已合并
internvl自动化用例服务拉起失败bug fix #1085
Snoopy99创建于 6月30日
internvl自动化用例服务拉起失败bug fix #1085
已合并
Snoopy99创建于 6月30日
3 个文件变更+17-3
@@ -10,7 +10,7 @@
10from atb_llm.models.base.graph_manager.graph_wrapper import ATBGraphWrapper10from atb_llm.models.base.graph_manager.graph_wrapper import ATBGraphWrapper
11from atb_llm.models.base.graph_manager.compatible_matrix import FeatureType11from atb_llm.models.base.graph_manager.compatible_matrix import FeatureType
12from atb_llm.models.base.flash_causal_lm import FlashForCausalLM12from atb_llm.models.base.flash_causal_lm import FlashForCausalLM
13-from atb_llm.runner.model_runner import generate_mem_pool_event_key13+from atb_llm.utils.mempool_utils import generate_mem_pool_event_key
14from mindie_llm.text_generator.plugins.plugin_manager import MemPoolType14from mindie_llm.text_generator.plugins.plugin_manager import MemPoolType
15 15 
16 16 
@@ -10,11 +10,12 @@
10from atb_llm.models.base.graph_manager.graph_wrapper import ATBGraphWrapper10from atb_llm.models.base.graph_manager.graph_wrapper import ATBGraphWrapper
11from atb_llm.models.base.graph_manager.compatible_matrix import FeatureType11from atb_llm.models.base.graph_manager.compatible_matrix import FeatureType
12from atb_llm.models.base.flash_causal_lm import FlashForCausalLM12from atb_llm.models.base.flash_causal_lm import FlashForCausalLM
13-from atb_llm.runner.model_runner import generate_mem_pool_event_key13+from atb_llm.utils.mempool_utils import generate_mem_pool_event_key
14 14 
15 15 
16class SplitFuseGraphWrapper(ATBGraphWrapper):16class SplitFuseGraphWrapper(ATBGraphWrapper):
17 "ATBGraphWrapper class for prefixcache and splitfuse"17 "ATBGraphWrapper class for prefixcache and splitfuse"
18+ 
18 pipe_key: str = generate_mem_pool_event_key(only_save_kv=False)19 pipe_key: str = generate_mem_pool_event_key(only_save_kv=False)
19 20 
20 def __init__(self):21 def __init__(self):
@@ -24,7 +25,7 @@ class SplitFuseGraphWrapper(ATBGraphWrapper):
24 25 
25 def activate(self, context: FlashForCausalLM, runtime_params, **kwargs) -> bool:26 def activate(self, context: FlashForCausalLM, runtime_params, **kwargs) -> bool:
26 pa_enable = False if context.inference_mode is None else context.inference_mode.enable_prefill_pa27 pa_enable = False if context.inference_mode is None else context.inference_mode.enable_prefill_pa
27- q_lens = "\"qLen\"" in runtime_params28+ q_lens = '"qLen"' in runtime_params
28 is_prefill = kwargs.get("is_prefill", False)29 is_prefill = kwargs.get("is_prefill", False)
29 if q_lens and is_prefill and pa_enable:30 if q_lens and is_prefill and pa_enable:
30 return True31 return True
@@ -0,0 +1,13 @@
1+# Copyright (c) Huawei Technologies Co., Ltd. 2024. All rights reserved.
2+# MindIE is licensed under Mulan PSL v2.
3+# You can use this software according to the terms and conditions of the Mulan PSL v2.
4+# You may obtain a copy of Mulan PSL v2 at:
5+# http://license.coscl.org.cn/MulanPSL2
6+# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
7+# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
8+# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
9+# See the Mulan PSL v2 for more details.
10+ 
11+ 
12+def generate_mem_pool_event_key(only_save_kv: bool) -> str:
13+ return "only_save_kv" if only_save_kv else "both_save_kv"