[1/N][Attention] Restructure attention: move files (#31916)
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
@@ -9,7 +9,6 @@ from itertools import islice
|
||||
import torch
|
||||
from torch import nn
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig, get_current_vllm_config
|
||||
@@ -50,6 +49,7 @@ from vllm.model_executor.models.utils import (
|
||||
maybe_prefix,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
@@ -8,10 +8,10 @@ from collections.abc import Iterable
|
||||
import torch
|
||||
import torch.nn as nn
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.distributed.utils import divide
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
|
||||
@@ -32,13 +32,14 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import ApertusConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
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_world_size
|
||||
from vllm.model_executor.layers.activation import XIELU
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
@@ -57,6 +58,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
maybe_remap_kv_scale_name,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import SupportsLoRA, SupportsPP
|
||||
from .utils import (
|
||||
|
||||
@@ -7,11 +7,13 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import BertConfig
|
||||
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, PoolerConfig, VllmConfig
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
QKVParallelLinear,
|
||||
|
||||
@@ -6,7 +6,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.distributed import (
|
||||
@@ -16,6 +15,9 @@ from vllm.distributed import (
|
||||
tensor_model_parallel_all_reduce,
|
||||
)
|
||||
from vllm.model_executor.layers.activation import get_act_and_mul_fn, get_act_fn
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.fused_moe import activation_without_mul, fused_topk
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -9,9 +9,9 @@ import torch
|
||||
import torch.nn as nn
|
||||
from transformers import Blip2VisionConfig, BlipVisionConfig
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.distributed import divide, get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -15,11 +15,11 @@ from transformers import (
|
||||
)
|
||||
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions, MultiModalConfig
|
||||
from vllm.distributed import divide, get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -4,12 +4,12 @@ from copy import deepcopy
|
||||
from math import lcm
|
||||
from typing import TYPE_CHECKING
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.models import ModelRegistry
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils.math_utils import cdiv, round_up
|
||||
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.v1.kv_cache_interface import FullAttentionSpec, MambaSpec, MLAAttentionSpec
|
||||
|
||||
if TYPE_CHECKING:
|
||||
|
||||
@@ -18,8 +18,8 @@ import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
from transformers import CLIPVisionConfig
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import MultiModalConfig
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.quantization import QuantizationConfig
|
||||
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
|
||||
|
||||
@@ -33,9 +33,7 @@ from torch import nn
|
||||
from transformers import DeepseekV2Config, DeepseekV3Config
|
||||
|
||||
from vllm._aiter_ops import rocm_aiter_ops
|
||||
from vllm.attention.backends.abstract import AttentionBackend
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.ops.common import pack_seq_triton, unpack_seq_triton
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, ParallelConfig, VllmConfig, get_current_vllm_config
|
||||
from vllm.distributed import (
|
||||
@@ -78,10 +76,12 @@ from vllm.platforms import current_platform
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.deep_gemm import fp8_mqa_logits, fp8_paged_mqa_logits
|
||||
from vllm.utils.torch_utils import direct_register_custom_op
|
||||
from vllm.v1.attention.backend import AttentionBackend
|
||||
from vllm.v1.attention.backends.mla.indexer import (
|
||||
DeepseekV32IndexerBackend,
|
||||
DeepseekV32IndexerMetadata,
|
||||
)
|
||||
from vllm.v1.attention.ops.common import pack_seq_triton, unpack_seq_triton
|
||||
from vllm.v1.kv_cache_interface import KVCacheSpec, MLAAttentionSpec
|
||||
from vllm.v1.worker.workspace import current_workspace_manager
|
||||
|
||||
@@ -679,7 +679,9 @@ def sparse_attn_indexer(
|
||||
)
|
||||
fp8_mqa_logits_func = fp8_mqa_logits
|
||||
if current_platform.is_rocm():
|
||||
from vllm.attention.ops.rocm_aiter_mla_sparse import rocm_fp8_mqa_logits
|
||||
from vllm.v1.attention.ops.rocm_aiter_mla_sparse import (
|
||||
rocm_fp8_mqa_logits,
|
||||
)
|
||||
|
||||
fp8_mqa_logits_func = rocm_fp8_mqa_logits
|
||||
logits = fp8_mqa_logits_func(
|
||||
@@ -729,7 +731,7 @@ def sparse_attn_indexer(
|
||||
num_padded_tokens = batch_size * next_n
|
||||
fp8_paged_mqa_logits_func = fp8_paged_mqa_logits
|
||||
if current_platform.is_rocm():
|
||||
from vllm.attention.ops.rocm_aiter_mla_sparse import (
|
||||
from vllm.v1.attention.ops.rocm_aiter_mla_sparse import (
