Example (llm): layerwise rotations can't be combined with FX

[nickfraser]

If you do, you get this:

Applying model quantization...                                                                                                                                                                                                                 
Traceback (most recent call last):                                                                                                                                                                                                             
  File "/scratch/users/nfraser/workspace/brevitas/src/brevitas_examples/llm/main.py", line 856, in <module>                                                                                                                                    
    main(args)                                                                                                                                                                                                                                 
  File "/scratch/users/nfraser/workspace/brevitas/src/brevitas_examples/llm/main.py", line 377, in main                                                                                                                                        
    last_layer_kwargs = layer_map[type(last_module)][1]                                                                                                                                                                                        
                        ~~~~~~~~~^^^^^^^^^^^^^^^^^^^                                                                                                                                                                                           
KeyError: <class 'brevitas.nn.equalized_layer.RotatedModule'>

I suspect the same will happen with SmoothQuant.