Improve performance docs (#17750)

* add skeleton files

* fix cpu inference link

* add hint to make clear that single gpu section contains general info

* add new files to ToC

* update toctree to have subsection for performance

* add "coming soon" to the still empty sections

* fix missing title

* fix typo

* add reference to empty documents

* Apply suggestions from code review

Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>

* Apply suggestions from code review

Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>

Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>

docs/source/en/perf_train_gpu_many.mdx

@@ -13,6 +13,12 @@ an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express o
 
 When training on a single GPU is too slow or the model weights don't fit in a single GPUs memory we use a mutli-GPU setup. Switching from a single GPU to multiple requires some form of parallelism as the work needs to be distributed. There are several techniques to achieve parallism such as data, tensor, or pipeline parallism. However, there is no one solution to fit them all and which settings works best depends on the hardware you are running on. While the main concepts most likely will apply to any other framework, this article is focused on PyTorch-based implementations.
 
+<Tip>
+
+ Note: Most of the strategies introduced in the [single GPU section](perf_train_gpu_one) (such as mixed precision training or gradient accumulation) are generic and apply to training models in general so make sure to have a look at it before diving into the following sections such as multi-GPU or CPU training.
+
+</Tip>
+
 We will first discuss in depth various 1D parallelism techniques and their pros and cons and then look at how they can be combined into 2D and 3D parallelism to enable an even faster training and to support even bigger models. Various other powerful alternative approaches will be presented.
 
 ## Concepts