diff --git a/requirements-common.txt b/requirements-common.txt
index 3b8d473c1fe7a..746ba8d31552f 100644
--- a/requirements-common.txt
+++ b/requirements-common.txt
@@ -18,7 +18,7 @@ prometheus_client >= 0.18.0
 prometheus-fastapi-instrumentator >= 7.0.0
 tiktoken >= 0.6.0  # Required for DBRX tokenizer
 lm-format-enforcer == 0.10.3
-outlines >= 0.0.43, < 0.1 # Requires torch >= 2.1.0
+outlines >= 0.0.46, < 0.1 # Requires torch >= 2.1.0
 typing_extensions
 filelock >= 3.10.4 # filelock starts to support `mode` argument from 3.10.4
 pyzmq
diff --git a/vllm/model_executor/guided_decoding/outlines_logits_processors.py b/vllm/model_executor/guided_decoding/outlines_logits_processors.py
index 1c8f6cccb3e9a..c527f23ced1f4 100644
--- a/vllm/model_executor/guided_decoding/outlines_logits_processors.py
+++ b/vllm/model_executor/guided_decoding/outlines_logits_processors.py
@@ -21,12 +21,12 @@
 from typing import Callable, DefaultDict, Dict, List, Union
 
 import torch
-from outlines.caching import cache
 from outlines.fsm.guide import CFGGuide, Generate, Guide, RegexGuide, Write
 from outlines.fsm.json_schema import build_regex_from_schema
 from pydantic import BaseModel
 from transformers import PreTrainedTokenizerBase
 
+from vllm.hpu.utils import with_mark_steps
 
 class BaseLogitsProcessor:
 
@@ -34,6 +34,7 @@ def __init__(self, guide: Guide):
         self._guide: Guide = guide
         self._fsm_state: DefaultDict[int, int] = defaultdict(int)
 
+    @with_mark_steps
     def __call__(self, input_ids: List[int],
                  scores: torch.Tensor) -> torch.Tensor:
         """Use the FSM to bias the logits before sampling the next token."""
@@ -61,14 +62,14 @@ def __call__(self, input_ids: List[int],
                           -math.inf,
                           device=scores.device)
         mask[allowed_tokens] = 0
-        scores.add_(mask)
+        scores = scores.add(mask)
         return scores
 
 
 class RegexLogitsProcessor(BaseLogitsProcessor):
 
     @classmethod
-    @cache()
+    @lru_cache(maxsize=32)
     def _get_guide(cls, regex_string: str,
                    tokenizer: PreTrainedTokenizerBase) -> Guide:
         tokenizer = _adapt_tokenizer(tokenizer)
@@ -127,7 +128,7 @@ def __init__(self, schema: Union[str, Dict, BaseModel],
 class CFGLogitsProcessor(BaseLogitsProcessor):
 
     @classmethod
-    @cache()
+    @lru_cache(maxsize=32)
     def _get_guide(cls, cfg: str, tokenizer: PreTrainedTokenizerBase) -> Guide:
         tokenizer = _adapt_tokenizer(tokenizer)
         return CFGGuide(cfg, tokenizer)