|
||||
rocm_fp8_paged_mqa_logits,
|
||||
)
|
||||
|
||||
|
||||
@@ -8,10 +8,6 @@ import torch.nn as nn
|
||||
from torch.nn import LayerNorm
|
||||
from transformers.models.qwen2_vl import Qwen2VLProcessor
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.attention.layers.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import utils as dist_utils
|
||||
@@ -20,6 +16,9 @@ from vllm.distributed.parallel_state import (
|
||||
get_tensor_model_parallel_world_size,
|
||||
)
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
@@ -59,6 +58,7 @@ from vllm.multimodal.inputs import MultiModalDataDict
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.transformers_utils.configs.dotsocr import DotsOCRConfig, DotsVisionConfig
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .vision import run_dp_sharded_mrope_vision_model
|
||||
|
||||
|
||||
@@ -36,16 +36,15 @@ import torch.nn.functional as F
|
||||
from einops import rearrange
|
||||
from transformers import BatchFeature
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.attention.layers.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions, VideoDummyOptions
|
||||
from vllm.distributed import parallel_state
|
||||
from vllm.distributed import utils as dist_utils
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import QuickGELU
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
@@ -74,6 +73,7 @@ from vllm.multimodal.processing import (
|
||||
from vllm.multimodal.profiling import BaseDummyInputsBuilder
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .ernie45_vl_moe import Ernie4_5_VLMoeForCausalLM
|
||||
from .interfaces import (
|
||||
|
||||
@@ -22,13 +22,15 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import Gemma3TextConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
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_world_size
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import GeluAndMul
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.layernorm import GemmaRMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
MergedColumnParallelLinear,
|
||||
@@ -47,8 +49,8 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
maybe_remap_kv_scale_name,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from ...attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
from .interfaces import SupportsLoRA, SupportsPP
|
||||
from .utils import (
|
||||
AutoWeightsLoader,
|
||||
|
||||
@@ -29,7 +29,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import Glm4Config
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
@@ -41,6 +40,7 @@ from vllm.model_executor.layers.quantization import QuantizationConfig
|
||||
from vllm.model_executor.layers.rotary_embedding import get_rope
|
||||
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import SupportsLoRA, SupportsPP
|
||||
from .llama import LlamaMLP as Glm4MLP
|
||||
|
||||
@@ -46,15 +46,14 @@ from transformers.models.glm4v.image_processing_glm4v import (
|
||||
from transformers.models.glm4v.video_processing_glm4v import Glm4vVideoProcessor
|
||||
from transformers.video_utils import VideoMetadata
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.attention.layers.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions, VideoDummyOptions
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size, parallel_state
|
||||
from vllm.distributed import utils as dist_utils
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer, Conv3dLayer
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
@@ -89,6 +88,7 @@ from vllm.multimodal.processing import (
|
||||
from vllm.multimodal.profiling import BaseDummyInputsBuilder
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from ..layers.activation import SiluAndMul
|
||||
from .interfaces import (
|
||||
|
||||
@@ -19,11 +19,11 @@ from transformers import BatchFeature, PreTrainedTokenizer, TensorType
|
||||
from transformers.image_utils import ImageInput
|
||||
from transformers.tokenization_utils_base import TextInput
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import SiluAndMul, get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -11,12 +11,12 @@ from transformers import BatchFeature
|
||||
from transformers.models.glmasr import GlmAsrConfig, GlmAsrProcessor
|
||||
from transformers.models.whisper import WhisperFeatureExtractor
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import ModelConfig, SpeechToTextConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed.parallel_state import get_tensor_model_parallel_world_size
|
||||
from vllm.inputs.data import PromptType
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
QKVParallelLinear,
|
||||
|
||||
@@ -7,7 +7,6 @@ import torch.distributed as dist
|
||||
from torch import nn
|
||||
from transformers import GptOssConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
@@ -37,6 +36,7 @@ from vllm.model_executor.models.utils import sequence_parallel_chunk
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.math_utils import cdiv
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import SupportsEagle3, SupportsLoRA, SupportsPP
|
||||
from .utils import (
|
||||
|
||||
