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

Add a RoutableProvider protocol, support for multiple routing keys (#163)

Browse source 
* Update configure.py to use multiple routing keys for safety
* Refactor distribution/datatypes into a providers/datatypes
* Cleanup
This commit is contained in:
Ashwin Bharambe 2024-09-30 17:30:21 -07:00 • committed by GitHub
parent 73decb3781
commit eb2d8a31a5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
24 changed files with 600 additions and 577 deletions
Download patch file Download diff file Expand all files Collapse all files

34
llama_stack/providers/adapters/inference/bedrock/bedrock.py
View file

@ -12,20 +12,21 @@ from botocore.config import Config
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
from llama_models.sku_list import resolve_model
from llama_stack.providers.utils.inference.routable import RoutableProviderForModels
from llama_stack.apis.inference import * # noqa: F403
from llama_stack.providers.adapters.inference.bedrock.config import BedrockConfig
# mapping of Model SKUs to ollama models
BEDROCK_SUPPORTED_MODELS = {
"Meta-Llama3.1-8B-Instruct": "meta.llama3-1-8b-instruct-v1:0",
"Meta-Llama3.1-70B-Instruct": "meta.llama3-1-70b-instruct-v1:0",
"Meta-Llama3.1-405B-Instruct": "meta.llama3-1-405b-instruct-v1:0",
"Llama3.1-8B-Instruct": "meta.llama3-1-8b-instruct-v1:0",
"Llama3.1-70B-Instruct": "meta.llama3-1-70b-instruct-v1:0",
"Llama3.1-405B-Instruct": "meta.llama3-1-405b-instruct-v1:0",
}
class BedrockInferenceAdapter(Inference):
class BedrockInferenceAdapter(Inference, RoutableProviderForModels):
@staticmethod
def _create_bedrock_client(config: BedrockConfig) -> BaseClient:
@ -68,6 +69,9 @@ class BedrockInferenceAdapter(Inference):
return boto3_session.client("bedrock-runtime", config=boto3_config)
def __init__(self, config: BedrockConfig) -> None:
RoutableProviderForModels.__init__(
self, stack_to_provider_models_map=BEDROCK_SUPPORTED_MODELS
)
self._config = config
self._client = BedrockInferenceAdapter._create_bedrock_client(config)
@ -94,22 +98,6 @@ class BedrockInferenceAdapter(Inference):
) -> Union[CompletionResponse, CompletionResponseStreamChunk]:
raise NotImplementedError()
@staticmethod
def resolve_bedrock_model(model_name: str) -> str:
model = resolve_model(model_name)
assert (
model is not None
and model.descriptor(shorten_default_variant=True)
in BEDROCK_SUPPORTED_MODELS
), (
f"Unsupported model: {model_name}, use one of the supported models: "
f"{','.join(BEDROCK_SUPPORTED_MODELS.keys())}"
)
return BEDROCK_SUPPORTED_MODELS.get(
model.descriptor(shorten_default_variant=True)
)
@staticmethod
def _bedrock_stop_reason_to_stop_reason(bedrock_stop_reason: str) -> StopReason:
if bedrock_stop_reason == "max_tokens":
@ -350,7 +338,7 @@ class BedrockInferenceAdapter(Inference):
) -> (
AsyncGenerator
): # Union[ChatCompletionResponse, ChatCompletionResponseStreamChunk]:
bedrock_model = BedrockInferenceAdapter.resolve_bedrock_model(model)
bedrock_model = self.map_to_provider_model(model)
inference_config = BedrockInferenceAdapter.get_bedrock_inference_config(
sampling_params
)