apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[Optimize] Optimize ttft for ep (#6098)

Browse Source 
* optimize ttft

* fix

* fix

* fix ci

* fix ci

* fix

* fix bug

* fix

* add comments

* fix ci

* fix
This commit is contained in:
chenjian
2026-02-04 15:03:29 +08:00
committed by GitHub
parent 6e96bd0bd2
commit 90db0bdd0d
10 changed files with 118 additions and 142 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/splitwise/internal_adapter_utils.py
+4 -1
View File
@@ -53,6 +53,9 @@ class InternalAdapter:
available_batch_size = min(self.cfg.max_prefill_batch, self.engine.resource_manager.available_batch())
available_block_num = self.engine.resource_manager.available_block_num()
unhandled_request_num = self.engine.scheduler.get_unhandled_request_num()
if envs.ENABLE_V1_KVCACHE_SCHEDULER:
unhandled_request_num = max(unhandled_request_num, len(self.engine.resource_manager.waiting))
server_info = {
"splitwise_role": self.cfg.scheduler_config.splitwise_role,
"block_size": int(self.cfg.cache_config.block_size),
@@ -62,7 +65,7 @@ class InternalAdapter:
"available_resource": float(1.0 * available_block_num / self.cfg.cache_config.total_block_num),
"max_batch_size": int(available_batch_size),
"max_input_token_num": self.cfg.model_config.max_model_len,
"unhandled_request_num": self.engine.scheduler.get_unhandled_request_num(),
"unhandled_request_num": unhandled_request_num,
"available_batch": int(self.engine.resource_manager.available_batch()),
}
return server_info