Update User Guide (#2388)

Author: TobyRoseman

docs-guides/_sources/source/flexible-inputs.md

@@ -168,15 +168,15 @@ If the source PyTorch model is exported by [`torch.export.export`](https://pytor
 
 ## Reshape Frequency Optimization Hint
 
-Setting the Reshape Frequency Optimization Hint to `Frequent` can allow flexible shaped models to run on the Neural Engine. This option can be set when loading your model:
+Setting the Reshape Frequency Optimization Hint to `Infrequent` can allow flexible shaped models to run on the Neural Engine, with iOS 17.4 or later.
+This option can be set when loading your model:
 
 ```python
 model = ct.model.MLModel(
-    'path/to/the/saved/model.mlmodel',
-    optimization_hints={ 'reshapeFrequency': ct.ReshapeFrequency.Frequent }
+    'path/to/the/saved/model.mlpackage',
+    optimization_hints={ 'reshapeFrequency': ct.ReshapeFrequency.Infrequent }
 )
-p```
-
+```
 
 ## Enable Unbounded Ranges
 

docs-guides/_sources/source/opt-stable-diffusion.md

@@ -179,4 +179,4 @@ The `sparse_palettized_model` generated above can be converted to produce a join
 | 50% Sparsity, 4-bit (group_size=16) | 933 MB     | <img height="250px" src="images/sdxl-0.5sparse-4bit-gs16.png" width="250px"/> |
 
 To summarize, we were able to use post training palettization to compress SDXL UNet model from `4.8GB` to `1.2GB` in size. We were able to regain accuracy lost from 4-bit palettization by using per grouped channel LUT. 
-With joint compression, we are able to further bring down the model size to `1GB`, while still generating decent accuracy images with 40% sparsity.
+With joint compression, we are able to further bring down the model size to `1GB`, while still generating decent accuracy images with 40% sparsity.

docs-guides/_static/scripts/bootstrap.js.map

@@ -0,0 +1,5 @@
+/*!
+ * Font Awesome Free 6.5.2 by @fontawesome - https://fontawesome.com
+ * License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License)
+ * Copyright 2024 Fonticons, Inc.
+ */