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

migrate model to Resource and new registration signature (#410)

Browse source 
* resource oriented object design for models

* add back llama_model field

* working tests

* register singature fix

* address feedback

---------

Co-authored-by: Dinesh Yeduguru <dineshyv@fb.com>
This commit is contained in:
Dinesh Yeduguru 2024-11-08 16:12:57 -08:00 • committed by GitHub
parent bd0622ef10
commit ec644d3418
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
17 changed files with 99 additions and 90 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama_stack/providers/datatypes.py
View file

@ -14,7 +14,7 @@ from pydantic import BaseModel, Field
from llama_stack.apis.datasets import DatasetDef
from llama_stack.apis.eval_tasks import EvalTaskDef
from llama_stack.apis.memory_banks import MemoryBankDef
from llama_stack.apis.models import ModelDef
from llama_stack.apis.models import Model
from llama_stack.apis.scoring_functions import ScoringFnDef
from llama_stack.apis.shields import Shield
@ -43,9 +43,7 @@ class Api(Enum):
class ModelsProtocolPrivate(Protocol):
async def list_models(self) -> List[ModelDef]: ...
async def register_model(self, model: ModelDef) -> None: ...
async def register_model(self, model: Model) -> None: ...
class ShieldsProtocolPrivate(Protocol):

15
llama_stack/providers/inline/inference/meta_reference/inference.py
View file

@ -12,7 +12,7 @@ from llama_models.sku_list import resolve_model
from llama_models.llama3.api.datatypes import * # noqa: F403
from llama_stack.apis.inference import * # noqa: F403
from llama_stack.providers.datatypes import ModelDef, ModelsProtocolPrivate
from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate
from llama_stack.providers.utils.inference.prompt_adapter import (
convert_image_media_to_url,
@ -45,16 +45,11 @@ class MetaReferenceInferenceImpl(Inference, ModelsProtocolPrivate):
else:
self.generator = Llama.build(self.config)
async def register_model(self, model: ModelDef) -> None:
raise ValueError("Dynamic model registration is not supported")
async def list_models(self) -> List[ModelDef]:
return [
ModelDef(
identifier=self.model.descriptor(),
llama_model=self.model.descriptor(),
async def register_model(self, model: Model) -> None:
if model.identifier != self.model.descriptor():
raise ValueError(
f"Model mismatch: {model.identifier} != {self.model.descriptor()}"
)
]
async def shutdown(self) -> None:
if self.config.create_distributed_process_group:

12
llama_stack/providers/inline/inference/vllm/vllm.py
View file

@ -20,7 +20,7 @@ from vllm.sampling_params import SamplingParams as VLLMSamplingParams
from llama_stack.apis.inference import * # noqa: F403
from llama_stack.providers.datatypes import ModelDef, ModelsProtocolPrivate
from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate
from llama_stack.providers.utils.inference.openai_compat import (
OpenAICompatCompletionChoice,
OpenAICompatCompletionResponse,
@ -83,19 +83,11 @@ class VLLMInferenceImpl(Inference, ModelsProtocolPrivate):
if self.engine:
self.engine.shutdown_background_loop()
async def register_model(self, model: ModelDef) -> None:
async def register_model(self, model: Model) -> None:
raise ValueError(
"You cannot dynamically add a model to a running vllm instance"
)
async def list_models(self) -> List[ModelDef]:
return [
ModelDef(
identifier=self.config.model,
llama_model=self.config.model,
)
]
def _sampling_params(self, sampling_params: SamplingParams) -> VLLMSamplingParams:
if sampling_params is None:
return VLLMSamplingParams(max_tokens=self.config.max_tokens)

12
llama_stack/providers/remote/inference/ollama/ollama.py
View file

@ -15,7 +15,7 @@ from llama_models.llama3.api.tokenizer import Tokenizer
from ollama import AsyncClient
from llama_stack.apis.inference import * # noqa: F403
from llama_stack.providers.datatypes import ModelsProtocolPrivate
from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate
from llama_stack.providers.utils.inference.openai_compat import (
get_sampling_options,
@ -65,10 +65,11 @@ class OllamaInferenceAdapter(Inference, ModelsProtocolPrivate):
async def shutdown(self) -> None:
pass
async def register_model(self, model: ModelDef) -> None:
raise ValueError("Dynamic model registration is not supported")
async def register_model(self, model: Model) -> None:
if model.identifier not in OLLAMA_SUPPORTED_MODELS:
raise ValueError(f"Model {model.identifier} is not supported by Ollama")
async def list_models(self) -> List[ModelDef]:
async def list_models(self) -> List[Model]:
ollama_to_llama = {v: k for k, v in OLLAMA_SUPPORTED_MODELS.items()}
ret = []
@ -80,9 +81,8 @@ class OllamaInferenceAdapter(Inference, ModelsProtocolPrivate):
llama_model = ollama_to_llama[r["model"]]
ret.append(
ModelDef(
Model(
identifier=llama_model,
llama_model=llama_model,
metadata={
"ollama_model": r["model"],
},

