shards llama3 model fail

[AndreaChiChengdu]

🐛 Describe the bug

When I was using the Qualcomm-quantized LLaMa 3 8B model, I found that if I used the sharding number, corresponding to lines 779 and 797 in the file example/model/export_llama_lib.py, I needed to import 'canonicalize_program' from 'executorch.backends.qualcomm.utils.utils'. However, there is no module named 'canonicalize_program' in the corresponding Qualcomm backend utils.py. Where could the issue be?
https://github.com/pytorch/executorch/blob/main/backends/qualcomm/utils/utils.py
thanks and Merry Christmas！

Versions

v0.5 main branch

[AndreaChiChengdu]

I found that this bug has been discussed in #6955 ，but not fixed