feat(core): APP use new SDK component (#1050)

dbgpt/util/benchmarks/llm/fastchat_benchmarks_inference.py

@@ -4,9 +4,10 @@ For benchmarks.
 
 """
 import gc
-from typing import Iterable, Dict
+from typing import Dict, Iterable
 
 import torch
+from fastchat.utils import get_context_length, is_partial_stop, is_sentence_complete
 from transformers.generation.logits_process import (
     LogitsProcessorList,
     RepetitionPenaltyLogitsProcessor,
@@ -16,9 +17,6 @@ from transformers.generation.logits_process import (
 )
 
 
-from fastchat.utils import is_partial_stop, is_sentence_complete, get_context_length
-
-
 def prepare_logits_processor(
     temperature: float, repetition_penalty: float, top_p: float, top_k: int
 ) -> LogitsProcessorList:

dbgpt/util/benchmarks/llm/llm_benchmarks.py

@@ -1,24 +1,22 @@
-from typing import Dict, List
+import argparse
 import asyncio
+import csv
+import logging
 import os
 import sys
 import time
-import csv
-import argparse
-import logging
 import traceback
-from dbgpt.configs.model_config import ROOT_PATH, LLM_MODEL_CONFIG
 from datetime import datetime
+from typing import Dict, List
 
-from dbgpt.model.cluster.worker.manager import (
-    run_worker_manager,
-    initialize_worker_manager_in_client,
-    WorkerManager,
-)
-
-from dbgpt.core import ModelOutput, ModelInferenceMetrics
+from dbgpt.configs.model_config import LLM_MODEL_CONFIG, ROOT_PATH
+from dbgpt.core import ModelInferenceMetrics, ModelOutput
 from dbgpt.core.interface.message import ModelMessage, ModelMessageRoleType
-
+from dbgpt.model.cluster.worker.manager import (
+    WorkerManager,
+    initialize_worker_manager_in_client,
+    run_worker_manager,
+)
 
 model_name = "vicuna-7b-v1.5"
 model_path = LLM_MODEL_CONFIG[model_name]