phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

Allow models to be registered as long as llama model is provided (#472)

Browse source 
This PR allows models to be registered with provider as long as the user
specifies a llama model, even though the model does not match our
prebuilt provider specific mapping.
Test:
pytest -v -s
llama_stack/providers/tests/inference/test_model_registration.py -m
"together" --env TOGETHER_API_KEY=<KEY>

---------

Co-authored-by: Dinesh Yeduguru <dineshyv@fb.com>
This commit is contained in:
Dinesh Yeduguru 2024-11-18 15:05:29 -08:00 • committed by GitHub
parent 2a31163178
commit 57a9b4d57f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 72 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

38
llama_stack/providers/utils/inference/model_registry.py
View file

@ -11,6 +11,10 @@ from llama_models.sku_list import all_registered_models
from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate
from llama_stack.providers.utils.inference import (
ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR,
)
ModelAlias = namedtuple("ModelAlias", ["provider_model_id", "aliases", "llama_model"])
@ -51,7 +55,7 @@ class ModelRegistryHelper(ModelsProtocolPrivate):
if identifier in self.alias_to_provider_id_map:
return self.alias_to_provider_id_map[identifier]
else:
raise ValueError(f"Unknown model: `{identifier}`")
return None
def get_llama_model(self, provider_model_id: str) -> str:
if provider_model_id in self.provider_id_to_llama_model_map:
@ -60,8 +64,34 @@ class ModelRegistryHelper(ModelsProtocolPrivate):
return None
async def register_model(self, model: Model) -> Model:
model.provider_resource_id = self.get_provider_model_id(
model.provider_resource_id
)
provider_resource_id = self.get_provider_model_id(model.provider_resource_id)
if provider_resource_id:
model.provider_resource_id = provider_resource_id
else:
if model.metadata.get("llama_model") is None:
raise ValueError(
f"Model '{model.provider_resource_id}' is not available and no llama_model was specified in metadata. "
"Please specify a llama_model in metadata or use a supported model identifier"
)
existing_llama_model = self.get_llama_model(model.provider_resource_id)
if existing_llama_model:
if existing_llama_model != model.metadata["llama_model"]:
raise ValueError(
f"Provider model id '{model.provider_resource_id}' is already registered to a different llama model: '{existing_llama_model}'"
)
else:
if (
model.metadata["llama_model"]
not in ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR
):
raise ValueError(
f"Invalid llama_model '{model.metadata['llama_model']}' specified in metadata. "
f"Must be one of: {', '.join(ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR.keys())}"
)
self.provider_id_to_llama_model_map[model.provider_resource_id] = (
ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR[
model.metadata["llama_model"]
]
)
return model