fix: move to VLLM

backends/smoldocling.py

@@ -16,9 +16,7 @@ PROMPT_TEXT = "Convert page to Docling."
 
 # Initialize processor and model
 # Initialize LLM
-llm = LLM(
-    model=MODEL_PATH, limit_mm_per_prompt={"image": 1}, gpu_memory_utilization=0.4
-)
+llm = LLM(model=MODEL_PATH, limit_mm_per_prompt={"image": 1})
 chat_template = f"<|im_start|>User:<image>{PROMPT_TEXT}<end_of_utterance>\nAssistant:"
 
 sampling_params = SamplingParams(