diff --git a/imperative/python/megengine/core/tensor/megbrain_graph.py b/imperative/python/megengine/core/tensor/megbrain_graph.py
index 1bff175b254734107273b33380412a172b4b56f0..e11e0f29ca8dd76978a0ab6327e0f1e34f0598f8 100644
--- a/imperative/python/megengine/core/tensor/megbrain_graph.py
+++ b/imperative/python/megengine/core/tensor/megbrain_graph.py
@@ -27,7 +27,8 @@ from .core import TensorBase
 def set_priority_to_id(dest_vars):
     """
     For all oprs in the subgraph constructed by dest_vars,
-       sets its priority to id if its original priority is zero.
+    sets its priority to id if its original priority is zero.
+    
     :param dest_vars: target vars representing the graph.
     """
     dest_vec = []
diff --git a/imperative/python/megengine/data/collator.py b/imperative/python/megengine/data/collator.py
index 270a93af6ac8cbb17d137ef4694f44a5500b73ca..113507f364faf8b2538d985fbf41b24c2d7947b4 100644
--- a/imperative/python/megengine/data/collator.py
+++ b/imperative/python/megengine/data/collator.py
@@ -40,7 +40,7 @@ class Collator:
 
     def apply(self, inputs):
         """
-        :param input: sequence_N(tuple(CHW, C, CK)).
+        :param inputs: sequence_N(tuple(CHW, C, CK)).
         :return: tuple(NCHW, NC, NCK).
         """
         elem = inputs[0]
diff --git a/imperative/python/megengine/data/dataloader.py b/imperative/python/megengine/data/dataloader.py
index 2eaad6e0db52684fe6ee6770f8a242cf3c38cea6..88b37c5901839599c09b66d615520ed05646bb76 100644
--- a/imperative/python/megengine/data/dataloader.py
+++ b/imperative/python/megengine/data/dataloader.py
@@ -43,8 +43,29 @@ def raise_timeout_error():
 
 
 class DataLoader:
-    r"""
-    Provides a convenient way to iterate on a given dataset.
+    r"""Provides a convenient way to iterate on a given dataset.
+
+    DataLoader combines a dataset with 
+    :class:`~.Sampler`, :class:`~.Transform` and :class:`~.Collator`,
+    make it flexible to get minibatch continually from a dataset.
+
+    :param dataset: dataset from which to load the minibatch.
+    :param sampler: defines the strategy to sample data from the dataset.
+    :param transform: defined the transforming strategy for a sampled batch.
+        Default: None
+    :param collator: defined the merging strategy for a transformed batch.
+        Default: None
+    :param num_workers: the number of sub-process to load, transform and collate
+        the batch. ``0`` means using single-process. Default: 0
+    :param timeout: if positive, means the timeout value(second) for collecting a
+        batch from workers. Default: 0
+    :param timeout_event: callback function triggered by timeout, default to raise
+        runtime error.
+    :param divide: define the paralleling strategy in multi-processing mode.
+        ``True`` means one batch is divided into :attr:`num_workers` pieces, and
+        the workers will process these pieces parallelly. ``False`` means
+        different sub-process will process different batch. Default: False
+
     """
     __initialized = False
 
@@ -59,36 +80,6 @@ class DataLoader:
         timeout_event: Callable = raise_timeout_error,
         divide: bool = False,
     ):
-        r"""
-        `DataLoader` combines a dataset with `sampler`, `transform` and `collator`,
-        make it flexible to get minibatch continually from a dataset.
-
-        :type dataset: Dataset
-        :param dataset: dataset from which to load the minibatch.
-        :type sampler: Sampler
-        :param sampler: defines the strategy to sample data from the dataset.
-        :type transform: Transform
-        :param transform: defined the transforming strategy for a sampled batch.
-            Default: None
-        :type collator: Collator
-        :param collator: defined the merging strategy for a transformed batch.
-            Default: None
-        :type num_workers: int
-        :param num_workers: the number of sub-process to load, transform and collate
-            the batch. ``0`` means using single-process. Default: 0
-        :type timeout: int
-        :param timeout: if positive, means the timeout value(second) for collecting a
-            batch from workers. Default: 0
-        :type timeout_event: Callable
-        :param timeout_event: callback function triggered by timeout, default to raise
-            runtime error.
-        :type divide: bool
-        :param divide: define the paralleling strategy in multi-processing mode.
-            ``True`` means one batch is divided into :attr:`num_workers` pieces, and
-            the workers will process these pieces parallelly. ``False`` means
-            different sub-process will process different batch. Default: False
-
-        """
         if num_workers < 0:
             raise ValueError("num_workers should not be negative")
 
