html refactor (#555)

colossalai/amp/__init__.py

@@ -19,7 +19,7 @@ def convert_to_amp(model: nn.Module, optimizer: Optimizer, criterion: _Loss, mod
         optimizer (:class:`torch.optim.Optimizer`): your optimizer object.
         criterion (:class:`torch.nn.modules.loss._Loss`): your loss function object.
         mode (:class:`colossalai.amp.AMP_TYPE`): amp mode.
-        amp_config (:class:`colossalai.context.Config` or dict): configuration for different amp modes
+        amp_config (Union[:class:`colossalai.context.Config`, dict]): configuration for different amp modes.
 
     Returns:
         A tuple (model, optimizer, criterion).

colossalai/amp/apex_amp/__init__.py

@@ -9,7 +9,7 @@ def convert_to_apex_amp(model: nn.Module, optimizer: Optimizer, amp_config):
     Args:
         model (:class:`torch.nn.Module`): your model object.
         optimizer (:class:`torch.optim.Optimizer`): your optimizer object.
-        amp_config (:class: colossalai.context.Config or dict): configuration for initializing apex_amp.
+        amp_config (Union[:class:`colossalai.context.Config`, dict]): configuration for initializing apex_amp.
 
     The ``amp_config`` should include parameters below:
     ::

colossalai/builder/builder.py

@@ -29,8 +29,8 @@ def build_from_registry(config, registry: Registry):
     is specified by `registry`.
 
     Note:
-        the `config` is used to construct the return object such as `LAYERS`,
-         `OPTIMIZERS` and other support types in `registry`. The `config` should contain
+        the `config` is used to construct the return object such as `LAYERS`, `OPTIMIZERS`
+        and other support types in `registry`. The `config` should contain
         all required parameters of corresponding object. The details of support
         types in `registry` and the `mod_type` in `config` could be found in
         `registry <https://github.com/hpcaitech/ColossalAI/blob/main/colossalai/registry/__init__.py>`_.
@@ -40,10 +40,11 @@ def build_from_registry(config, registry: Registry):
             used in the construction of the return object.
         registry (:class:`Registry`): A registry specifying the type of the return object
 
-    Returns: A Python object specified by `registry`
+    Returns:
+        A Python object specified by `registry`.
 
     Raises:
-        Exception: Raises an Exception if an error occurred when building from registry
+        Exception: Raises an Exception if an error occurred when building from registry.
     """
     config_ = config.copy()  # keep the original config untouched
     assert isinstance(

colossalai/builder/pipeline.py

@@ -163,12 +163,14 @@ def count_layer_params(layers):
 
 
 def build_pipeline_model_from_cfg(config, num_chunks: int = 1, partition_method: str = 'parameter', verbose: bool = False):
-    """An intializer to split the model into different stages for pipeline parallelism.
+    """An initializer to split the model into different stages for pipeline parallelism.
 
     An example for the model config is shown below. The class VisionTransformerFromConfig should
     inherit colossalai.nn.model.ModelFromConfig to allow this initializer to build model from a sequence
     of layer configurations.
 
+    ::
+
         model_config = dict(
             type='VisionTransformerFromConfig',
             embedding_cfg=dict(...),

colossalai/communication/utils.py

@@ -45,7 +45,7 @@ def recv_tensor_meta(tensor_shape, prev_rank=None):
         prev_rank (int): The rank of the source of the tensor.
 
     Returns:
-        torch.Size: The shape of the tensor to be received.
+        :class:`torch.Size`: The shape of the tensor to be received.
     """
     if tensor_shape is None:
         if prev_rank is None:
@@ -71,7 +71,7 @@ def split_tensor_into_1d_equal_chunks(tensor, new_buffer=False):
         new_buffer (bool, optional): Whether to use a new buffer to store sliced tensor.
 
     Returns:
-        torch.Tensor: The split tensor
+        :class:`torch.Size`: The split tensor
     """
     partition_size = torch.numel(tensor) // gpc.get_world_size(ParallelMode.PARALLEL_1D)
     start_index = partition_size * gpc.get_local_rank(ParallelMode.PARALLEL_1D)
@@ -92,7 +92,7 @@ def gather_split_1d_tensor(tensor):
     Args:
         tensor (torch.Tensor): Tensor to be gathered after communication.
     Returns:
-        gathered (torch.Tensor): The gathered tensor
+        :class:`torch.Size`: The gathered tensor.
     """
     world_size = gpc.get_world_size(ParallelMode.PARALLEL_1D)
     numel = torch.numel(tensor)

colossalai/context/random/seed_manager.py

@@ -34,6 +34,7 @@ class SeedManager:
 
     def set_state(self, parallel_mode: ParallelMode, state: Tensor):
         """Sets the state of the seed manager for `parallel_mode`.
+
         Args:
             parallel_mode (:class:`colossalai.context.ParallelMode`): The chosen parallel mode.
             state (:class:`torch.Tensor`): the state to be set.
@@ -66,9 +67,9 @@ class SeedManager:
             seed (int): The seed to be added.
             overwrtie (bool, optional): Whether allows to overwrite the seed that has been set already
 
-        Raises
-            AssertionError: Raises an AssertionError if `parallel_mode` is not an instance of
-                :class:`colossalai.context.ParallelMode` or the seed for `parallel_mode` has been added.
+        Raises:
+            AssertionError: Raises an AssertionError if `parallel_mode` is not an instance of :class:`colossalai.context.ParallelMode`
+                or the seed for `parallel_mode` has been added.
         """
         assert isinstance(parallel_mode, ParallelMode), 'A valid ParallelMode must be provided'
         if overwrtie is False:

colossalai/nn/loss/loss_2d.py

@@ -27,7 +27,7 @@ class CrossEntropyLoss2D(_Loss):
         reduce (bool, optional)
         label_smoothing (float, optional)
 
