Fix link to pytorch documents (pytorch#1294)

* Fix link to pytorch documents

* Fix too long lines

Co-authored-by: vfdev <[email protected]>

examples/notebooks/FashionMNIST.ipynb

@@ -234,10 +234,10 @@
       "source": [
         "Explanation of Model Architecture\n",
         "\n",
-        "* [Convolutional layers](https://pytorch.org/docs/stable/nn.html#conv2d), the Convolutional layer is used to create a convolution kernel that is convolved with the layer input to produce a tensor of outputs.\n",
-        "* [Maxpooling layers](https://pytorch.org/docs/stable/nn.html#maxpool2d), the Maxpooling layer is used to downsample an input representation keeping the most active pixels from the previous layer.\n",
-        "* The usual [Linear](https://pytorch.org/docs/stable/nn.html#linear) + [Dropout](https://pytorch.org/docs/stable/nn.html#dropout2d) layers to avoid overfitting and produce a 10-dim output.\n",
-        "* We had used [Relu](https://pytorch.org/docs/stable/nn.html#id27) Non Linearity for the model and [logsoftmax](https://pytorch.org/docs/stable/nn.html#log-softmax) at the last layer because we are going to use the [NLLL loss](https://pytorch.org/docs/stable/nn.html#nllloss).\n"
+        "* [Convolutional layers](https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html), the Convolutional layer is used to create a convolution kernel that is convolved with the layer input to produce a tensor of outputs.\n",
+        "* [Maxpooling layers](https://pytorch.org/docs/stable/generated/torch.nn.MaxPool2d.html), the Maxpooling layer is used to downsample an input representation keeping the most active pixels from the previous layer.\n",
+        "* The usual [Linear](https://pytorch.org/docs/stable/generated/torch.nn.Linear.html) + [Dropout](https://pytorch.org/docs/stable/generated/torch.nn.Dropout2d.html) layers to avoid overfitting and produce a 10-dim output.\n",
+        "* We had used [Relu](https://pytorch.org/docs/stable/generated/torch.nn.ReLU.html) Non Linearity for the model and [logsoftmax](https://pytorch.org/docs/stable/generated/torch.nn.LogSoftmax.html) at the last layer because we are going to use the [NLLL loss](https://pytorch.org/docs/stable/generated/torch.nn.NLLLoss.html).\n"
       ]
     },
     {

ignite/contrib/handlers/tensorboard_logger.py

@@ -384,16 +384,16 @@ class TensorboardLogger(BaseLogger):
 
     otherwise, it falls back to using
     `PyTorch's SummaryWriter
-    <https://pytorch.org/docs/stable/tensorboard.html#torch.utils.tensorboard.writer.SummaryWriter>`_
+    <https://pytorch.org/docs/stable/tensorboard.html>`_
     (>=v1.2.0).
 
     Args:
         *args: Positional arguments accepted from
             `SummaryWriter
-            <https://pytorch.org/docs/stable/tensorboard.html#torch.utils.tensorboard.writer.SummaryWriter>`_.
+            <https://pytorch.org/docs/stable/tensorboard.html>`_.
         **kwargs: Keyword arguments accepted from
             `SummaryWriter
-            <https://pytorch.org/docs/stable/tensorboard.html#torch.utils.tensorboard.writer.SummaryWriter>`_.
+            <https://pytorch.org/docs/stable/tensorboard.html>`_.
             For example, `log_dir` to setup path to the directory where to log.
 
     Examples:

ignite/distributed/utils.py

@@ -293,7 +293,7 @@ def train_fn(local_rank, a, b, c, d=12):
               | and `node_rank=0` are tolerated and ignored, otherwise an exception is raised.
 
     .. _dist.init_process_group: https://pytorch.org/docs/stable/distributed.html#torch.distributed.init_process_group
-    .. _mp.start_processes: https://pytorch.org/docs/stable/_modules/torch/multiprocessing/spawn.html#spawn
+    .. _mp.start_processes: https://pytorch.org/docs/stable/multiprocessing.html#torch.multiprocessing.spawn
     .. _xmp.spawn: http://pytorch.org/xla/release/1.6/index.html#torch_xla.distributed.xla_multiprocessing.spawn
     .. _hvd_run: https://horovod.readthedocs.io/en/latest/api.html#module-horovod.run
 

ignite/handlers/checkpoint.py

@@ -99,8 +99,9 @@ class Checkpoint(Serializable):
         include_self (bool): Whether to include the `state_dict` of this object in the checkpoint. If `True`, then
             there must not be another object in ``to_save`` with key ``checkpointer``.
 
-    .. _DistributedDataParallel: https://pytorch.org/docs/stable/nn.html#torch.nn.parallel.DistributedDataParallel
-    .. _DataParallel: https://pytorch.org/docs/stable/nn.html#torch.nn.DataParallel
+    .. _DistributedDataParallel: https://pytorch.org/docs/stable/generated/
+        torch.nn.parallel.DistributedDataParallel.html
+    .. _DataParallel: https://pytorch.org/docs/stable/generated/torch.nn.DataParallel.html
 
     Note:
         This class stores a single file as a dictionary of provided objects to save.
@@ -475,8 +476,9 @@ def load_objects(to_load: Mapping, checkpoint: Mapping, **kwargs) -> None:
             **kwargs: Keyword arguments accepted for `nn.Module.load_state_dict()`. Passing `strict=False` enables
                 the user to load part of the pretrained model (useful for example, in Transfer Learning)
 
-        .. _DistributedDataParallel: https://pytorch.org/docs/stable/nn.html#torch.nn.parallel.DistributedDataParallel
-        .. _DataParallel: https://pytorch.org/docs/stable/nn.html#torch.nn.DataParallel
+        .. _DistributedDataParallel: https://pytorch.org/docs/stable/generated/
+            torch.nn.parallel.DistributedDataParallel.html
+        .. _DataParallel: https://pytorch.org/docs/stable/generated/torch.nn.DataParallel.html
 
         """
         Checkpoint._check_objects(to_load, "load_state_dict")