Sparse Semi Sturcutred Tutorial audit (#3599)

Fixes [T237454985](https://www.internalfb.com/intern/tasks/?t=237454985)

## Description
Realized that we have essentially two versions of the same tutorial,
with the only difference being one is a jupyter notebook.


[here](https://docs.pytorch.org/tutorials/unstable/semi_structured_sparse.html)
and
[here](https://docs.pytorch.org/tutorials/advanced/semi_structured_sparse.html)
for reference.

The PR removes the `unstable_source` rst tutorial and makes a couple of
small updates of the jupyter notebook.
## Checklist
<!--- Make sure to add `x` to all items in the following checklist: -->
- [x] The issue that is being fixed is referred in the description (see
above "Fixes #ISSUE_NUMBER")
- [x] Only one issue is addressed in this pull request
- [x] Labels from the issue that this PR is fixing are added to this
pull request
- [x] No unnecessary issues are included into this pull request.

Author: jcaip

advanced_source/semi_structured_sparse.py

@@ -43,6 +43,8 @@
 # -  A NVIDIA GPU with semi-structured sparsity support (Compute
 #    Capability 8.0+).
 #
+#  .. note:: This tutorial is tested on an NVIDIA A100 80GB GPU. You may not see similar speedups on newer GPU architectures, For the latest information on semi-structured sparsity support, please refer to the README `here <https://github.com/pytorch/ao/tree/main/torchao/sparsity#torchao-sparsity>
+#
 # This tutorial is designed for beginners to semi-structured sparsity and
 # sparsity in general. For users with existing 2:4 sparse models,
 # accelerating ``nn.Linear`` layers for inference with
@@ -52,7 +54,6 @@
 import torch
 from torch.sparse import to_sparse_semi_structured, SparseSemiStructuredTensor
 from torch.utils.benchmark import Timer
-SparseSemiStructuredTensor._FORCE_CUTLASS = True
 
 # mask Linear weight to be 2:4 sparse
 mask = torch.Tensor([0, 0, 1, 1]).tile((3072, 2560)).cuda().bool()
@@ -207,7 +208,6 @@
 import transformers
 
 # force CUTLASS use if ``cuSPARSELt`` is not available
-SparseSemiStructuredTensor._FORCE_CUTLASS = True
 torch.manual_seed(100)
 
 # Set default device to "cuda:0"

redirects.py

@@ -48,6 +48,7 @@
     "recipes/torchserve_vertexai_tutorial.html": "../index.html",
     "recipes/amx.html": "../index.html",
     "unstable_source/vulkan_workflow.rst": "../index.html",
+    "unstable/semi_structured_sparse.html": "https://docs.pytorch.org/tutorials/advanced/semi_structured_sparse.html",
     "unstable/skip_param_init.html": "https://docs.pytorch.org/tutorials/recipes/recipes/module_load_state_dict_tips.html",
     "unstable_source/backend_config_tutorial.rst": "../index.html",
 }

unstable_index.rst

@@ -36,15 +36,6 @@ decide if we want to upgrade the level of commitment or to fail fast.
 
 .. Add prototype tutorial cards below this line
 
-.. Sparsity
-
-.. customcarditem::
-   :header: (prototype) Accelerating BERT with semi-structured (2:4) sparsity
-   :card_description: Prune BERT to be 2:4 sparse and accelerate for inference.
-   :image: _static/img/thumbnails/cropped/generic-pytorch-logo.png
-   :link: unstable/semi_structured_sparse.html
-   :tags: Model-Optimiziation
-
 .. vmap
 
 .. customcarditem::