[Chore] Use Transformers v5 names and drop redundant processor use_fast (#58550)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Harry Mellor
2026-09-24 14:14:10 +00:00
committed by GitHub
co-authored by Claude Opus 5.5
parent 484c211fe8
commit d881812500
218 changed files with 869 additions and 813 deletions
@@ -4,7 +4,7 @@ import copy
import torch
import torch.nn.functional as F
from transformers.configuration_utils import PretrainedConfig
from transformers.configuration_utils import PreTrainedConfig
from vllm.config import (
VllmConfig,
@@ -37,7 +37,7 @@ from vllm.triton_utils import tl, triton
from vllm.v1.attention.backends.linear_attn import LinearAttentionMetadata
def _build_rope_parameters(config: PretrainedConfig) -> dict | None:
def _build_rope_parameters(config: PreTrainedConfig) -> dict | None:
rope_parameters = copy.deepcopy(getattr(config, "rope_parameters", None)) or {}
if "rope_theta" not in rope_parameters and hasattr(config, "rope_theta"):
rope_parameters["rope_theta"] = config.rope_theta
@@ -385,7 +385,7 @@ class BailingMoELinearAttention(LinearAttention):
# --8<-- [end:bailing_moe_linear_attention]
def __init__(
self,
config: PretrainedConfig,
config: PreTrainedConfig,
vllm_config: VllmConfig,
prefix: str = "linear_attn",
):
@@ -1,7 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import torch
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from vllm.config import (
VllmConfig,
@@ -24,7 +24,7 @@ class LinearAttention(PluggableLayer, MambaBase):
"""Base class for Linear attention layer."""
def __init__(
self, config: PretrainedConfig, vllm_config: VllmConfig, prefix: str = ""
self, config: PreTrainedConfig, vllm_config: VllmConfig, prefix: str = ""
):
super().__init__()
self.layer_idx = extract_layer_index(prefix)