Organize code (rename, movement) (#953)

python/sglang/srt/models/llava.py

@@ -32,13 +32,12 @@ from vllm.config import CacheConfig
 from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
-from sglang.srt.managers.schedule_batch import ForwardMode
 from sglang.srt.mm_utils import (
     get_anyres_image_grid_shape,
     unpad_image,
     unpad_image_shape,
 )
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import ForwardMode, InputMetadata
 from sglang.srt.models.llama2 import LlamaForCausalLM
 from sglang.srt.models.mistral import MistralForCausalLM
 from sglang.srt.models.qwen2 import Qwen2ForCausalLM

python/sglang/srt/models/llavavid.py

@@ -26,13 +26,12 @@ from vllm.config import CacheConfig
 from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
-from sglang.srt.managers.schedule_batch import ForwardMode
 from sglang.srt.mm_utils import (
     get_anyres_image_grid_shape,
     unpad_image,
     unpad_image_shape,
 )
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import ForwardMode, InputMetadata
 from sglang.srt.models.llama2 import LlamaForCausalLM
 
 

python/sglang/srt/models/minicpm.py

@@ -39,7 +39,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import InputMetadata
 
 
 class MiniCPMMLP(nn.Module):

python/sglang/srt/models/mixtral.py

@@ -50,7 +50,7 @@ from vllm.utils import print_warning_once
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import InputMetadata
 
 
 class MixtralMoE(nn.Module):

python/sglang/srt/models/mixtral_quant.py

@@ -45,7 +45,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import InputMetadata
 
 
 class MixtralMLP(nn.Module):

python/sglang/srt/models/qwen.py

@@ -39,7 +39,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import InputMetadata
 
 
 class QWenMLP(nn.Module):

python/sglang/srt/models/qwen2.py

@@ -39,7 +39,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import InputMetadata
 
 Qwen2Config = None
 

python/sglang/srt/models/qwen2_moe.py

@@ -51,7 +51,7 @@ from vllm.sequence import IntermediateTensors, SamplerOutput
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import InputMetadata
 
 
 class Qwen2MoeMLP(nn.Module):

python/sglang/srt/models/stablelm.py

@@ -40,7 +40,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
-from sglang.srt.model_executor.model_runner import InputMetadata
+from sglang.srt.model_executor.forward_batch_info import InputMetadata
 
 
 class StablelmMLP(nn.Module):