feat: remove vllm distributed (#2907)

Co-authored-by: Zhangyi <1109276519@qq.com>

python/sglang/srt/layers/activation.py

@@ -25,13 +25,13 @@ from sglang.srt.utils import is_flashinfer_available
 if is_flashinfer_available():
     from flashinfer.activation import gelu_and_mul, gelu_tanh_and_mul, silu_and_mul
 
-from vllm.distributed import (
+from vllm.model_executor.custom_op import CustomOp
+
+from sglang.srt.distributed import (
     divide,
     get_tensor_model_parallel_rank,
     get_tensor_model_parallel_world_size,
 )
-from vllm.model_executor.custom_op import CustomOp
-
 from sglang.srt.layers.custom_op_util import register_custom_op
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
 from sglang.srt.utils import set_weight_attrs

python/sglang/srt/layers/dp_attention.py

@@ -1,5 +1,6 @@
 import torch
-from vllm.distributed import GroupCoordinator, get_tp_group
+
+from sglang.srt.distributed import GroupCoordinator, get_tp_group
 
 _ATTN_TP_GROUP = None
 _ATTN_TP_RANK = None

python/sglang/srt/layers/linear.py

@@ -7,7 +7,8 @@ from typing import Dict, List, Optional, Tuple
 import torch
 import torch.nn.functional as F
 from torch.nn.parameter import Parameter, UninitializedParameter
-from vllm.distributed import (
+
+from sglang.srt.distributed import (
     divide,
     get_tensor_model_parallel_rank,
     get_tensor_model_parallel_world_size,
@@ -15,7 +16,6 @@ from vllm.distributed import (
     tensor_model_parallel_all_gather,
     tensor_model_parallel_all_reduce,
 )
-
 from sglang.srt.layers.parameter import (
     BasevLLMParameter,
     PackedColumnParameter,

python/sglang/srt/layers/logits_processor.py

@@ -20,11 +20,11 @@ import torch
 import triton
 import triton.language as tl
 from torch import nn
-from vllm.distributed import (
+
+from sglang.srt.distributed import (
     get_tensor_model_parallel_world_size,
     tensor_model_parallel_all_gather,
 )
-
 from sglang.srt.layers.vocab_parallel_embedding import VocabParallelEmbedding
 from sglang.srt.model_executor.forward_batch_info import (
     CaptureHiddenMode,

python/sglang/srt/layers/moe/ep_moe/layer.py

@@ -4,13 +4,13 @@ from typing import Callable, List, Optional, Tuple
 import torch
 from torch.nn import Module
 from vllm import _custom_ops as ops
-from vllm.distributed import (
-    get_tensor_model_parallel_rank,
-    get_tensor_model_parallel_world_size,
-)
 from vllm.model_executor.custom_op import CustomOp
 from vllm.model_executor.layers.quantization.fp8 import Fp8Config, Fp8MoEMethod
 
+from sglang.srt.distributed import (
+    get_tensor_model_parallel_rank,
+    get_tensor_model_parallel_world_size,
+)
 from sglang.srt.layers.custom_op_util import register_custom_op
 from sglang.srt.layers.moe.ep_moe.kernels import (
     grouped_gemm_triton,

python/sglang/srt/layers/moe/fused_moe_triton/layer.py

@@ -5,13 +5,13 @@ from enum import Enum
 from typing import Callable, List, Optional, Tuple
 
 import torch
-from vllm.distributed import (
+from vllm.model_executor.custom_op import CustomOp
+
+from sglang.srt.distributed import (
     get_tensor_model_parallel_rank,
     get_tensor_model_parallel_world_size,
     tensor_model_parallel_all_reduce,
 )
-from vllm.model_executor.custom_op import CustomOp
-
 from sglang.srt.layers.custom_op_util import register_custom_op
 from sglang.srt.layers.moe.fused_moe_native import moe_forward_native
 from sglang.srt.layers.moe.topk import select_experts

python/sglang/srt/layers/parameter.py

@@ -6,7 +6,8 @@ from typing import Callable, Optional, Union
 
 import torch
 from torch.nn import Parameter
-from vllm.distributed import get_tensor_model_parallel_rank
+
+from sglang.srt.distributed import get_tensor_model_parallel_rank
 
 __all__ = [
     "BasevLLMParameter",

python/sglang/srt/layers/quantization/fp8.py

@@ -8,7 +8,6 @@ import torch.nn.functional as F
 from torch.nn import Module
 from torch.nn.parameter import Parameter
 from vllm import _custom_ops as ops
-from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.quantization.kv_cache import BaseKVCacheMethod
 from vllm.model_executor.layers.quantization.utils.marlin_utils_fp8 import (
     apply_fp8_marlin_linear,
@@ -24,6 +23,7 @@ from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
     requantize_with_max_scale,
 )
 
+from sglang.srt.distributed import get_tensor_model_parallel_world_size
 from sglang.srt.layers.linear import (
     LinearBase,
     LinearMethodBase,

python/sglang/srt/layers/vocab_parallel_embedding.py

@@ -6,13 +6,13 @@ from typing import List, Optional, Sequence, Tuple
 import torch
 import torch.nn.functional as F
 from torch.nn.parameter import Parameter, UninitializedParameter
-from vllm.distributed import (
+
+from sglang.srt.distributed import (
     divide,
     get_tensor_model_parallel_rank,
     get_tensor_model_parallel_world_size,
     tensor_model_parallel_all_reduce,
 )
-
 from sglang.srt.layers.parameter import BasevLLMParameter
 from sglang.srt.layers.quantization.base_config import (
     QuantizationConfig,