Fix for issue #81

[horheynm]

SUMMARY:
Fix for #81

steps:

1. clone repo
2. create and activate virtual eng
3. pip3 install -e .[dev]
4. change llm-compressor/examples/big_model_offloading/big_model_fp8.py line 34 from
   model_stub = "meta-llama/Meta-Llama-3-70B-Instruct"
   to
   model_stub = "microsoft/Phi-3-medium-128k-instruct"
5. run the code with

# determine which layers to offload to cpu based on available resources
device_map = calculate_offload_device_map(
    model_stub, reserve_for_hessians=False, num_gpus=1, torch_dtype=torch.float16, trust_remote_code=True
)

the output config.json will get rid of line 201 from
"type": "longrope"
to
"type": "su"

TEST PLAN:
Run the code to completion, make sure that the issue does not come up.

[Lin-K76]

Tested changes, fixes error regarding the rope type and the configuration python files.

[Lin-K76]

Looks good to me!

[Lin-K76]

Example changes look good to me