[Misc][Refactor] Introduce ExecuteModelData (#4540)
This commit is contained in:
@@ -3,7 +3,7 @@ from typing import Any, Dict, List, Optional, Set, Tuple
|
||||
from vllm.executor.executor_base import ExecutorAsyncBase, ExecutorBase
|
||||
from vllm.logger import init_logger
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.sequence import SamplerOutput, SequenceGroupMetadata
|
||||
from vllm.sequence import ExecuteModelRequest, SamplerOutput
|
||||
from vllm.utils import (get_distributed_init_method, get_ip, get_open_port,
|
||||
make_async)
|
||||
from vllm.worker.worker_base import WorkerWrapperBase
|
||||
@@ -117,20 +117,9 @@ class GPUExecutor(ExecutorBase):
|
||||
self.driver_worker.initialize_cache(num_gpu_blocks, num_cpu_blocks)
|
||||
|
||||
def execute_model(
|
||||
self,
|
||||
seq_group_metadata_list: List[SequenceGroupMetadata],
|
||||
blocks_to_swap_in: Dict[int, int],
|
||||
blocks_to_swap_out: Dict[int, int],
|
||||
blocks_to_copy: Dict[int, List[int]],
|
||||
num_lookahead_slots: int,
|
||||
) -> List[SamplerOutput]:
|
||||
output = self.driver_worker.execute_model(
|
||||
seq_group_metadata_list=seq_group_metadata_list,
|
||||
blocks_to_swap_in=blocks_to_swap_in,
|
||||
blocks_to_swap_out=blocks_to_swap_out,
|
||||
blocks_to_copy=blocks_to_copy,
|
||||
num_lookahead_slots=num_lookahead_slots,
|
||||
)
|
||||
self,
|
||||
execute_model_req: ExecuteModelRequest) -> List[SamplerOutput]:
|
||||
output = self.driver_worker.execute_model(execute_model_req)
|
||||
return output
|
||||
|
||||
def add_lora(self, lora_request: LoRARequest) -> bool:
|
||||
@@ -154,16 +143,8 @@ class GPUExecutorAsync(GPUExecutor, ExecutorAsyncBase):
|
||||
|
||||
async def execute_model_async(
|
||||
self,
|
||||
seq_group_metadata_list: List[SequenceGroupMetadata],
|
||||
blocks_to_swap_in: Dict[int, int],
|
||||
blocks_to_swap_out: Dict[int, int],
|
||||
blocks_to_copy: Dict[int, List[int]],
|
||||
num_lookahead_slots: int,
|
||||
execute_model_req: ExecuteModelRequest,
|
||||
) -> List[SamplerOutput]:
|
||||
output = await make_async(self.driver_worker.execute_model)(
|
||||
seq_group_metadata_list=seq_group_metadata_list,
|
||||
blocks_to_swap_in=blocks_to_swap_in,
|
||||
blocks_to_swap_out=blocks_to_swap_out,
|
||||
blocks_to_copy=blocks_to_copy,
|
||||
num_lookahead_slots=num_lookahead_slots)
|
||||
output = await make_async(self.driver_worker.execute_model
|
||||
)(execute_model_req=execute_model_req, )
|
||||
return output
|
||||
|
||||
Reference in New Issue
Block a user