using the suggest_int syntax and providing a range. I need to figure out a way to define this so that I can pass arguments that can be passed to clearML as hyperparameters, but also how to define that hyperparameter within the hyper_parameters list inside the HyperParameterOptimizer arguments. In my mind I'm imagining a brute force solution but I'm really hoping there's a cleaner, more clever way to go about it.

If anyone has any suggestions - or if I did a terrible job explaining this and you could use some more information - please let know. Thanks!