CustomPresetDataDictionary
Table of contents
There are 100 preset slots available by default. These presets automatically store/load the UI values.
That said, you may want to store some tensors along side these UI values.
As a result, you can use the CustomPresetDataDictionary struct to store any tensor data you want along side the UI values.
Tracking Tensors for Saving Presets
To track the content of a given tensor that should be stored in a preset, do the following in the deploy() method
auto some_tensor = torch::rand({1, 1, 1, 1});
CustomPresetData->tensor("input_tensor", some_tensor);
auto second_tensor = torch::rand({128});
CustomPresetData->tensor("second_tensor", second_tensor);
Once a new preset is saved, the content of these tensors will be saved along side the UI values.
Accessing Tensors from a Loaded Preset
once a new preset is loaded, you will be notified via the new_preset_loaded_since_last_call flag.
You can then access the content of the stored tensors as follows:
Approach 1: Iterate over all tensors
if (new_preset_loaded_since_last_call) {
for (const auto& pair : CustomPresetData->tensors()) {
const std::string& key = pair.first;
const torch::Tensor& t = pair.second;
}
}
Approach 2: Access a specific tensor using the specified key
if (new_preset_loaded_since_last_call) {
auto t = CustomPresetData->tensor(key);
if (t != std::nullopt) { <-- if key doesnt exist, it will return std::nullopt, so ALWAYS check for this
auto value = *t;
}
}