[VLM] Refactor MultiModalConfig initialization and profiling (#7530)

This commit is contained in:
Roger Wang
2024-08-17 13:30:55 -07:00
committed by GitHub
parent 1ef13cf92f
commit bbf55c4805
29 changed files with 143 additions and 190 deletions
+2 -5
View File
@@ -3,9 +3,8 @@ from typing import List, Optional
import torch
from vllm.config import (CacheConfig, DeviceConfig, LoadConfig, LoRAConfig,
ModelConfig, MultiModalConfig, ParallelConfig,
PromptAdapterConfig, SchedulerConfig,
SpeculativeConfig)
ModelConfig, ParallelConfig, PromptAdapterConfig,
SchedulerConfig, SpeculativeConfig)
from vllm.executor.executor_base import ExecutorAsyncBase
from vllm.executor.gpu_executor import GPUExecutor
from vllm.logger import init_logger
@@ -29,7 +28,6 @@ class XPUExecutor(GPUExecutor):
device_config: DeviceConfig,
load_config: LoadConfig,
lora_config: Optional[LoRAConfig],
multimodal_config: Optional[MultiModalConfig],
prompt_adapter_config: Optional[PromptAdapterConfig],
speculative_config: Optional[SpeculativeConfig],
) -> None:
@@ -46,7 +44,6 @@ class XPUExecutor(GPUExecutor):
self.parallel_config = parallel_config
self.scheduler_config = scheduler_config
self.device_config = device_config
self.multimodal_config = multimodal_config
self.prompt_adapter_config = prompt_adapter_config
self.speculative_config = None