[1/N] Remove CacheConfig import in all model files (#1658)

python/sglang/srt/models/deepseek.py

@@ -21,7 +21,6 @@ from typing import Any, Dict, Iterable, Optional, Tuple
 import torch
 from torch import nn
 from transformers import PretrainedConfig
-from vllm.config import CacheConfig
 from vllm.distributed import (
     get_tensor_model_parallel_rank,
     get_tensor_model_parallel_world_size,
@@ -185,7 +184,7 @@ class DeepseekAttention(nn.Module):
         rope_theta: float = 10000,
         rope_scaling: Optional[Dict[str, Any]] = None,
         max_position_embeddings: int = 8192,
-        cache_config: Optional[CacheConfig] = None,
+        cache_config=None,
         quant_config: Optional[QuantizationConfig] = None,
     ) -> None:
         super().__init__()
@@ -262,7 +261,7 @@ class DeepseekDecoderLayer(nn.Module):
         self,
         config: PretrainedConfig,
         layer_id: int,
-        cache_config: Optional[CacheConfig] = None,
+        cache_config=None,
         quant_config: Optional[QuantizationConfig] = None,
     ) -> None:
         super().__init__()
@@ -331,7 +330,7 @@ class DeepseekModel(nn.Module):
     def __init__(
         self,
         config: PretrainedConfig,
-        cache_config: Optional[CacheConfig] = None,
+        cache_config=None,
         quant_config: Optional[QuantizationConfig] = None,
     ) -> None:
         super().__init__()
@@ -374,7 +373,7 @@ class DeepseekForCausalLM(nn.Module):
     def __init__(
         self,
         config: PretrainedConfig,
-        cache_config: Optional[CacheConfig] = None,
+        cache_config=None,
         quant_config: Optional[QuantizationConfig] = None,
     ) -> None:
         super().__init__()