Skip to content
Closed
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 13 additions & 0 deletions components/src/dynamo/sglang/request_handlers/handler_base.py
Original file line number Diff line number Diff line change
Expand Up @@ -413,6 +413,18 @@ async def load_lora(self, request: Optional[Dict[str, Any]] = None):
else:
lora_worker_type = WorkerType.Aggregated
lora_needs = []

# Reuse the base-model metadata builder so LoRA
# cards advertise the same token-overflow policy,
# parser configuration, and routing capabilities.
# Lazy import: static test collection lacks parts of SGLang.
from dynamo.sglang.register import _get_runtime_config

runtime_config = await _get_runtime_config(
self.engine,
self.config.server_args,
self.config.dynamo_args,
)
await register_llm(
model_input=ModelInput.Tokens,
model_type=lora_model_type,
Expand All @@ -424,6 +436,7 @@ async def load_lora(self, request: Optional[Dict[str, Any]] = None):
base_model_path=self.config.server_args.model_path,
worker_type=lora_worker_type,
needs=lora_needs,
runtime_config=runtime_config,
# Publish the worker's per-worker LoRA slot budget so the frontend
# allocator sizes placement against real capacity instead of the
# hard-coded default.
Expand Down
Loading