[Platform][Refactor] Extract func get_default_attn_backend to Platform (#10358)

Signed-off-by: Mengqing Cao <cmq0113@163.com>
This commit is contained in:
Mengqing Cao
2024-11-19 11:22:26 +08:00
committed by GitHub
parent 7eb719df13
commit 8c1fb50705
14 changed files with 99 additions and 69 deletions
+1 -1
View File
@@ -13,7 +13,6 @@ from torch.nn import functional as F
from transformers import PretrainedConfig
from vllm.attention import Attention, AttentionMetadata
from vllm.attention.selector import _Backend
from vllm.compilation.decorators import support_torch_compile
from vllm.config import CacheConfig, VllmConfig
from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
@@ -38,6 +37,7 @@ from vllm.model_executor.layers.vocab_parallel_embedding import (
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalKwargs
from vllm.multimodal.utils import cached_get_tokenizer
from vllm.platforms import _Backend
from vllm.sequence import (VLLM_TOKEN_ID_ARRAY_TYPE, IntermediateTensors,
SequenceData)
from vllm.transformers_utils.processor import get_processor
+1 -1
View File
@@ -39,7 +39,6 @@ from transformers.models.qwen2_vl.image_processing_qwen2_vl import (
make_batched_images, make_batched_videos, smart_resize)
from vllm.attention import AttentionMetadata
from vllm.attention.selector import _Backend
from vllm.config import VllmConfig
from vllm.distributed import get_pp_group, parallel_state
from vllm.distributed import utils as dist_utils
@@ -65,6 +64,7 @@ from vllm.multimodal.image import cached_get_image_processor
from vllm.multimodal.inputs import (MultiModalData, MultiModalDataDict,
MultiModalKwargs)
from vllm.multimodal.utils import cached_get_tokenizer
from vllm.platforms import _Backend
from vllm.sequence import IntermediateTensors, PoolerOutput, SequenceData
from vllm.transformers_utils.config import uses_mrope
from vllm.transformers_utils.processor import cached_get_processor
+2 -2
View File
@@ -9,13 +9,13 @@ from torch.func import functional_call
from transformers import PretrainedConfig
import vllm.envs as envs
from vllm.attention.selector import (_Backend, backend_name_to_enum,
from vllm.attention.selector import (backend_name_to_enum,
get_global_forced_attn_backend)
from vllm.config import VllmConfig
from vllm.logger import init_logger
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.multimodal import MultiModalPlaceholderMap, NestedTensors
from vllm.platforms import current_platform
from vllm.platforms import _Backend, current_platform
from vllm.sequence import IntermediateTensors
from vllm.utils import is_pin_memory_available