Tiny move files to utils folder (#11166)

python/sglang/srt/model_executor/model_runner.py

@@ -29,7 +29,6 @@ from typing import List, Optional, Tuple, Union
 import torch
 import torch.distributed as dist
 
-from sglang.srt import slow_rank_detector
 from sglang.srt.configs.device_config import DeviceConfig
 from sglang.srt.configs.load_config import LoadConfig, LoadFormat
 from sglang.srt.configs.model_config import AttentionArch, ModelConfig
@@ -115,7 +114,6 @@ from sglang.srt.offloader import (
     get_offloader,
     set_offloader,
 )
-from sglang.srt.patch_torch import monkey_patch_torch_reductions
 from sglang.srt.sampling.sampling_batch_info import SamplingBatchInfo
 from sglang.srt.server_args import ServerArgs
 from sglang.srt.speculative.spec_info import SpeculativeAlgorithm
@@ -140,7 +138,9 @@ from sglang.srt.utils import (
     monkey_patch_p2p_access_check,
     monkey_patch_vllm_gguf_config,
     set_cuda_arch,
+    slow_rank_detector,
 )
+from sglang.srt.utils.patch_torch import monkey_patch_torch_reductions
 from sglang.srt.weight_sync.tensor_bucket import (
     FlattenedTensorBucket,
     FlattenedTensorMetadata,

python/sglang/srt/models/dots_ocr.py

@@ -9,7 +9,6 @@ import torch.nn as nn
 from transformers.activations import ACT2FN
 
 from sglang.srt.configs import DotsOCRConfig
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
 from sglang.srt.layers.vocab_parallel_embedding import ParallelLMHead
@@ -23,6 +22,7 @@ from sglang.srt.model_loader.weight_utils import default_weight_loader
 from sglang.srt.models.dots_vlm_vit import DotsVisionTransformer
 from sglang.srt.models.qwen2 import Qwen2ForCausalLM
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/gemma3_mm.py

@@ -23,7 +23,6 @@ import torch
 from torch import nn
 from transformers import Gemma3Config, PreTrainedModel
 
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.layernorm import Gemma3RMSNorm
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
@@ -44,6 +43,7 @@ from sglang.srt.model_loader.weight_utils import (
 from sglang.srt.models.gemma3_causal import Gemma3ForCausalLM
 from sglang.srt.models.siglip import SiglipVisionModel
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/gemma3n_mm.py

@@ -14,7 +14,6 @@ from transformers import (
 )
 from transformers.models.auto.modeling_auto import AutoModel
 
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.layernorm import RMSNorm
 from sglang.srt.layers.linear import ColumnParallelLinear, RowParallelLinear
 from sglang.srt.layers.logits_processor import LogitsProcessor
@@ -38,6 +37,7 @@ from sglang.srt.model_loader.weight_utils import (
 from sglang.srt.models.gemma3n_audio import Gemma3nAudioEncoder
 from sglang.srt.models.gemma3n_causal import Gemma3nRMSNorm, Gemma3nTextModel
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/glm4v.py

@@ -7,7 +7,6 @@ import torch.nn as nn
 import torch.nn.functional as F
 from transformers.models.glm4v.configuration_glm4v import Glm4vConfig, Glm4vVisionConfig
 
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.activation import SiluAndMul
 from sglang.srt.layers.attention import vision_utils
 from sglang.srt.layers.layernorm import RMSNorm
@@ -28,6 +27,7 @@ from sglang.srt.models.qwen2_5_vl import (
     Qwen2_5_VLForConditionalGeneration,
 )
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/glm4v_moe.py

@@ -10,7 +10,6 @@ from sglang.srt.distributed import (
     get_moe_expert_parallel_world_size,
     get_tensor_model_parallel_world_size,
 )
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.attention import vision_utils
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.moe.fused_moe_triton import FusedMoE
@@ -22,6 +21,7 @@ from sglang.srt.model_loader.weight_utils import default_weight_loader
 from sglang.srt.models.glm4_moe import Glm4MoeModel
 from sglang.srt.models.glm4v import Glm4vForConditionalGeneration, Glm4vVisionModel
 from sglang.srt.utils import add_prefix, is_cuda, log_info_on_rank0
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 _is_cuda = is_cuda()
 

python/sglang/srt/models/qwen2_5_vl.py

@@ -40,7 +40,6 @@ from transformers.models.qwen2_5_vl.modeling_qwen2_5_vl import (
     Qwen2_5_VisionRotaryEmbedding,
 )
 
