mirror of
https://github.com/meta-llama/llama-stack.git
synced 2025-10-07 12:47:37 +00:00
d23ed26238
# What does this PR do? - implement get_api_key instead of relying on LiteLLMOpenAIMixin.get_api_key - remove use of LiteLLMOpenAIMixin - add default initialize/shutdown methods to OpenAIMixin - remove __init__s to allow proper pydantic construction - remove dead code from vllm adapter and associated / duplicate unit tests - update vllm adapter to use openaimixin for model registration - remove ModelRegistryHelper from fireworks & together adapters - remove Inference from nvidia adapter - complete type hints on embedding_model_metadata - allow extra fields on OpenAIMixin, for model_store, __provider_id__, etc - new recordings for ollama - enhance the list models error handling - update cerebras (remove cerebras-cloud-sdk) and anthropic (custom model listing) inference adapters - parametrized test_inference_client_caching - remove cerebras, databricks, fireworks, together from blanket mypy exclude - removed unnecessary litellm deps ## Test Plan ci
86 lines
3.1 KiB
Python
86 lines
3.1 KiB
Python
# Copyright (c) Meta Platforms, Inc. and affiliates.
|
|
# All rights reserved.
|
|
#
|
|
# This source code is licensed under the terms described in the LICENSE file in
|
|
# the root directory of this source tree.
|
|
|
|
|
|
from collections.abc import Iterable
|
|
|
|
from huggingface_hub import AsyncInferenceClient, HfApi
|
|
from pydantic import SecretStr
|
|
|
|
from llama_stack.apis.inference import OpenAIEmbeddingsResponse
|
|
from llama_stack.log import get_logger
|
|
from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
|
|
|
|
from .config import InferenceAPIImplConfig, InferenceEndpointImplConfig, TGIImplConfig
|
|
|
|
log = get_logger(name=__name__, category="inference::tgi")
|
|
|
|
|
|
class _HfAdapter(OpenAIMixin):
|
|
url: str
|
|
api_key: SecretStr
|
|
|
|
hf_client: AsyncInferenceClient
|
|
max_tokens: int
|
|
model_id: str
|
|
|
|
overwrite_completion_id = True # TGI always returns id=""
|
|
|
|
def get_api_key(self):
|
|
return self.api_key.get_secret_value()
|
|
|
|
def get_base_url(self):
|
|
return self.url
|
|
|
|
async def list_provider_model_ids(self) -> Iterable[str]:
|
|
return [self.model_id]
|
|
|
|
async def openai_embeddings(
|
|
self,
|
|
model: str,
|
|
input: str | list[str],
|
|
encoding_format: str | None = "float",
|
|
dimensions: int | None = None,
|
|
user: str | None = None,
|
|
) -> OpenAIEmbeddingsResponse:
|
|
raise NotImplementedError()
|
|
|
|
|
|
class TGIAdapter(_HfAdapter):
|
|
async def initialize(self, config: TGIImplConfig) -> None:
|
|
if not config.url:
|
|
raise ValueError("You must provide a URL in run.yaml (or via the TGI_URL environment variable) to use TGI.")
|
|
log.info(f"Initializing TGI client with url={config.url}")
|
|
self.hf_client = AsyncInferenceClient(model=config.url, provider="hf-inference")
|
|
endpoint_info = await self.hf_client.get_endpoint_info()
|
|
self.max_tokens = endpoint_info["max_total_tokens"]
|
|
self.model_id = endpoint_info["model_id"]
|
|
self.url = f"{config.url.rstrip('/')}/v1"
|
|
self.api_key = SecretStr("NO_KEY")
|
|
|
|
|
|
class InferenceAPIAdapter(_HfAdapter):
|
|
async def initialize(self, config: InferenceAPIImplConfig) -> None:
|
|
self.hf_client = AsyncInferenceClient(model=config.huggingface_repo, token=config.api_token.get_secret_value())
|
|
endpoint_info = await self.hf_client.get_endpoint_info()
|
|
self.max_tokens = endpoint_info["max_total_tokens"]
|
|
self.model_id = endpoint_info["model_id"]
|
|
# TODO: how do we set url for this?
|
|
|
|
|
|
class InferenceEndpointAdapter(_HfAdapter):
|
|
async def initialize(self, config: InferenceEndpointImplConfig) -> None:
|
|
# Get the inference endpoint details
|
|
api = HfApi(token=config.api_token.get_secret_value())
|
|
endpoint = api.get_inference_endpoint(config.endpoint_name)
|
|
# Wait for the endpoint to be ready (if not already)
|
|
endpoint.wait(timeout=60)
|
|
|
|
# Initialize the adapter
|
|
self.hf_client = endpoint.async_client
|
|
self.model_id = endpoint.repository
|
|
self.max_tokens = int(endpoint.raw["model"]["image"]["custom"]["env"]["MAX_TOTAL_TOKENS"])
|
|
# TODO: how do we set url for this?
|