apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-22 16:07:51 +08:00
Code Issues Actions 15 Packages Projects Releases Wiki Activity

[BugFix][XPU] Fix kv_cache management bug (#7420)

Browse Source
This commit is contained in:
ddchenhao66
2026-04-16 15:45:45 +08:00
committed by GitHub
parent 6e16438a57
commit e9527208d9
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/worker/xpu_model_runner.py
+5 -1
View File
@@ -1259,7 +1259,11 @@ class XPUModelRunner(ModelRunnerBase):
# Check if gpu runner needs to create kv cache
# 1. During profiling, it creates its own kv cache.
# 2. GPU runner creates kv cache tensor unless p/d disaggregation is enabled.
create_cache_tensor = profile or self.scheduler_config.splitwise_role == "mixed"
create_cache_tensor = profile or not (
self.fd_config.cache_config.num_cpu_blocks > 0
or self.fd_config.cache_config.kvcache_storage_backend
or self.fd_config.scheduler_config.splitwise_role != "mixed"
)
if not create_cache_tensor:
logger.info(f"Waiting for cache managers to create kv cache.. {cache_ready_signal.value}")
while cache_ready_signal.value[local_rank] != 1: