From 7a865f2325b948f9b6bc6523b1ab4dfe2aa267a0 Mon Sep 17 00:00:00 2001
From: Nick Hill <nhill@redhat.com>
Date: Tue, 28 Oct 2025 04:17:45 -0700
Subject: [PATCH] [V0 Deprecation] Remove vestigial V0 logits_processors.py
 file (#27601)

Signed-off-by: Nick Hill <nhill@redhat.com>
---
 vllm/entrypoints/openai/logits_processors.py | 92 --------------------
 1 file changed, 92 deletions(-)
 delete mode 100644 vllm/entrypoints/openai/logits_processors.py

diff --git a/vllm/entrypoints/openai/logits_processors.py b/vllm/entrypoints/openai/logits_processors.py
deleted file mode 100644
index dedbc23ec83fa..0000000000000
--- a/vllm/entrypoints/openai/logits_processors.py
+++ /dev/null
@@ -1,92 +0,0 @@
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-
-from collections.abc import Iterable
-from functools import lru_cache, partial
-
-import torch
-
-from vllm.sampling_params import LogitsProcessor
-from vllm.transformers_utils.tokenizer import AnyTokenizer
-
-
-class AllowedTokenIdsLogitsProcessor:
-    """Logits processor for constraining generated tokens to a
-    specific set of token ids."""
-
-    def __init__(self, allowed_ids: Iterable[int]):
-        self.allowed_ids: list[int] | None = list(allowed_ids)
-        self.mask: torch.Tensor | None = None
-
-    def __call__(self, token_ids: list[int], logits: torch.Tensor) -> torch.Tensor:
-        if self.mask is None:
-            self.mask = torch.ones(
-                (logits.shape[-1],), dtype=torch.bool, device=logits.device
-            )
-            self.mask[self.allowed_ids] = False
-            self.allowed_ids = None
-        logits.masked_fill_(self.mask, float("-inf"))
-        return logits
-
-
-@lru_cache(maxsize=32)
-def _get_allowed_token_ids_logits_processor(
-    allowed_token_ids: frozenset[int],
-    vocab_size: int,
-) -> LogitsProcessor:
-    if not allowed_token_ids:
-        raise ValueError("Empty allowed_token_ids provided")
-    if not all(0 <= tid < vocab_size for tid in allowed_token_ids):
-        raise ValueError("allowed_token_ids contains out-of-vocab token id")
-    return AllowedTokenIdsLogitsProcessor(allowed_token_ids)
-
-
-def logit_bias_logits_processor(
-    logit_bias: dict[int, float],
-    token_ids: list[int],
-    logits: torch.Tensor,
-) -> torch.Tensor:
-    for token_id, bias in logit_bias.items():
-        logits[token_id] += bias
-    return logits
-
-
-def get_logits_processors(
-    logit_bias: dict[int, float] | dict[str, float] | None,
-    allowed_token_ids: list[int] | None,
-    tokenizer: AnyTokenizer,
-) -> list[LogitsProcessor]:
-    logits_processors: list[LogitsProcessor] = []
-    if logit_bias:
-        try:
-            # Convert token_id to integer
-            # Clamp the bias between -100 and 100 per OpenAI API spec
-            clamped_logit_bias: dict[int, float] = {
-                int(token_id): min(100.0, max(-100.0, bias))
-                for token_id, bias in logit_bias.items()
-            }
-        except ValueError as exc:
-            raise ValueError(
-                "Found token_id in logit_bias that is not "
-                "an integer or string representing an integer"
-            ) from exc
-
-        # Check if token_id is within the vocab size
-        for token_id, bias in clamped_logit_bias.items():
-            if token_id < 0 or token_id >= len(tokenizer):
-                raise ValueError(
-                    f"token_id {token_id} in logit_bias contains out-of-vocab token id"
-                )
-
-        logits_processors.append(
-            partial(logit_bias_logits_processor, clamped_logit_bias)
-        )
-
-    if allowed_token_ids is not None:
-        logits_processors.append(
-            _get_allowed_token_ids_logits_processor(
-                frozenset(allowed_token_ids), len(tokenizer)
-            )
-        )
-
-    return logits_processors