apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 17:11:21 +08:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

Write the cache of preempted req to storage (#7113)

Browse Source
This commit is contained in:
jc
2026-04-01 13:16:12 +08:00
committed by GitHub
parent 6d0d404a9b
commit bd48640b4b
5 changed files with 16 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/cache_manager/prefix_cache_manager.py
+7 -5
View File
@@ -872,7 +872,7 @@ class PrefixCacheManager:
read_storage_task = ReadStorageTask(
task_id=req_id,
keys=no_match_block_keys,
token_ids=input_token_ids,
token_ids=input_token_ids if self.kvcache_storage_backend == "attention_store" else None,
gpu_block_ids=gpu_recv_storage_block_ids,
start_read_block_idx=match_token_num // block_size,
)
@@ -1111,7 +1111,9 @@ class PrefixCacheManager:
if isinstance(token_ids, np.ndarray):
token_ids = token_ids.tolist()
if self.config.cache_config.enable_output_caching:
token_ids += request.output_token_ids
input_token_ids = token_ids + request.output_token_ids
else:
input_token_ids = token_ids
req_id = request.request_id
keys = []
@@ -1128,7 +1130,7 @@ class PrefixCacheManager:
write_storage_task = WriteStorageTask(
task_id=req_id,
keys=keys,
token_ids=token_ids,
token_ids=input_token_ids if self.kvcache_storage_backend == "attention_store" else None,
gpu_block_ids=gpu_block_ids,
)
logger.debug(f"issue write storage task: {write_storage_task}")
@@ -2067,7 +2069,7 @@ class PrefixCacheManager:
event_type = data[0]
if event_type.value == CacheStatus.STORAGE2GPU.value:
logger.info(f"recv_data_transfer_result: {data}")
logger.debug(f"recv_data_transfer_result: {data}")
task_id, hash_keys, block_ids = data[1:]
if task_id not in self.storage_prefetch_block_ids:
self.storage_prefetch_block_ids[task_id] = []
@@ -2078,7 +2080,7 @@ class PrefixCacheManager:
if task_id in self.task_prefetch_event:
self.task_prefetch_event[task_id].set()
elif event_type.value == CacheStatus.GPU2STORAGE.value:
logger.info(f"recv_data_transfer_result: {data}")
logger.debug(f"recv_data_transfer_result: {data}")
task_id, hash_keys, block_ids = data[1:]
if task_id in self.task_write_back_event:
self.task_write_back_event[task_id].set()