diff --git a/imperative/python/megengine/data/sampler.py b/imperative/python/megengine/data/sampler.py
index c31efb98d76bd6a65d90dd662346de9bc97d1e4c..7cec2873c78aa14fd778757ca133753b83158c40 100644
--- a/imperative/python/megengine/data/sampler.py
+++ b/imperative/python/megengine/data/sampler.py
@@ -30,22 +30,15 @@ class MapSampler(Sampler):
     r"""
     Sampler for map dataset.
 
-    :type dataset: `dataset`
     :param dataset: dataset to sample from.
-    :type batch_size: positive integer
     :param batch_size: batch size for batch method.
-    :type drop_last: bool
     :param drop_last: set ``True`` to drop the last incomplete batch,
         if the dataset size is not divisible by the batch size. If ``False`` and 
         the size of dataset is not divisible by the batch_size, then the last batch will
         be smaller. Default: False
-    :type num_samples: positive integer
     :param num_samples: number of samples assigned to one rank.
-    :type world_size: positive integer
     :param world_size: number of ranks.
-    :type rank: non-negative integer within 0 and world_size
     :param rank: rank id, non-negative interger within 0 and ``world_size``.
-    :type seed: non-negative integer
     :param seed: seed for random operators.
     """
 
@@ -166,7 +159,7 @@ class StreamSampler(Sampler):
         different data. But this class cannot do it yet, please build your own
         dataset and sampler to achieve this goal.
 
-    Usually, meth::`~.StreamDataset.__iter__` can return different iterator by
+    Usually, :meth:`~.StreamDataset.__iter__` can return different iterator by
     ``rank = dist.get_rank()``. So that they will get different data.
 
     """
@@ -184,6 +177,16 @@ class StreamSampler(Sampler):
 class SequentialSampler(MapSampler):
     r"""
     Sample elements sequentially.
+
+    :param dataset: dataset to sample from.
+    :param batch_size: batch size for batch method.
+    :param drop_last: set ``True`` to drop the last incomplete batch,
+        if the dataset size is not divisible by the batch size. If ``False`` and 
+        the size of dataset is not divisible by the batch_size, then the last batch will
+        be smaller. Default: False
+    :param indices: indice of samples.
+    :param world_size: number of ranks.
+    :param rank: rank id, non-negative interger within 0 and ``world_size``.
     """
 
     def __init__(
@@ -216,6 +219,17 @@ class SequentialSampler(MapSampler):
 class RandomSampler(MapSampler):
     r"""
     Sample elements randomly without replacement.
+
+    :param dataset: dataset to sample from.
+    :param batch_size: batch size for batch method.
+    :param drop_last: set ``True`` to drop the last incomplete batch,
+        if the dataset size is not divisible by the batch size. If ``False`` and 
+        the size of dataset is not divisible by the batch_size, then the last batch will
+        be smaller. Default: False
+    :param indices: indice of samples.
+    :param world_size: number of ranks.
+    :param rank: rank id, non-negative interger within 0 and ``world_size``.
+    :param seed: seed for random operators.
     """
 
     def __init__(
@@ -247,8 +261,17 @@ class ReplacementSampler(MapSampler):
     r"""
     Sample elements randomly with replacement.
 
