phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-08-02 08:44:44 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

address comments

Browse source
This commit is contained in:
Botao Chen 2024-12-18 14:12:57 -08:00
parent d021983b0e
commit 0000e1e8c6
2 changed files with 7 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama_stack/providers/inline/inference/meta_reference/inference.py
View file

@ -73,7 +73,7 @@ class MetaReferenceInferenceImpl(
self.model_id = None
self.llama_model = None
async def initialize(self, model_id, llama_model) -> None:
async def load_model(self, model_id, llama_model) -> None:
log.info(f"Loading model `{model_id}`")
if self.config.create_distributed_process_group:
self.generator = LlamaModelParallelGenerator(
@ -127,9 +127,9 @@ class MetaReferenceInferenceImpl(
if model.model_type == ModelType.embedding:
self._load_sentence_transformer_model(model.provider_resource_id)
if "skip_initialize" in model.metadata and model.metadata["skip_initialize"]:
if "skip_load" in model.metadata and model.metadata["skip_load"]:
return model
await self.initialize(model.identifier, llama_model)
await self.load_model(model.identifier, llama_model)
return model
async def completion(

8
llama_stack/providers/tests/inference/test_model_registration.py
View file

@ -55,7 +55,7 @@ class TestModelRegistration:
model_id="custom-model",
metadata={
"llama_model": "meta-llama/Llama-2-7b",
"skip_initialize": True,
"skip_load": True,
},
)
@ -73,16 +73,16 @@ class TestModelRegistration:
_, models_impl = inference_stack
with patch(
"llama_stack.providers.inline.inference.meta_reference.inference.MetaReferenceInferenceImpl.initialize",
"llama_stack.providers.inline.inference.meta_reference.inference.MetaReferenceInferenceImpl.load_model",
new_callable=AsyncMock,
) as mock_initialize:
) as mock_load_model:
_ = await models_impl.register_model(
model_id="Llama3.1-8B-Instruct",
metadata={
"llama_model": "meta-llama/Llama-3.1-8B-Instruct",
},
)
mock_initialize.assert_called_once()
mock_load_model.assert_called_once()
@pytest.mark.asyncio
async def test_register_with_invalid_llama_model(self, inference_stack):