@@ -33,7 +33,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig, get_current_vllm_config
|
||||
@@ -65,6 +64,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
maybe_remap_kv_scale_name,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import MixtureOfExperts, SupportsLoRA, SupportsPP
|
||||
from .utils import (
|
||||
|
||||
@@ -33,14 +33,13 @@ import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
from transformers import BatchFeature
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import parallel_state
|
||||
from vllm.distributed import utils as dist_utils
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
@@ -81,6 +80,7 @@ from vllm.transformers_utils.configs.hunyuan_vl import (
|
||||
from vllm.transformers_utils.processors.hunyuan_vl import HunYuanVLProcessor
|
||||
from vllm.transformers_utils.processors.hunyuan_vl_image import smart_resize
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .interfaces import (
|
||||
MultiModalEmbeddings,
|
||||
|
||||
@@ -27,9 +27,9 @@ from transformers.models.idefics2.configuration_idefics2 import (
|
||||
Idefics2VisionConfig,
|
||||
)
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -15,7 +15,6 @@ import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.distributed import (
|
||||
divide,
|
||||
get_tensor_model_parallel_rank,
|
||||
@@ -24,6 +23,7 @@ from vllm.distributed import (
|
||||
tensor_model_parallel_all_gather,
|
||||
)
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
|
||||
@@ -14,8 +14,8 @@ import torch.nn as nn
|
||||
from transformers import PretrainedConfig
|
||||
from transformers.utils import torch_int
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import ColumnParallelLinear, RowParallelLinear
|
||||
|
||||
@@ -24,7 +24,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
@@ -48,6 +47,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
)
|
||||
from vllm.model_executor.models.llama import LlamaMLP
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .utils import (
|
||||
AutoWeightsLoader,
|
||||
|
||||
@@ -16,11 +16,11 @@ from transformers.image_processing_utils import BatchFeature
|
||||
from transformers.tokenization_utils import TensorType
|
||||
from typing_extensions import TypedDict, Unpack
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.model import ModelConfig
|
||||
from vllm.distributed import parallel_state
|
||||
from vllm.distributed import utils as dist_utils
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
QKVParallelLinear,
|
||||
|
||||
@@ -16,13 +16,13 @@ from transformers.feature_extraction_utils import BatchFeature
|
||||
from transformers.modeling_outputs import BaseModelOutput, BaseModelOutputWithPooling
|
||||
from transformers.utils import torch_int
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -31,13 +31,14 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import LlamaConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
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_world_size
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
MergedColumnParallelLinear,
|
||||
@@ -56,6 +57,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
maybe_remap_kv_scale_name,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .adapters import as_embedding_model, as_seq_cls_model
|
||||
from .interfaces import (
|
||||
|
||||
@@ -25,7 +25,6 @@ from torch import nn
|
||||
from transformers import Llama4TextConfig
|
||||
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.chunked_local_attention import ChunkedLocalAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.distributed import (
|
||||
@@ -34,6 +33,9 @@ from vllm.distributed import (
|
||||
tensor_model_parallel_all_gather,
|
||||
)
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.attention.chunked_local_attention import (
|
||||
ChunkedLocalAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.fused_moe import SharedFusedMoE
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
|
||||
@@ -6,7 +6,6 @@ from itertools import islice
|
||||
import torch
|
||||
from torch import nn
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.config import (
|
||||
CacheConfig,
|
||||
@@ -43,6 +42,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
)
|
||||
from vllm.model_executor.models.utils import sequence_parallel_chunk
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import MixtureOfExperts, SupportsPP
|
||||
from .utils import (
|
||||
|
||||
@@ -14,7 +14,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import MiniMaxConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionMetadata
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, ModelConfig, VllmConfig
|
||||
@@ -48,6 +47,7 @@ from vllm.model_executor.layers.vocab_parallel_embedding import (
|
||||
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
|
||||
from vllm.model_executor.models.utils import maybe_prefix
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionMetadata
|
||||
|
||||
from .interfaces import HasInnerState, IsHybrid
|
||||
from .utils import PPMissingLayer, is_pp_missing_parameter, make_layers
|
||||
|
||||
@@ -31,10 +31,10 @@ from transformers.models.llama4.image_processing_llama4_fast import (
