[Bugfix] Fix transform_config parsing in Compressed Tensors

[kylesayrs]

Purpose

• Fix compressed tensors config parsing

Traceback (most recent call last):
  File "/home/kyle/vllm/vllm/entrypoints/llm.py", line 270, in __init__
    self.llm_engine = LLMEngine.from_engine_args(
  File "/home/kyle/vllm/vllm/engine/llm_engine.py", line 485, in from_engine_args
    vllm_config = engine_args.create_engine_config(usage_context)
  File "/home/kyle/vllm/vllm/engine/arg_utils.py", line 1382, in create_engine_config
    config = VllmConfig(
  File "/home/kyle/vllm/env/lib/python3.10/site-packages/pydantic/_internal/_dataclasses.py", line 123, in __init__
    s.__pydantic_validator__.validate_python(ArgsKwargs(args, kwargs), self_instance=s)
pydantic_core._pydantic_core.ValidationError: 1 validation error for VllmConfig
config_groups
  Field required [type=missing, input_value={}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/missing

"transform_config": {},

Changes

• Support parsing transform_configs which are empty dictionaries (by skipping them)

Testing

• The following model fails on main, but is runnable on this branch

from vllm import LLM
llm = LLM("nm-testing/TinyLlama-1.1B-Chat-v1.0-W8A8-Static-Asym-e2e")
llm.generate("Hello there")

[mgoin]

Thanks!

[yewentao256]

LGTM, thanks for the work!

[DarkLight1337]

Can you merge from main to fix the CI failures?

[tlrmchlsmth]

Thank you!