Misc clean up; Remove the support of jump forward (#4032)

python/sglang/srt/layers/attention/double_sparsity_backend.py

@@ -4,7 +4,7 @@ from typing import TYPE_CHECKING
 
 import torch
 
-from sglang.srt.layers.attention import AttentionBackend
+from sglang.srt.layers.attention.base_attn_backend import AttentionBackend
 from sglang.srt.managers.schedule_batch import global_server_args_dict
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch
 

python/sglang/srt/layers/attention/flashinfer_backend.py

@@ -19,9 +19,8 @@ import triton
 import triton.language as tl
 
 from sglang.global_config import global_config
-from sglang.srt.layers.attention import AttentionBackend
+from sglang.srt.layers.attention.base_attn_backend import AttentionBackend
 from sglang.srt.layers.dp_attention import get_attention_tp_size
-from sglang.srt.managers.schedule_batch import global_server_args_dict
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch, ForwardMode
 from sglang.srt.utils import is_flashinfer_available
 

python/sglang/srt/layers/attention/flashinfer_mla_backend.py

@@ -15,7 +15,7 @@ from typing import TYPE_CHECKING, Optional, Union
 import torch
 
 from sglang.global_config import global_config
-from sglang.srt.layers.attention import AttentionBackend
+from sglang.srt.layers.attention.base_attn_backend import AttentionBackend
 from sglang.srt.layers.attention.flashinfer_backend import (
     create_flashinfer_kv_indices_triton,
 )
@@ -34,7 +34,6 @@ if is_flashinfer_available():
         BatchMLAPagedAttentionWrapper,
         BatchPrefillWithRaggedKVCacheWrapper,
     )
-    from flashinfer.cascade import merge_state
 
 
 @dataclass

python/sglang/srt/layers/attention/torch_native_backend.py

@@ -5,7 +5,7 @@ from typing import TYPE_CHECKING
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 
-from sglang.srt.layers.attention import AttentionBackend
+from sglang.srt.layers.attention.base_attn_backend import AttentionBackend
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch
 
 if TYPE_CHECKING:

python/sglang/srt/layers/attention/triton_backend.py

@@ -5,7 +5,7 @@ from typing import TYPE_CHECKING, Optional, Union
 import torch
 import triton
 
-from sglang.srt.layers.attention import AttentionBackend
+from sglang.srt.layers.attention.base_attn_backend import AttentionBackend
 from sglang.srt.layers.attention.flashinfer_backend import (
     create_flashinfer_kv_indices_triton,
 )