|
||||
get_best_fit,
|
||||
)
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.fused_moe import FusedMoE
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -7,10 +7,12 @@ from torch import nn
|
||||
from transformers import ModernBertConfig
|
||||
from transformers.activations import ACT2FN
|
||||
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.linear import QKVParallelLinear, RowParallelLinear
|
||||
from vllm.model_executor.layers.pooler import DispatchPooler
|
||||
from vllm.model_executor.layers.pooler.seqwise import (
|
||||
|
||||
@@ -18,7 +18,6 @@ from transformers.image_utils import ImageInput
|
||||
from transformers.tokenization_utils_base import TextInput
|
||||
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
@@ -30,6 +29,7 @@ from vllm.distributed import (
|
||||
tensor_model_parallel_all_gather,
|
||||
)
|
||||
from vllm.model_executor.layers.activation import MulAndSilu, QuickGELU, SiluAndMul
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -52,9 +52,9 @@ import torch.nn.functional as F
|
||||
from transformers.activations import ACT2FN
|
||||
from transformers.modeling_utils import PreTrainedModel
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import MultiModalConfig
|
||||
from vllm.distributed import divide, get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -31,7 +31,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import LlamaConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.distributed import get_pp_group
|
||||
@@ -49,6 +48,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
)
|
||||
from vllm.model_executor.models.llama import LlamaAttention, LlamaMLP
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import HasNoOps, SupportsLoRA, SupportsPP
|
||||
from .utils import (
|
||||
|
||||
@@ -26,7 +26,6 @@ from transformers import (
|
||||
TensorType,
|
||||
)
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.config.lora import LoRAConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
@@ -63,6 +62,7 @@ from vllm.multimodal.profiling import BaseDummyInputsBuilder
|
||||
from vllm.transformers_utils.configs.radio import RadioConfig
|
||||
from vllm.transformers_utils.tokenizer import AnyTokenizer
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
logger = init_logger(__name__)
|
||||
DEFAULT_FINAL_IMAGE_SIZE = (2048, 1648)
|
||||
|
||||
@@ -30,7 +30,6 @@ from torch import nn
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.layer import Attention, AttentionType
|
||||
from vllm.attention.layers.static_sink_attention import StaticSinkAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, ParallelConfig, VllmConfig
|
||||
from vllm.distributed import (
|
||||
@@ -42,6 +41,9 @@ from vllm.distributed import (
|
||||
tensor_model_parallel_all_gather,
|
||||
)
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.attention.static_sink_attention import (
|
||||
StaticSinkAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.fused_moe import SharedFusedMoE
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
|
||||
@@ -33,7 +33,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
@@ -57,6 +56,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
maybe_remap_kv_scale_name,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import SupportsLoRA
|
||||
from .utils import (
|
||||
|
||||
@@ -30,14 +30,13 @@ from transformers.modeling_outputs import (
|
||||
)
|
||||
from transformers.utils import torch_int
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.attention.layers.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import parallel_state
|
||||
from vllm.distributed import utils as dist_utils
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import (
|
||||
MMEncoderAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
QKVParallelLinear,
|
||||
@@ -72,6 +71,7 @@ from vllm.multimodal.processing import (
|
||||
from vllm.multimodal.profiling import BaseDummyInputsBuilder
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .ernie45 import Ernie4_5ForCausalLM
|
||||
from .interfaces import MultiModalEmbeddings, SupportsMRoPE, SupportsMultiModal
|
||||
|
||||
@@ -9,7 +9,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionMetadata
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import VllmConfig, get_current_vllm_config
|
||||
@@ -66,6 +65,7 @@ from vllm.model_executor.models.utils import (
|
||||
from vllm.model_executor.utils import set_weight_attrs
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.torch_utils import direct_register_custom_op
|
||||
from vllm.v1.attention.backend import AttentionMetadata
|
||||
from vllm.v1.attention.backends.mamba2_attn import Mamba2AttentionMetadata
|
||||
|
||||
|
||||
|
||||
@@ -33,13 +33,14 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import Qwen2Config
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
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_world_size
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
MergedColumnParallelLinear,
|
||||
@@ -59,6 +60,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.transformers_utils.config import is_interleaved, set_default_rope_theta
