phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-08-03 17:29:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Added a comment explaining the client handling in register_model.

Browse source
This commit is contained in:
ilya-kolchinsky 2025-04-16 15:29:23 +02:00
parent f1fd382d51
commit 6a0ee7180b
1 changed files with 3 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

3
llama_stack/providers/remote/inference/vllm/vllm.py
View file

@ -368,6 +368,9 @@ class VLLMInferenceAdapter(Inference, ModelsProtocolPrivate):
yield chunk yield chunk
async def register_model(self, model: Model) -> Model: async def register_model(self, model: Model) -> Model:
# register_model is called during Llama Stack initialization, hence we cannot init self.client if not initialized yet.
# self.client should only be created after the initialization is complete to avoid asyncio cross-context errors.
# Changing this may lead to unpredictable behavior.
client = self._create_client() if self.client is None else self.client client = self._create_client() if self.client is None else self.client
model = await self.register_helper.register_model(model) model = await self.register_helper.register_model(model)
res = await client.models.list() res = await client.models.list()