update api urls (#509)

doombeaker · web-flow · commit 0d0fbb1356a0 · 2022-07-29T10:16:23.000+08:00
diff --git a/cn/docs/cookies/activation_checkpointing.md b/cn/docs/cookies/activation_checkpointing.md
@@ -40,7 +40,7 @@ optimizer = flow.optim.SGD([{'params': model_part1.parameters()},
                            lr=1e-3)
 ```
 
-如果要开启 activation checkpointing，只需在 [nn.Graph](../basics/08_nn_graph.md) 模型中的 Eager 模型成员 (即 nn.Module 对象) 上指定 `.config.activation_checkpointing = True`。此 API 详见：[activation_checkpointing](https://oneflow.readthedocs.io/en/master/graph.html#oneflow.nn.graph.block_config.BlockConfig.activation_checkpointing)。对于每个打开 "activation checkpointing" 的 nn.Module，其输入 activation 将会被保留，而其它中间 activation 在反向传播过程中被使用时会被重新计算。
+如果要开启 activation checkpointing，只需在 [nn.Graph](../basics/08_nn_graph.md) 模型中的 Eager 模型成员 (即 nn.Module 对象) 上指定 `.config.activation_checkpointing = True`。此 API 详见：[activation_checkpointing](https://oneflow.readthedocs.io/en/v0.8.1/generated/oneflow.nn.graph.block_config.BlockConfig.activation_checkpointing.html)。对于每个打开 "activation checkpointing" 的 nn.Module，其输入 activation 将会被保留，而其它中间 activation 在反向传播过程中被使用时会被重新计算。
 
 ```python
 class CustomGraph(flow.nn.Graph):
diff --git a/cn/docs/cookies/amp.md b/cn/docs/cookies/amp.md
@@ -27,7 +27,7 @@ loss_fn = nn.CrossEntropyLoss().to(DEVICE)
 optimizer = flow.optim.SGD(model.parameters(), lr=1e-3)
 ```
 
-如果要开启 AMP 模式，只需在 [nn.Graph](../basics/08_nn_graph.md) 模型中添加 `self.config.enable_amp(True)`，此 API 详见： [enable_amp](https://oneflow.readthedocs.io/en/master/graph.html#oneflow.nn.graph.graph_config.GraphConfig.enable_amp)。
+如果要开启 AMP 模式，只需在 [nn.Graph](../basics/08_nn_graph.md) 模型中添加 `self.config.enable_amp(True)`，此 API 详见： [enable_amp](https://oneflow.readthedocs.io/en/v0.8.1/generated/oneflow.nn.graph.graph_config.GraphConfig.enable_amp.html)。
 
 ```python
 class CustomGraph(flow.nn.Graph):
@@ -61,7 +61,7 @@ for _ in range(100):
 
 **Gradient Scaling (梯度缩放)** 是一种用于解决 FP16 易导致数值溢出问题的方法，其基本原理是在反向传播的过程中使用一个 scale factor 对损失和梯度进行缩放，以改变其数值的量级，从而尽可能缓解数值溢出问题。
 
-OneFlow 提供了 `GradScaler` 来在 AMP 模式下使用 Gradient Scaling，只需要在 nn.Graph 模型的 `__init__` 方法中实例化一个`GradScaler` 对象，然后通过 [set_grad_scaler](https://oneflow.readthedocs.io/en/master/graph.html#oneflow.nn.Graph.set_grad_scaler) 接口进行指定即可，nn.Graph 将会自动管理 Gradient Scaling 的整个过程。以上文中的 `CustomGraph` 为例，我们需要在其 `__init__` 方法中添加：
+OneFlow 提供了 `GradScaler` 来在 AMP 模式下使用 Gradient Scaling，只需要在 nn.Graph 模型的 `__init__` 方法中实例化一个`GradScaler` 对象，然后通过 [set_grad_scaler](https://oneflow.readthedocs.io/en/v0.8.1/generated/oneflow.nn.Graph.set_grad_scaler.html) 接口进行指定即可，nn.Graph 将会自动管理 Gradient Scaling 的整个过程。以上文中的 `CustomGraph` 为例，我们需要在其 `__init__` 方法中添加：
 
 ```python
 grad_scaler = flow.amp.GradScaler(
diff --git a/en/docs/basics/02_tensor.md b/en/docs/basics/02_tensor.md
@@ -164,7 +164,7 @@ oneflow.int32 cuda:0
 
 ## Operations on Tensors
 
-A large number of operators are provided in OneFlow, most of which are in the namespaces of [oneflow](https://oneflow.readthedocs.io/en/v0.8.1/oneflow.html), [oneflow.Tensor](https://oneflow.readthedocs.io/en/v0.8.1/tensor.html), [oneflow.nn](https://oneflow.readthedocs.io/en/master/nn.html), and [oneflow.nn.functional](https://oneflow.readthedocs.io/en/v0.8.1/nn.functional.html).
+A large number of operators are provided in OneFlow, most of which are in the namespaces of [oneflow](https://oneflow.readthedocs.io/en/v0.8.1/oneflow.html), [oneflow.Tensor](https://oneflow.readthedocs.io/en/v0.8.1/tensor.html), [oneflow.nn](https://oneflow.readthedocs.io/en/v0.8.1/nn.html), and [oneflow.nn.functional](https://oneflow.readthedocs.io/en/v0.8.1/nn.functional.html).
 
 Tensors in OneFlow are as easy to use as the NumPy arrays. For example, slicing in NumPy style is supported:
 
diff --git a/en/docs/cookies/activation_checkpointing.md b/en/docs/cookies/activation_checkpointing.md
@@ -40,7 +40,7 @@ optimizer = flow.optim.SGD([{'params': model_part1.parameters()},
                            lr=1e-3)
 ```
 