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import SupportsEagle3, SupportsLoRA, SupportsPP
|
||||
from .utils import (
|
||||
|
||||
@@ -41,8 +41,6 @@ from transformers.models.qwen2_5_vl.configuration_qwen2_5_vl import (
|
||||
Qwen2_5_VLVisionConfig,
|
||||
)
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.distributed import parallel_state
|
||||
@@ -50,6 +48,7 @@ from vllm.distributed import utils as dist_utils
|
||||
from vllm.forward_context import set_forward_context
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import get_act_and_mul_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv3dLayer
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (
|
||||
@@ -83,6 +82,7 @@ from vllm.multimodal.processing import PromptReplacement, PromptUpdate
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.platform_utils import is_pin_memory_available
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .interfaces import (
|
||||
MultiModalEmbeddings,
|
||||
|
||||
@@ -43,14 +43,13 @@ from transformers.models.qwen2_vl.configuration_qwen2_vl import (
|
||||
from transformers.models.qwen2_vl.image_processing_qwen2_vl import smart_resize
|
||||
from transformers.models.qwen2_vl.video_processing_qwen2_vl import Qwen2VLVideoProcessor
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import parallel_state, tensor_model_parallel_all_gather
|
||||
from vllm.distributed import utils as dist_utils
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import QuickGELU
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv3dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
@@ -90,6 +89,7 @@ from vllm.multimodal.profiling import BaseDummyInputsBuilder
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.tokenizers import TokenizerLike
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .interfaces import (
|
||||
MultiModalEmbeddings,
|
||||
|
||||
@@ -30,7 +30,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import Qwen3Config
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
@@ -44,6 +43,7 @@ from vllm.model_executor.layers.rotary_embedding import get_rope
|
||||
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.transformers_utils.config import set_default_rope_theta
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import SupportsEagle3, SupportsLoRA, SupportsPP
|
||||
from .qwen2 import Qwen2MLP as Qwen3MLP
|
||||
|
||||
@@ -10,7 +10,6 @@ from einops import rearrange
|
||||
from torch import nn
|
||||
from transformers.activations import ACT2FN
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionMetadata
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import (
|
||||
@@ -75,6 +74,7 @@ from vllm.sequence import IntermediateTensors
|
||||
from vllm.transformers_utils.configs import Qwen3NextConfig
|
||||
from vllm.triton_utils import tl, triton
|
||||
from vllm.utils.torch_utils import direct_register_custom_op
|
||||
from vllm.v1.attention.backend import AttentionMetadata
|
||||
from vllm.v1.attention.backends.gdn_attn import GDNAttentionMetadata
|
||||
|
||||
from .interfaces import (
|
||||
|
||||
@@ -46,7 +46,6 @@ from transformers.models.qwen3_omni_moe.processing_qwen3_omni_moe import (
|
||||
)
|
||||
from transformers.models.whisper import WhisperFeatureExtractor
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.distributed import get_pp_group
|
||||
@@ -75,6 +74,7 @@ from vllm.multimodal.processing import (
|
||||
PromptUpdateDetails,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .interfaces import (
|
||||
MultiModalEmbeddings,
|
||||
|
||||
@@ -48,7 +48,6 @@ from transformers.models.qwen3_vl.video_processing_qwen3_vl import (
|
||||
)
|
||||
from transformers.video_utils import VideoMetadata
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import MultiModalConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions, VideoDummyOptions
|
||||
@@ -92,6 +91,7 @@ from vllm.multimodal.processing import (
|
||||
from vllm.multimodal.profiling import BaseDummyInputsBuilder
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils.collection_utils import is_list_of
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
from .interfaces import (
|
||||
MultiModalEmbeddings,
|
||||
|
||||
@@ -30,7 +30,6 @@ import torch
|
||||
from torch import nn
|
||||
from transformers import PretrainedConfig as SeedOssConfig
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
@@ -56,6 +55,7 @@ from vllm.model_executor.model_loader.weight_utils import (
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.transformers_utils.config import set_default_rope_theta
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
from .interfaces import SupportsLoRA, SupportsPP
|
||||
from .utils import (
|
||||
|
||||
@@ -15,12 +15,14 @@ from transformers import (
|
||||
SiglipVisionConfig,
|
||||
)
|
||||
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions, MultiModalConfig
|
||||
from vllm.distributed import divide, get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -10,11 +10,11 @@ from torch import nn
|
||||
from torch.nn import functional as F
|
||||
from transformers import Siglip2VisionConfig