-    :type weights: List
+    :param dataset: dataset to sample from.
+    :param batch_size: batch size for batch method.
+    :param drop_last: set ``True`` to drop the last incomplete batch,
+        if the dataset size is not divisible by the batch size. If ``False`` and 
+        the size of dataset is not divisible by the batch_size, then the last batch will
+        be smaller. Default: False
+    :param num_samples: number of samples assigned to one rank.
     :param weights: weights for sampling indices, it could be unnormalized weights.
+    :param world_size: number of ranks.
+    :param rank: rank id, non-negative interger within 0 and ``world_size``.
+    :param seed: seed for random operators.
     """
 
     def __init__(
diff --git a/imperative/python/megengine/functional/vision.py b/imperative/python/megengine/functional/vision.py
index 8ed39a732c082467178d8855a9cfbc72758e6c2a..cb9b339311a489bd89c2d2e0f6437bb076955b56 100644
--- a/imperative/python/megengine/functional/vision.py
+++ b/imperative/python/megengine/functional/vision.py
@@ -224,7 +224,7 @@ def nms(
     :param scores: tensor of shape `(N,)`, the score of boxes.
     :param max_output: the maximum number of boxes to keep; it is optional if this operator is not traced
         otherwise it required to be specified; if it is not specified, all boxes are kept.
-    :return: indices of the elements that have been kept by NMS.
+    :return: indices of the elements that have been kept by NMS, sorted by scores.
 
     Examples:
 
diff --git a/imperative/python/megengine/module/conv.py b/imperative/python/megengine/module/conv.py
index cd156eeceeff3372ad83385b582852d9982ad31d..b7c6007d2eaddb75a2a7211fd1c2eb055080c43a 100644
--- a/imperative/python/megengine/module/conv.py
+++ b/imperative/python/megengine/module/conv.py
@@ -409,7 +409,7 @@ class Conv3d(_ConvNd):
 
     For instance, given an input of the size :math:`(N, C_{\text{in}}, T, H, W)`,
     this layer generates an output of the size
-    :math:`(N, C_{\text{out}}, T_{\text{out}}}, H_{\text{out}}}, W_{\text{out}}})` through the
+    :math:`(N, C_{\text{out}}, T_{\text{out}}, H_{\text{out}}, W_{\text{out}})` through the
     process described as below:
 
     .. math::
diff --git a/imperative/python/megengine/module/module.py b/imperative/python/megengine/module/module.py
index 1569e2e359b5c26f0517bad46600f639f3efcddc..6560302fa403f072f3671ed3f1fe5c35eeee5e6c 100644
--- a/imperative/python/megengine/module/module.py
+++ b/imperative/python/megengine/module/module.py
@@ -91,7 +91,7 @@ class Module(metaclass=ABCMeta):
     def __init__(self, name=None):
         """
         :param name: module's name, can be initialized by the ``kwargs`` parameter
-        of child class.
+            of child class.
         """
         self._modules = []
 
@@ -122,7 +122,7 @@ class Module(metaclass=ABCMeta):
         Registers a hook to handle forward inputs. `hook` should be a function.
 
         :param hook: a function that receive `module` and `inputs`, then return
-        a modified `inputs` or `None`.
+            a modified `inputs` or `None`.
         :return: a handler with :meth:`~.HookHandler.remove` interface to delete the hook.
         """
         return HookHandler(self._forward_pre_hooks, hook)
diff --git a/imperative/python/megengine/tensor.py b/imperative/python/megengine/tensor.py
index c343e2f094eee947f0413788733bfb6916d36391..574e785110674eac661eb7fd5abd4a67b06f3cdf 100644
--- a/imperative/python/megengine/tensor.py
+++ b/imperative/python/megengine/tensor.py
@@ -174,7 +174,7 @@ class Tensor(_Tensor, ArrayMethodMixin):
     def set_value(self, value):
         self._reset(value)
 
-    @deprecated(version="1.0", reason="use *= 0 instead")
+    @deprecated(version="1.0", reason="use ``*= 0`` instead")
     def reset_zero(self):
         self *= 0