biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[v1] Hybrid Memory Allocator (#17996)

Browse Source 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
This commit is contained in:
Chen Zhang
2025-06-06 11:47:09 +08:00
committed by GitHub
parent 3465b87ef8
commit f8a1a2d108
21 changed files with 1605 additions and 440 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/engine/arg_utils.py
View File

@@ -387,6 +387,9 @@ class EngineArgs:
bool] = SchedulerConfig.enable_chunked_prefill
disable_chunked_mm_input: bool = SchedulerConfig.disable_chunked_mm_input
disable_hybrid_kv_cache_manager: bool = (
SchedulerConfig.disable_hybrid_kv_cache_manager)
guided_decoding_backend: GuidedDecodingBackend = DecodingConfig.backend
guided_decoding_disable_fallback: bool = DecodingConfig.disable_fallback
guided_decoding_disable_any_whitespace: bool = \
@@ -849,6 +852,9 @@ class EngineArgs:
**scheduler_kwargs["disable_chunked_mm_input"])
scheduler_group.add_argument("--scheduler-cls",
**scheduler_kwargs["scheduler_cls"])
scheduler_group.add_argument(
"--disable-hybrid-kv-cache-manager",
**scheduler_kwargs["disable_hybrid_kv_cache_manager"])
# vLLM arguments
vllm_kwargs = get_kwargs(VllmConfig)
@@ -1174,6 +1180,8 @@ class EngineArgs:
max_num_partial_prefills=self.max_num_partial_prefills,
max_long_partial_prefills=self.max_long_partial_prefills,
long_prefill_token_threshold=self.long_prefill_token_threshold,
disable_hybrid_kv_cache_manager=self.
disable_hybrid_kv_cache_manager,
)
lora_config = LoRAConfig(