docs(mge/optimizer): refine the docstring of several apis

GitOrigin-RevId: a97fe5b68ace2983a16b747f5b4d73562feb8e9e

python_module/megengine/optimizer/adam.py

@@ -14,7 +14,7 @@ from .optimizer import Optimizer
 
 
 class Adam(Optimizer):
-    r"""Implements Adam algorithm.
+    r"""Implements Adam algorithm proposed in `"Adam: A Method for Stochastic Optimization" <https://arxiv.org/abs/1412.6980>`_.
 
     :param params: iterable of parameters to optimize or dicts defining
             parameter groups.

python_module/megengine/optimizer/internal.py

@@ -21,7 +21,7 @@ def add_update_fastpath(
     beta: Union[Tensor, float, int] = 1.0,
     bias: Union[Tensor, float, int] = 0.0
 ):
-    """a fast-path ONLY used to update parameters in optimzier, since it
+    """a fast-path ONLY used to update parameters in optimizer, since it
     would bypass computing graph and launch dnn/add_update kernel directly,
     it is more efficient than functional/add_update.
     """

python_module/megengine/optimizer/lr_scheduler.py

@@ -4,7 +4,7 @@ from .optimizer import Optimizer
 
 
 class LRScheduler(metaclass=ABCMeta):
-    r"""Base class for all lr_schedulers.
+    r"""Base class for all learning rate based schedulers.
 
     :param optimizer: Wrapped optimizer.
     :param current_epoch: The index of current epoch. Default: -1

python_module/megengine/optimizer/sgd.py

@@ -17,7 +17,7 @@ class SGD(Optimizer):
     r"""Implements stochastic gradient descent.
 
     Nesterov momentum is based on the formula from
-    `On the importance of initialization and momentum in deep learning`.
+    `"On the importance of initialization and momentum in deep learning" <http://www.cs.toronto.edu/%7Ehinton/absps/momentum.pdf>`_ .
 
     :param params: iterable of parameters to optimize or dicts defining
             parameter groups.