[Feature][Perf] Support Selective CPU Weight Offloading (#34535)

Signed-off-by: wzhao18 <wzhao18.sz@gmail.com>
This commit is contained in:
Wei Zhao
2026-02-13 20:02:24 -08:00
committed by GitHub
parent a0638d052d
commit b37b679770
4 changed files with 44 additions and 2 deletions
+5 -1
View File
@@ -345,9 +345,13 @@ class GPUModelRunner(
self.speculative_config = vllm_config.speculative_config
self.observability_config = vllm_config.observability_config
from vllm.model_executor.models.utils import set_cpu_offload_max_bytes
from vllm.model_executor.models.utils import (
set_cpu_offload_max_bytes,
set_cpu_offload_params,
)
set_cpu_offload_max_bytes(int(self.cache_config.cpu_offload_gb * 1024**3))
set_cpu_offload_params(self.cache_config.cpu_offload_params)
model_config = self.model_config
cache_config = self.cache_config