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import MultiModalConfig
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
QKVParallelLinear,
|
||||
|
||||
@@ -11,10 +11,10 @@ from torch.nn import functional as F
|
||||
from transformers import Siglip2VisionConfig
|
||||
from transformers.configuration_utils import PretrainedConfig
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import MultiModalConfig
|
||||
from vllm.distributed import divide, get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -15,11 +15,11 @@ from torchvision import transforms
|
||||
from torchvision.transforms.functional import InterpolationMode
|
||||
from transformers import BatchFeature, PretrainedConfig, TensorType
|
||||
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.conv import Conv2dLayer
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
|
||||
@@ -27,13 +27,14 @@ from torch import nn
|
||||
from transformers import AutoModel
|
||||
from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS
|
||||
|
||||
from vllm.attention.backends.abstract import AttentionType
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.encoder_only_attention import EncoderOnlyAttention
|
||||
from vllm.config.utils import getattr_iter
|
||||
from vllm.distributed import get_pp_group, get_tp_group
|
||||
from vllm.distributed.utils import get_pp_indices
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.attention.encoder_only_attention import (
|
||||
EncoderOnlyAttention,
|
||||
)
|
||||
from vllm.model_executor.layers.vocab_parallel_embedding import VocabParallelEmbedding
|
||||
from vllm.model_executor.models.interfaces import (
|
||||
SupportsEagle,
|
||||
@@ -59,6 +60,7 @@ from vllm.model_executor.models.utils import (
|
||||
maybe_prefix,
|
||||
)
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.v1.attention.backend import AttentionType
|
||||
|
||||
if TYPE_CHECKING:
|
||||
from transformers import PreTrainedModel
|
||||
|
||||
@@ -10,7 +10,6 @@ from typing import Final, Generic, Literal, Protocol, TypeAlias, TypeVar
|
||||
import torch
|
||||
from transformers import PretrainedConfig
|
||||
|
||||
from vllm.attention.backends.registry import AttentionBackendEnum
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.distributed import (
|
||||
get_tensor_model_parallel_rank,
|
||||
@@ -19,6 +18,7 @@ from vllm.distributed import (
|
||||
)
|
||||
from vllm.logger import init_logger
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.v1.attention.backends.registry import AttentionBackendEnum
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
@@ -18,18 +18,15 @@ from transformers import (
|
||||
)
|
||||
from transformers.models.whisper.modeling_whisper import sinusoids
|
||||
|
||||
from vllm.attention.backends.abstract import (
|
||||
AttentionType,
|
||||
)
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.layers.cross_attention import CrossAttention
|
||||
from vllm.attention.layers.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.config import CacheConfig, ModelConfig, SpeechToTextConfig, VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import get_tensor_model_parallel_world_size
|
||||
from vllm.inputs.data import PromptType
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import get_act_fn
|
||||
from vllm.model_executor.layers.attention.cross_attention import CrossAttention
|
||||
from vllm.model_executor.layers.attention.mm_encoder_attention import MMEncoderAttention
|
||||
from vllm.model_executor.layers.linear import (
|
||||
ColumnParallelLinear,
|
||||
QKVParallelLinear,
|
||||
@@ -62,6 +59,9 @@ from vllm.transformers_utils.processor import cached_processor_from_config
|
||||
from vllm.utils.jsontree import json_map_leaves
|
||||
from vllm.utils.tensor_schema import TensorSchema, TensorShape
|
||||
from vllm.utils.torch_utils import set_default_torch_dtype
|
||||
from vllm.v1.attention.backend import (
|
||||
AttentionType,
|
||||
)
|
||||
|
||||
from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsTranscription
|
||||
from .utils import (
|
||||
|
||||
@@ -9,20 +9,20 @@ import torch
|
||||
import torch.nn.functional as F
|
||||
from torch import nn
|
||||
|
||||
from vllm.attention.backends.abstract import (
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.model_executor.layers.quantization import QuantizationConfig
|
||||
from vllm.v1.attention.backend import (
|
||||
AttentionBackend,
|
||||
AttentionMetadata,
|
||||
AttentionType,
|
||||
)
|
||||
from vllm.attention.layer import Attention
|
||||
from vllm.attention.selector import get_attn_backend
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.model_executor.layers.quantization import QuantizationConfig
|
||||
from vllm.v1.attention.backends.flash_attn import FlashAttentionBackend
|
||||
from vllm.v1.attention.backends.utils import (
|
||||
CommonAttentionMetadata,
|
||||
subclass_attention_backend_with_overrides,
|
||||
)
|
||||
from vllm.v1.attention.selector import get_attn_backend
|
||||
from vllm.v1.kv_cache_interface import AttentionSpec
|
||||
|
||||
# From https://platform.openai.com/docs/guides/speech-to-text/supported-languages
|
||||
|
||||
Reference in New Issue
Block a user