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.attention.vision import VisionAttention
 from sglang.srt.layers.layernorm import RMSNorm
 from sglang.srt.layers.linear import (
@@ -61,6 +60,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch
 from sglang.srt.model_loader.weight_utils import default_weight_loader
 from sglang.srt.models.qwen2 import Qwen2Model
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/qwen2_audio.py

@@ -39,7 +39,6 @@ from transformers.models.qwen2_audio.modeling_qwen2_audio import (
     Qwen2AudioMultiModalProjector,
 )
 
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.activation import QuickGELU
 from sglang.srt.layers.attention.vision import VisionAttention
 from sglang.srt.layers.linear import ColumnParallelLinear, RowParallelLinear
@@ -61,6 +60,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch
 from sglang.srt.model_loader.weight_utils import default_weight_loader
 from sglang.srt.models.qwen2 import Qwen2ForCausalLM
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/qwen2_vl.py

@@ -33,7 +33,6 @@ from einops import rearrange
 from transformers import Qwen2VLConfig
 from transformers.models.qwen2_vl.configuration_qwen2_vl import Qwen2VLVisionConfig
 
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.activation import QuickGELU
 from sglang.srt.layers.attention.vision import VisionAttention
 from sglang.srt.layers.linear import ColumnParallelLinear, RowParallelLinear
@@ -50,6 +49,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch
 from sglang.srt.model_loader.weight_utils import default_weight_loader
 from sglang.srt.models.qwen2 import Qwen2Model
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/qwen3_vl.py

@@ -28,7 +28,6 @@ from transformers.models.qwen2_5_vl.modeling_qwen2_5_vl import (
 )
 
 from sglang.srt.configs.qwen3_vl import Qwen3VLConfig, Qwen3VLVisionConfig
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.attention.vision import VisionAttention
 from sglang.srt.layers.linear import ColumnParallelLinear, RowParallelLinear
 from sglang.srt.layers.logits_processor import LogitsProcessor
@@ -45,6 +44,7 @@ from sglang.srt.model_loader.weight_utils import default_weight_loader
 from sglang.srt.models.qwen2_vl import Qwen2VLVideoInputs
 from sglang.srt.models.qwen3 import Qwen3Model
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/models/qwen3_vl_moe.py

@@ -34,7 +34,6 @@ from sglang.srt.distributed import (
     get_pp_group,
     get_tensor_model_parallel_rank,
 )
-from sglang.srt.hf_transformers_utils import get_processor
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE
 from sglang.srt.layers.pooler import Pooler, PoolingType
@@ -58,6 +57,7 @@ from sglang.srt.models.qwen3_vl import (
     Qwen3VLForConditionalGeneration,
 )
 from sglang.srt.utils import add_prefix
+from sglang.srt.utils.hf_transformers_utils import get_processor
 
 logger = logging.getLogger(__name__)
 

python/sglang/srt/server_args.py

@@ -24,7 +24,6 @@ from typing import List, Literal, Optional, Union
 
 from sglang.srt.connector import ConnectorType
 from sglang.srt.function_call.function_call_parser import FunctionCallParser
-from sglang.srt.hf_transformers_utils import check_gguf_file, get_config
 from sglang.srt.lora.lora_registry import LoRARef
 from sglang.srt.parser.reasoning_parser import ReasoningParser
 from sglang.srt.utils import (
@@ -47,6 +46,7 @@ from sglang.srt.utils import (
     nullable_str,
     parse_connector_type,
 )
+from sglang.srt.utils.hf_transformers_utils import check_gguf_file, get_config
 from sglang.utils import is_in_ci
 
 logger = logging.getLogger(__name__)

python/sglang/srt/utils/__init__.py

+# Temporarily do this to avoid changing all imports in the repo
+from .common import *

python/sglang/srt/weight_sync/utils.py

@@ -33,7 +33,7 @@ async def update_weights(
     """
     infer_tp_size = device_mesh[device_mesh_key].mesh.size()[0]
     infer_tp_rank = device_mesh[device_mesh_key].get_local_rank()
-    from sglang.srt.patch_torch import monkey_patch_torch_reductions
+    from sglang.srt.utils.patch_torch import monkey_patch_torch_reductions
 
     monkey_patch_torch_reductions()