-    More details about args, kwargs and torch.nn.functional.cross_entropy could be found in
+    More details about ``args``, ``kwargs`` and ``torch.nn.functional.cross_entropy`` could be found in
     `Cross_entropy <https://pytorch.org/docs/stable/generated/torch.nn.functional.cross_entropy.html#torch.nn.functional.cross_entropy>`_.
     """
 

colossalai/nn/loss/loss_2p5d.py

@@ -27,7 +27,7 @@ class CrossEntropyLoss2p5D(_Loss):
         reduce (bool, optional)
         label_smoothing (float, optional)
 
-    More details about args, kwargs and torch.nn.functional.cross_entropy could be found in
+    More details about ``args``, ``kwargs`` and ``torch.nn.functional.cross_entropy`` could be found in
     `Cross_entropy <https://pytorch.org/docs/stable/generated/torch.nn.functional.cross_entropy.html#torch.nn.functional.cross_entropy>`_.
     """
     def __init__(self, reduction=True, *args, **kwargs):

colossalai/nn/loss/loss_3d.py

@@ -27,7 +27,7 @@ class CrossEntropyLoss3D(_Loss):
         reduce (bool, optional)
         label_smoothing (float, optional)
 
-    More details about args, kwargs and torch.nn.functional.cross_entropy could be found in
+    More details about ``args``, ``kwargs`` and ``torch.nn.functional.cross_entropy`` could be found in
     `Cross_entropy <https://pytorch.org/docs/stable/generated/torch.nn.functional.cross_entropy.html#torch.nn.functional.cross_entropy>`_.
     """
 

colossalai/nn/loss/loss_moe.py

@@ -23,7 +23,7 @@ class MoeCrossEntropyLoss(_Loss):
         reduction (str, optional)
         label_smoothing (float, optional)
 
-    More details about args, kwargs and torch.nn.functional.cross_entropy could be found in
+    More details about ``args``, ``kwargs`` and ``torch.nn.functional.cross_entropy`` could be found in
     `Cross_entropy <https://pytorch.org/docs/stable/generated/torch.nn.functional.cross_entropy.html#torch.nn.functional.cross_entropy>`_.
     """
 
@@ -40,7 +40,7 @@ class MoeCrossEntropyLoss(_Loss):
             input (:class:`torch.tensor`): Predicted unnormalized scores (often referred to as logits).
             target (:class:`torch.tensor`): Ground truth class indices or class probabilities.
 
-        More details about args, kwargs and torch.nn.functional.cross_entropy could be found in
+        More details about ``args``, ``kwargs`` and ``torch.nn.functional.cross_entropy`` could be found in
         `Cross_entropy <https://pytorch.org/docs/stable/generated/torch.nn.functional.cross_entropy.html#torch.nn.functional.cross_entropy>`_.
         """
         main_loss = self.loss(*args)

colossalai/trainer/_trainer.py

@@ -307,8 +307,7 @@ class Trainer:
             max_steps (int, optional): Maximum number of running iterations.
             test_dataloader (:class:`torch.utils.data.DataLoader`, optional): DataLoader for validation.
             test_interval (int, optional): Interval of validation
-            hooks (list[`BaseHook <https://github.com/hpcaitech/ColossalAI/tree/main/colossalai/trainer/hooks>`_],
-                optional): A list of hooks used in training.
+            hooks (list[BaseHook], optional): A list of hooks used in training.
             display_progress (bool, optional): If True, a progress bar will be displayed.
         """
 

colossalai/utils/memory_tracer/async_memtracer.py

@@ -21,6 +21,7 @@ class AsyncMemoryMonitor:
     :type power: int
 
     Usage:
+    ::
 
         ```python
             async_mem_monitor = AsyncMemoryMonitor()

colossalai/utils/profiler/prof_utils.py

@@ -73,6 +73,7 @@ class ProfilerContext(object):
     """
     Profiler context manager
     Usage:
+    ::
 
         ```python
             world_size = 4

docs/colossalai/colossalai.amp.apex_amp.apex_amp.rst

-colossalai.amp.apex\_amp.apex\_amp
-==================================
-
-.. automodule:: colossalai.amp.apex_amp.apex_amp
-   :members:

docs/colossalai/colossalai.amp.apex_amp.rst

@@ -3,9 +3,3 @@ colossalai.amp.apex\_amp
 
 .. automodule:: colossalai.amp.apex_amp
    :members:
-
-
-.. toctree::
-   :maxdepth: 2
-
-   colossalai.amp.apex_amp.apex_amp

docs/colossalai/colossalai.amp.naive_amp.grad_scaler.base_grad_scaler.rst

-colossalai.amp.naive\_amp.grad\_scaler.base\_grad\_scaler
-=========================================================
-
-.. automodule:: colossalai.amp.naive_amp.grad_scaler.base_grad_scaler
-   :members:

docs/colossalai/colossalai.amp.naive_amp.grad_scaler.constant_grad_scaler.rst

-colossalai.amp.naive\_amp.grad\_scaler.constant\_grad\_scaler
-=============================================================
-
-.. automodule:: colossalai.amp.naive_amp.grad_scaler.constant_grad_scaler
-   :members:

docs/colossalai/colossalai.amp.naive_amp.grad_scaler.dynamic_grad_scaler.rst

-colossalai.amp.naive\_amp.grad\_scaler.dynamic\_grad\_scaler
-============================================================
-
-.. automodule:: colossalai.amp.naive_amp.grad_scaler.dynamic_grad_scaler
-   :members: