Docs/improved multigpu doc clarity (Lightning-AI#3194)

* + clarifying doc

* improved documentation clarity

* docs minor correction

* remove device from type_as

Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>

Author: humandotlearning

docs/source/multi_gpu.rst

@@ -48,7 +48,7 @@ This will make your code scale to any arbitrary number of GPUs or TPUs with Ligh
     # with lightning
     def forward(self, x):
         z = torch.Tensor(2, 3)
-        z = z.type_as(x, device=self.device)
+        z = z.type_as(x)
 
 The :class:`~pytorch_lightning.core.lightning.LightningModule` knows what device it is on. You can access the reference via `self.device`.
 Sometimes it is necessary to store tensors as module attributes. However, if they are not parameters they will
@@ -149,7 +149,7 @@ a comma separated list of GPU ids:
 .. testcode::
     :skipif: torch.cuda.device_count() < 2
 
-    # DEFAULT (int) specifies how many GPUs to use
+    # DEFAULT (int) specifies how many GPUs to use per node
     Trainer(gpus=k)
 
     # Above is equivalent to

pytorch_lightning/trainer/trainer.py

@@ -233,7 +233,7 @@ def __init__(
 
             num_nodes: number of GPU nodes for distributed training.
 
-            gpus: Which GPUs to train on.
+            gpus: number of gpus to train on (int) or which GPUs to train on (list or str) applied per node
 
             auto_select_gpus: