Different learning rate per encoding level

[aboelela924]

I have 16 levels of Hash Encoding is there a way of setting a different learning rate per level. I have alreadly estimated the shift in each level using the following code:
`
base_resolution = self.model.hash_encoder.encoding_config['base_resolution']
log2_hashmap_size = self.model.hash_encoder.encoding_config['log2_hashmap_size']
per_level_scale = self.model.hash_encoder.encoding_config['per_level_scale']
n_levels = self.model.hash_encoder.encoding_config['n_levels']
n_features_per_level = self.model.hash_encoder.encoding_config['n_features_per_level']
N_POS_DIMS = 3
offset = 0

    levels = torch.zeros_like(self.model.hash_encoder.params)
    self.offset_table = torch.zeros((n_levels, 2), dtype=torch.int32)
    for i in range(n_levels):
        resolution = math.ceil((math.exp(math.log(per_level_scale) * i) * base_resolution) - 1) + 1
        params_in_level = resolution ** N_POS_DIMS
        params_in_level = min(params_in_level, 2 ** log2_hashmap_size)
        params_in_level = math.ceil(params_in_level / 8) * 8
        levels[offset * n_features_per_level:(offset + params_in_level) * n_features_per_level] = i
        self.offset_table[i, 0] = offset * n_features_per_level
        self.offset_table[i, 1] = (offset + params_in_level) * n_features_per_level
        offset += params_in_level
    self.param_level = levels

`

but i can't find a way in pytroch to set a different learning rate per level