apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 00:57:33 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[fix] initialize available_gpu_block_num with max_gpu_block_num (#4193)

Browse Source
This commit is contained in:
李泳桦
2025-09-22 18:56:00 +08:00
committed by GitHub
parent 01f6934162
commit 0358329946
1 changed files with 5 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
fastdeploy/cache_manager/prefix_cache_manager.py
View File

@@ -113,6 +113,10 @@ class PrefixCacheManager:
+ f"{self.num_cpu_blocks}, bytes_per_layer_per_block {self.cache_config.bytes_per_layer_per_block}"
)
main_process_metrics.max_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.available_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.available_gpu_resource.set(1.0)
@property
def available_gpu_resource(self):
return len(self.gpu_free_block_list) / self.num_gpu_blocks if self.num_gpu_blocks > 0 else 0.0
@@ -269,6 +273,7 @@ class PrefixCacheManager:
self.node_id_pool = list(range(self.num_gpu_blocks + self.num_cpu_blocks))
main_process_metrics.max_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.available_gpu_block_num.set(self.num_gpu_blocks)
main_process_metrics.available_gpu_resource.set(1.0)
def can_allocate_gpu_blocks(self, num_blocks: int):