2
llama_stack/providers/remote/inference/sample/sample.py
View file

@ -14,7 +14,7 @@ class SampleInferenceImpl(Inference):
def __init__(self, config: SampleConfig):
self.config = config
async def register_model(self, model: ModelDef) -> None:
async def register_model(self, model: Model) -> None:
# these are the model names the Llama Stack will use to route requests to this provider
# perform validation here if necessary
pass

10
llama_stack/providers/remote/inference/tgi/tgi.py
View file

@ -16,7 +16,7 @@ from llama_models.sku_list import all_registered_models
from llama_stack.apis.inference import * # noqa: F403
from llama_stack.apis.models import * # noqa: F403
from llama_stack.providers.datatypes import ModelDef, ModelsProtocolPrivate
from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate
from llama_stack.providers.utils.inference.openai_compat import (
get_sampling_options,
@ -50,14 +50,14 @@ class _HfAdapter(Inference, ModelsProtocolPrivate):
if model.huggingface_repo
}
async def register_model(self, model: ModelDef) -> None:
raise ValueError("Model registration is not supported for HuggingFace models")
async def register_model(self, model: Model) -> None:
pass
async def list_models(self) -> List[ModelDef]:
async def list_models(self) -> List[Model]:
repo = self.model_id
identifier = self.huggingface_repo_to_llama_model_id[repo]
return [
ModelDef(
Model(
identifier=identifier,
llama_model=identifier,
metadata={

8
llama_stack/providers/remote/inference/vllm/vllm.py
View file

@ -13,7 +13,7 @@ from llama_models.sku_list import all_registered_models, resolve_model
from openai import OpenAI
from llama_stack.apis.inference import * # noqa: F403
from llama_stack.providers.datatypes import ModelsProtocolPrivate
from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate
from llama_stack.providers.utils.inference.openai_compat import (
get_sampling_options,
@ -44,13 +44,13 @@ class VLLMInferenceAdapter(Inference, ModelsProtocolPrivate):
async def initialize(self) -> None:
self.client = OpenAI(base_url=self.config.url, api_key=self.config.api_token)
async def register_model(self, model: ModelDef) -> None:
async def register_model(self, model: Model) -> None:
raise ValueError("Model registration is not supported for vLLM models")
async def shutdown(self) -> None:
pass
async def list_models(self) -> List[ModelDef]:
async def list_models(self) -> List[Model]:
models = []
for model in self.client.models.list():
repo = model.id
@ -60,7 +60,7 @@ class VLLMInferenceAdapter(Inference, ModelsProtocolPrivate):
identifier = self.huggingface_repo_to_llama_model_id[repo]
models.append(
ModelDef(
Model(
identifier=identifier,
llama_model=identifier,
)

7
llama_stack/providers/tests/inference/fixtures.py
View file

@ -153,7 +153,7 @@ INFERENCE_FIXTURES = [
@pytest_asyncio.fixture(scope="session")
async def inference_stack(request):
async def inference_stack(request, inference_model):
fixture_name = request.param
inference_fixture = request.getfixturevalue(f"inference_{fixture_name}")
impls = await resolve_impls_for_test_v2(
@ -162,4 +162,9 @@ async def inference_stack(request):
inference_fixture.provider_data,
)
await impls[Api.models].register_model(
model_id=inference_model,
provider_model_id=inference_fixture.providers[0].provider_id,
)
return (impls[Api.inference], impls[Api.models])

2
llama_stack/providers/tests/inference/test_text_inference.py
View file

@ -69,7 +69,7 @@ class TestInference:
response = await models_impl.list_models()
assert isinstance(response, list)
assert len(response) >= 1
assert all(isinstance(model, ModelDefWithProvider) for model in response)
assert all(isinstance(model, Model) for model in response)
model_def = None
for model in response:

12
llama_stack/providers/utils/inference/model_registry.py
View file

@ -4,11 +4,11 @@
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
from typing import Dict, List
from typing import Dict
from llama_models.sku_list import resolve_model
from llama_stack.providers.datatypes import ModelDef, ModelsProtocolPrivate
from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate
class ModelRegistryHelper(ModelsProtocolPrivate):
@ -28,14 +28,8 @@ class ModelRegistryHelper(ModelsProtocolPrivate):
return self.stack_to_provider_models_map[identifier]
async def register_model(self, model: ModelDef) -> None:
async def register_model(self, model: Model) -> None:
if model.identifier not in self.stack_to_provider_models_map:
raise ValueError(
f"Unsupported model {model.identifier}. Supported models: {self.stack_to_provider_models_map.keys()}"
)
async def list_models(self) -> List[ModelDef]:
models = []
for llama_model, provider_model in self.stack_to_provider_models_map.items():
models.append(ModelDef(identifier=llama_model, llama_model=llama_model))
return models