mirror of
https://github.com/catlog22/Claude-Code-Workflow.git
synced 2026-03-21 19:08:17 +08:00
feat: 添加重排序模型配置,支持最大输入令牌数,优化 API 批处理能力
This commit is contained in:
@@ -400,6 +400,11 @@ class HybridSearchEngine:
|
||||
elif backend == "legacy":
|
||||
if not bool(getattr(self._config, "embedding_use_gpu", True)):
|
||||
device = "cpu"
|
||||
elif backend == "api":
|
||||
# Pass max_input_tokens for adaptive batching
|
||||
max_tokens = getattr(self._config, "reranker_max_input_tokens", None)
|
||||
if max_tokens:
|
||||
kwargs["max_input_tokens"] = max_tokens
|
||||
|
||||
return get_reranker(
|
||||
backend=backend,
|
||||
|
||||
Reference in New Issue
Block a user