-To turn on activation checkpointing, you only need to specify `.config.activation_checkpointing = True` on the Eager model member (i.e. the nn.Module object) in the [nn.Graph](../basics/08_nn_graph.md) model. For more details of this API, please refer to: [activation_checkpointing](https://oneflow.readthedocs.io/en/master/graph.html#oneflow.nn.graph.block_config.BlockConfig.activation_checkpointing). For each nn.Module with "activation checkpointing" turned on, its input activations will be preserved, while other intermediate activations will be recomputed when used during backpropagation.
+To turn on activation checkpointing, you only need to specify `.config.activation_checkpointing = True` on the Eager model member (i.e. the nn.Module object) in the [nn.Graph](../basics/08_nn_graph.md) model. For more details of this API, please refer to: [activation_checkpointing](https://oneflow.readthedocs.io/en/v0.8.1/generated/oneflow.nn.graph.block_config.BlockConfig.activation_checkpointing.html). For each nn.Module with "activation checkpointing" turned on, its input activations will be preserved, while other intermediate activations will be recomputed when used during backpropagation.
 
 ```python
 class CustomGraph(flow.nn.Graph):
diff --git a/en/docs/cookies/amp.md b/en/docs/cookies/amp.md
@@ -27,7 +27,7 @@ loss_fn = nn.CrossEntropyLoss().to(DEVICE)
 optimizer = flow.optim.SGD(model.parameters(), lr=1e-3)
 ```
 
-If you want to enable AMP mode, just add `self.config.enable_amp(True)` to the model [nn.Graph](../basics/08_nn_graph.md). The details of this API is at: [enable_amp](https://oneflow.readthedocs.io/en/master/graph.html#oneflow.nn.graph.graph_config.GraphConfig.enable_amp).
+If you want to enable AMP mode, just add `self.config.enable_amp(True)` to the model [nn.Graph](../basics/08_nn_graph.md). The details of this API is at: [enable_amp](https://oneflow.readthedocs.io/en/v0.8.1/generated/oneflow.nn.graph.graph_config.GraphConfig.enable_amp.html).
 
 ```python
 class CustomGraph(flow.nn.Graph):
@@ -61,7 +61,7 @@ for _ in range(100):
 
 **Gradient Scaling** is a method for solving the problem that FP16 is prone to numerical overflow. The basic principle is to use a scale factor to scale the loss and gradient in the process of backpropagation to change the magnitude of its value, thereby mitigate numerical overflow problems as much as possible.
 
-OneFlow provides `GradScaler` to use Gradient Scaling in AMP mode. You only need to instantiate a `GradScaler` object in the `__init__` method of the nn.Graph model, and then specify it through the interface [set_grad_scaler](https://oneflow.readthedocs.io/en/master/graph.html#oneflow.nn.Graph.set_grad_scaler). nn.Graph will automatically manage the whole process of Gradient Scaling. Taking the `CustomGraph` above as an example, you need to add the following code to its `__init__` method:
+OneFlow provides `GradScaler` to use Gradient Scaling in AMP mode. You only need to instantiate a `GradScaler` object in the `__init__` method of the nn.Graph model, and then specify it through the interface [set_grad_scaler](https://oneflow.readthedocs.io/en/v0.8.1/generated/oneflow.nn.Graph.set_grad_scaler.html). nn.Graph will automatically manage the whole process of Gradient Scaling. Taking the `CustomGraph` above as an example, you need to add the following code to its `__init__` method:
 
 ```python
 grad_scaler = flow.amp.GradScaler(
