[Model] Skip unused Jina V5 output layers (#52037)

Signed-off-by: kiroxu <148877251+BabyDrangoner@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
This commit is contained in:
kiroxu
2026-08-13 06:14:37 +00:00
committed by GitHub
co-authored by Codex
parent 7bc1660bd6
commit 89c8401c8a
+21 -3
View File
@@ -11,6 +11,8 @@ from safetensors.torch import load as safetensors_load
from torch import nn
from vllm.config import VllmConfig
from vllm.model_executor.layers.logits_processor import LogitsProcessor
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
from vllm.sequence import IntermediateTensors
from vllm.tasks import PoolingTask
from vllm.transformers_utils.repo_utils import get_hf_file_bytes
@@ -26,7 +28,13 @@ from .interfaces import SupportsLateInteraction
from .interfaces_base import VllmModelForPooling
from .llama import LlamaForCausalLM
from .qwen3 import Qwen3ForCausalLM, Qwen3Model
from .utils import AutoWeightsLoader, WeightsMapper, maybe_prefix
from .utils import (
AutoWeightsLoader,
StageMissingLayer,
WeightsMapper,
maybe_prefix,
no_init_weights,
)
logger = logging.getLogger(__name__)
@@ -267,7 +275,12 @@ class JinaEmbeddingsV5DecoderModel(Qwen3ForCausalLM, VllmModelForPooling):
)
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
super().__init__(vllm_config=vllm_config, prefix=prefix)
with no_init_weights(
self,
lambda mod: StageMissingLayer("output", mod),
targets=(LogitsProcessor, ParallelLMHead),
):
super().__init__(vllm_config=vllm_config, prefix=prefix)
_setup_jina_v5_task_and_pooler(self, vllm_config)
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
@@ -289,7 +302,12 @@ class JinaEmbeddingsV5EncoderModel(LlamaForCausalLM, VllmModelForPooling):
)
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
super().__init__(vllm_config=vllm_config, prefix=prefix)
with no_init_weights(
self,
lambda mod: StageMissingLayer("output", mod),
targets=(LogitsProcessor, ParallelLMHead),
):
super().__init__(vllm_config=vllm_config, prefix=prefix)
_setup_jina_v5_task_and_pooler(self, vllm_config)
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]: