From cdc4824a21eaa986d4d1fee90a7e6465c9f706e6 Mon Sep 17 00:00:00 2001 From: wangxiyuan Date: Fri, 14 Aug 2026 23:15:43 +0800 Subject: [PATCH] [Misc] Remove `override_attention_dtype` (#48684) Signed-off-by: wangxiyuan Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> --- vllm/config/model.py | 10 ---------- vllm/engine/arg_utils.py | 5 ----- 2 files changed, 15 deletions(-) diff --git a/vllm/config/model.py b/vllm/config/model.py index b0dd48b7e5..fcdce3ef58 100644 --- a/vllm/config/model.py +++ b/vllm/config/model.py @@ -1,7 +1,6 @@ # SPDX-License-Identifier: Apache-2.0 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project -import warnings from collections.abc import Callable from dataclasses import InitVar, field from functools import cached_property @@ -349,8 +348,6 @@ class ModelConfig: - "transformers" will use the Transformers model implementation. - "terratorch" will use the TerraTorch model implementation. """ - override_attention_dtype: str | None = None - """Override dtype for attention""" logits_processors: list[str | type[LogitsProcessor]] | None = None """One or more logits processors' fully-qualified class names or class definitions""" @@ -433,7 +430,6 @@ class ModelConfig: "config_format", "hf_token", "hf_overrides", - "override_attention_dtype", "logits_processors", "io_processor_plugin", "pooler_config", @@ -595,12 +591,6 @@ class ModelConfig: self.hf_token, ) - if self.override_attention_dtype is not None and not current_platform.is_rocm(): - warnings.warn( - "override-attention-dtype is set but not using ROCm platform", - stacklevel=2, - ) - if self.enable_sleep_mode: if not current_platform.is_sleep_mode_available(): raise ValueError("Sleep mode is not supported on current platform.") diff --git a/vllm/engine/arg_utils.py b/vllm/engine/arg_utils.py index 528c884d73..44ee837c6f 100644 --- a/vllm/engine/arg_utils.py +++ b/vllm/engine/arg_utils.py @@ -700,7 +700,6 @@ class EngineArgs: ModelConfig, "override_generation_config" ) model_impl: str = ModelConfig.model_impl - override_attention_dtype: str | None = ModelConfig.override_attention_dtype attention_backend: AttentionBackendEnum | None = AttentionConfig.backend kv_cache_dtype_skip_layers: list[str] = get_field( @@ -917,9 +916,6 @@ class EngineArgs: "--enable-cumem-allocator", **model_kwargs["enable_cumem_allocator"] ) model_group.add_argument("--model-impl", **model_kwargs["model_impl"]) - model_group.add_argument( - "--override-attention-dtype", **model_kwargs["override_attention_dtype"] - ) model_group.add_argument( "--logits-processors", **model_kwargs["logits_processors"] ) @@ -1787,7 +1783,6 @@ class EngineArgs: enable_sleep_mode=self.enable_sleep_mode, enable_cumem_allocator=self.enable_cumem_allocator, model_impl=self.model_impl, - override_attention_dtype=self.override_attention_dtype, logits_processors=self.logits_processors, video_pruning_rate=self.video_pruning_rate, video_pruning_method=self.video_pruning_method,