fix(scaffolding): use default LLM rather than trt backend LLM

Signed-off-by: Zhenhuan Chen <[email protected]>

Author: dc3671

tensorrt_llm/scaffolding/worker.py

@@ -4,7 +4,7 @@
 import openai
 from transformers import AutoTokenizer
 
-from tensorrt_llm._tensorrt_engine import LLM
+from tensorrt_llm import LLM
 from tensorrt_llm.executor import GenerationExecutor
 from tensorrt_llm.llmapi.llm_args import KvCacheConfig
 from tensorrt_llm.sampling_params import SamplingParams
@@ -150,7 +150,6 @@ def init_with_new_llm(
         )
 
         llm = LLM(model_dir,
-                  backend=backend,
                   tokenizer=tokenizer,
                   mixed_sampler=True,
                   disable_overlap_scheduler=disable_overlap_scheduler,