MyAdagrad¶
-
class
MyAdagrad(params, lr=0.01, lr_decay=0, init_accu_value=0.1, weight_decay=0)[source]¶ Modification of the Adagrad optimizer that allows to specify an initial accumulator value. This mimics the behavior of the default Adagrad implementation in Tensorflow. The default PyTorch Adagrad uses 0 for initial accumulator value.
- Parameters
params (iterable) – iterable of parameters to optimize or dicts defining parameter groups
lr (float, optional) – learning rate (default: 1e-2)
lr_decay (float, optional) – learning rate decay (default: 0)
init_accu_value (float, optional) – initial accumulater value.
weight_decay (float, optional) – weight decay (L2 penalty) (default: 0)