[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
@@ -14,7 +14,7 @@ from types import SimpleNamespace
from typing import cast
import pytest
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from vllm.config import ModelConfig
from vllm.model_executor.models.config import (
@@ -23,7 +23,7 @@ from vllm.model_executor.models.config import (
)
def _model_config(hf_config: PretrainedConfig) -> ModelConfig:
def _model_config(hf_config: PreTrainedConfig) -> ModelConfig:
"""Minimal stand-in for ModelConfig; only hf_config is read."""
return cast(ModelConfig, SimpleNamespace(hf_config=hf_config))
@@ -42,7 +42,7 @@ def test_encoder_backbone_enables_bidirectional_attention():
EncoderOnlyAttention; JinaEmbeddingsV5Model then dispatches to the encoder
implementation.
"""
hf_config = PretrainedConfig(is_decoder=False)
hf_config = PreTrainedConfig(is_decoder=False)
JinaEmbeddingsV5ModelConfig.verify_and_update_model_config(_model_config(hf_config))
@@ -55,13 +55,13 @@ def test_supported_decoder_backbone_is_accepted():
`-small` omits `is_decoder` entirely, so an absent attribute has to be
treated as a decoder. The first assertion pins that assumption: if
PretrainedConfig ever gains an `is_decoder=False` default, this fails here
PreTrainedConfig ever gains an `is_decoder=False` default, this fails here
rather than silently rejecting a supported checkpoint.
"""
absent = PretrainedConfig()
absent = PreTrainedConfig()
assert not hasattr(absent, "is_decoder")
JinaEmbeddingsV5ModelConfig.verify_and_update_model_config(_model_config(absent))
JinaEmbeddingsV5ModelConfig.verify_and_update_model_config(
_model_config(PretrainedConfig(is_decoder=True))
_model_config(PreTrainedConfig(is_decoder=True))
)
@@ -22,19 +22,19 @@ from unittest.mock import MagicMock
import numpy as np
import pytest
import torch
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from tests.models.registry import HF_EXAMPLE_MODELS
from vllm.config.multimodal import MultiModalDummyOptions
class MockAudioFlamingo3Config(PretrainedConfig):
class MockAudioFlamingo3Config(PreTrainedConfig):
model_type = "audioflamingo3"
def __init__(self, **kwargs):
super().__init__(**kwargs)
self.audio_config = PretrainedConfig()
self.text_config = PretrainedConfig()
self.audio_config = PreTrainedConfig()
self.text_config = PreTrainedConfig()
class MockAudioFlamingo3Processor:
@@ -8,7 +8,7 @@ import pytest
import torch
from PIL import Image
from tokenizers import Tokenizer, decoders, models, pre_tokenizers
from transformers import PreTrainedTokenizerFast
from transformers import TokenizersBackend
from vllm.model_executor.models.bailing_moe_v3_vl import (
IMAGE_PLACEHOLDER,
@@ -34,7 +34,7 @@ def processor():
add_prefix_space=False, use_regex=False
)
backend.decoder = decoders.ByteLevel()
tokenizer = PreTrainedTokenizerFast(
tokenizer = TokenizersBackend(
tokenizer_object=backend,
additional_special_tokens=[
"<|vision_start|>",
@@ -6,7 +6,7 @@ from collections.abc import Mapping
import pytest
from PIL import Image
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.image import rescale_image_size
@@ -17,7 +17,7 @@ from ...utils import build_model_context
def _get_expected_num_patches(
config: PretrainedConfig,
config: PreTrainedConfig,
image: Image.Image,
num_imgs: int,
min_num: int,
@@ -6,7 +6,7 @@ from collections.abc import Mapping
import pytest
from PIL import Image
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.image import rescale_image_size
@@ -17,7 +17,7 @@ from ...utils import build_model_context
def _get_expected_num_patches(
config: PretrainedConfig,
config: PreTrainedConfig,
image: Image.Image,
num_imgs: int,
min_num: int,
@@ -6,7 +6,7 @@ from collections.abc import Mapping
import pytest
from PIL import Image
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.image import rescale_image_size
@@ -17,7 +17,7 @@ from ...utils import build_model_context
def _get_expected_num_patches(
config: PretrainedConfig,
config: PreTrainedConfig,
image: Image.Image,
num_imgs: int,
min_num: int,
+2 -2
View File
@@ -7,7 +7,7 @@ from typing import Any, Literal
import pytest
from packaging.version import Version
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from transformers import __version__ as TRANSFORMERS_VERSION
from vllm.config.model import ModelDType, TokenizerMode
@@ -1233,7 +1233,7 @@ _MULTIMODAL_EXAMPLE_MODELS = {
# NemotronH layers are constructed via `hybrid_override_pattern`
use_original_num_layers=True,
hf_overrides={
"vision_config": PretrainedConfig(
"vision_config": PreTrainedConfig(
args={
"min_num_patches": 1,
"max_num_patches": 12,
+3 -3
View File
@@ -4,7 +4,7 @@
import pytest
import torch
from transformers import Gemma3Config, PretrainedConfig, Qwen2Config
from transformers import Gemma3Config, PreTrainedConfig, Qwen2Config
from vllm.model_executor.models.adapters import (
_create_pooling_model_cls,
@@ -266,14 +266,14 @@ def test_resolve_num_labels_defaults_when_undeclared():
config = _composite_config()
assert (
_resolve_num_labels(config, config.get_text_config())
== PretrainedConfig().num_labels
== PreTrainedConfig().num_labels
)
def test_resolve_num_labels_declared_on_outer_config():
"""Multimodal checkpoints keep id2label/problem_type on the top-level config."""
