To Reproduce Steps to reproduce the behavior:

1. Start latest AMD image
2. init config and download model
3. Run

   ilab model serve

4. Get traceback:

    $ ilab model serve
   [sudo] password for tflink: 
   INFO 2024-11-26 17:13:13,106 instructlab.model.serve_backend:56: Using model '/home/tflink/.cache/instructlab/models/granite-8b-lab-v1' with -1 gpu-layers and 4096 max context size.
   INFO 2024-11-26 17:13:13,106 instructlab.model.serve_backend:88: '--gpus' flag used alongside '--tensor-parallel-size' in the vllm_args section of the config file. Using value of the --gpus flag.
   INFO 2024-11-26 17:13:13,107 instructlab.model.backends.vllm:313: vLLM starting up on pid 4 at http://127.0.0.1:8000/v1
   INFO 11-26 17:13:17 importing.py:10] Triton not installed; certain GPU-related functions will not be available.
   Traceback (most recent call last):
     File "<frozen runpy>", line 189, in _run_module_as_main
     File "<frozen runpy>", line 112, in _get_module_details
     File "/opt/app-root/lib64/python3.11/site-packages/vllm/__init__.py", line 3, in <module>
       from vllm.engine.arg_utils import AsyncEngineArgs, EngineArgs
     File "/opt/app-root/lib64/python3.11/site-packages/vllm/engine/arg_utils.py", line 11, in <module>
       from vllm.config import (CacheConfig, ConfigFormat, DecodingConfig,
     File "/opt/app-root/lib64/python3.11/site-packages/vllm/config.py", line 16, in <module>
       from vllm.transformers_utils.config import (ConfigFormat, get_config,
     File "/opt/app-root/lib64/python3.11/site-packages/vllm/transformers_utils/config.py", line 11, in <module>
       from transformers.models.auto.image_processing_auto import (
     File "/opt/app-root/lib64/python3.11/site-packages/transformers/models/auto/image_processing_auto.py", line 27, in <module>
       from ...image_processing_utils import BaseImageProcessor, ImageProcessingMixin
     File "/opt/app-root/lib64/python3.11/site-packages/transformers/image_processing_utils.py", line 21, in <module>
       from .image_transforms import center_crop, normalize, rescale
     File "/opt/app-root/lib64/python3.11/site-packages/transformers/image_transforms.py", line 22, in <module>
       from .image_utils import (
     File "/opt/app-root/lib64/python3.11/site-packages/transformers/image_utils.py", line 58, in <module>
       from torchvision.transforms import InterpolationMode
     File "/opt/app-root/lib64/python3.11/site-packages/torchvision/__init__.py", line 6, in <module>
       from torchvision import _meta_registrations, datasets, io, models, ops, transforms, utils
     File "/opt/app-root/lib64/python3.11/site-packages/torchvision/_meta_registrations.py", line 163, in <module>
       @torch._custom_ops.impl_abstract("torchvision::nms")
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
     File "/opt/app-root/lib64/python3.11/site-packages/torch/library.py", line 654, in register
       use_lib._register_fake(op_name, func, _stacklevel=stacklevel + 1)
     File "/opt/app-root/lib64/python3.11/site-packages/torch/library.py", line 154, in _register_fake
       handle = entry.abstract_impl.register(func_to_register, source)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
     File "/opt/app-root/lib64/python3.11/site-packages/torch/_library/abstract_impl.py", line 31, in register
       if torch._C._dispatch_has_kernel_for_dispatch_key(self.qualname, "Meta"):
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   RuntimeError: operator torchvision::nms does not exist

Expected behavior

• ilab model serve should work

Device Info (please complete the following information):

• Hardware Specs: x86_64, x8 MI300X
• OS Version: RHEL AI 1.3 pre-release
• Python Version: Python 3.11.7
• InstructLab Version: ilab, version 0.21.0

Additional context

• Tested with instructlab-amd-rhel9:1.3-1732577202 and bootc-amd-rhel9:1.3-1732609273