config = _composite_config(outer_labels=20)
assert config.get_text_config().num_labels == PretrainedConfig().num_labels
assert config.get_text_config().num_labels == PreTrainedConfig().num_labels
assert _resolve_num_labels(config, config.get_text_config()) == 20
+5 -5
View File
@@ -5,7 +5,7 @@
from typing import Any
import pytest
from transformers import AutoConfig, PretrainedConfig
from transformers import AutoConfig, PreTrainedConfig
from vllm.config.speculative import SpeculativeConfig
@@ -15,19 +15,19 @@ _CHECKPOINTS = {
}
def _mtp_config(model_type: str) -> PretrainedConfig:
def _mtp_config(model_type: str) -> PreTrainedConfig:
"""Create a top-level MTP configuration with mtp_num_hidden_layers."""
kwargs: dict[str, Any] = {
"model_type": model_type,
"architectures": ["SomeArch"],
"mtp_num_hidden_layers": 1,
}
return PretrainedConfig(**kwargs)
return PreTrainedConfig(**kwargs)
def _multimodal_wrapper_mtp_config(
model_type: str, mtp_layers: int = 1
) -> PretrainedConfig:
) -> PreTrainedConfig:
"""Download a multimodal wrapper config via AutoConfig.from_pretrained
and configure mtp_num_hidden_layers in text_config.
@@ -36,7 +36,7 @@ def _multimodal_wrapper_mtp_config(
- MoE (qwen3_5_moe): Qwen/Qwen3.6-35B-A3B
"""
repo = _CHECKPOINTS[model_type]
config: PretrainedConfig = AutoConfig.from_pretrained(repo)
config: PreTrainedConfig = AutoConfig.from_pretrained(repo)
text_config = config.get_text_config()
text_config.mtp_num_hidden_layers = mtp_layers
return config
+2 -2
View File
@@ -13,7 +13,7 @@ from typing import Any
import pytest
import torch
import torch.nn as nn
from transformers import AutoConfig, AutoModel, PretrainedConfig
from transformers import AutoConfig, AutoModel, PreTrainedConfig
from vllm.config import ModelConfig, VllmConfig
from vllm.model_executor.models.interfaces import SupportsMultiModal
@@ -524,7 +524,7 @@ def replace_vocab_embeddings(model, **config_kwargs):
stub = nn.Module()
stub.model = model
stub.config = PretrainedConfig(
stub.config = PreTrainedConfig(
vocab_size=VOCAB_SIZE, num_positions=NUM_POSITIONS, **config_kwargs
)
embeddings = Base._vocab_embeddings(stub)
+3 -3
View File
@@ -8,7 +8,7 @@ from typing import Any
import torch
import torch.nn.functional as F
from transformers import PretrainedConfig
from transformers import PreTrainedConfig
from vllm.config.model import AttnTypeStr, ModelConfig, ModelDType, RunnerOption
from vllm.config.pooler import SequencePoolingType, TokenPoolingType
@@ -446,12 +446,12 @@ def get_vllm_extra_kwargs(model_info: ModelInfo, vllm_extra_kwargs):
def dummy_hf_overrides(
hf_config: PretrainedConfig,
hf_config: PreTrainedConfig,
*,
model_arch: str = "",
exist_overrides: dict[str, Any] | None = None,
use_original_num_layers: bool = False,
) -> PretrainedConfig:
) -> PreTrainedConfig:
"""Dummy HF overrides function used to create dummy model
with only minimum nums of layer.
"""