mirror of
https://github.com/meta-llama/llama-stack.git
synced 2025-12-16 10:49:26 +00:00
Ran precommit
This commit is contained in:
Omar Abdelwahab
parent
9886520b40
commit
9fc0d966f6
7 changed files with 153 additions and 310 deletions
|
|
@ -7,10 +7,9 @@
|
|||
|
||||
from llama_stack.apis.inference import * # noqa: F403
|
||||
from llama_stack.apis.inference import OpenAIEmbeddingsResponse
|
||||
|
||||
from llama_stack.providers.utils.inference.model_registry import (
|
||||
build_hf_repo_model_entry,
|
||||
ModelRegistryHelper,
|
||||
build_hf_repo_model_entry,
|
||||
)
|
||||
from llama_stack.providers.utils.inference.openai_compat import (
|
||||
get_sampling_options,
|
||||
|
|
@ -51,9 +50,7 @@ class RunpodInferenceAdapter(
|
|||
Inference,
|
||||
):
|
||||
def __init__(self, config: RunpodImplConfig) -> None:
|
||||
ModelRegistryHelper.__init__(
|
||||
self, stack_to_provider_models_map=RUNPOD_SUPPORTED_MODELS
|
||||
)
|
||||
ModelRegistryHelper.__init__(self, stack_to_provider_models_map=RUNPOD_SUPPORTED_MODELS)
|
||||
self.config = config
|
||||
|
||||
def _get_params(self, request: ChatCompletionRequest) -> dict:
|
||||
|
|
|
|||
|
|
@ -9,6 +9,7 @@ from typing import Any
|
|||
|
||||
from ibm_watsonx_ai.foundation_models import Model
|
||||
from ibm_watsonx_ai.metanames import GenTextParamsMetaNames as GenParams
|
||||
from openai import AsyncOpenAI
|
||||
|
||||
from llama_stack.apis.inference import (
|
||||
ChatCompletionRequest,
|
||||
|
|
@ -33,7 +34,6 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
|
|||
completion_request_to_prompt,
|
||||
request_has_media,
|
||||
)
|
||||
from openai import AsyncOpenAI
|
||||
|
||||
from . import WatsonXConfig
|
||||
from .models import MODEL_ENTRIES
|
||||
|
|
@ -65,9 +65,7 @@ class WatsonXInferenceAdapter(Inference, ModelRegistryHelper):
|
|||
self._project_id = self._config.project_id
|
||||
|
||||
def _get_client(self, model_id) -> Model:
|
||||
config_api_key = (
|
||||
self._config.api_key.get_secret_value() if self._config.api_key else None
|
||||
)
|
||||
config_api_key = self._config.api_key.get_secret_value() if self._config.api_key else None
|
||||
config_url = self._config.url
|
||||
project_id = self._config.project_id
|
||||
credentials = {"url": config_url, "apikey": config_api_key}
|
||||
|
|
@ -82,46 +80,28 @@ class WatsonXInferenceAdapter(Inference, ModelRegistryHelper):
|
|||
)
|
||||
return self._openai_client
|
||||
|
||||
|
||||
async def _get_params(self, request: ChatCompletionRequest) -> dict:
|
||||
|
||||
input_dict = {"params": {}}
|
||||
media_present = request_has_media(request)
|
||||
llama_model = self.get_llama_model(request.model)
|
||||
if isinstance(request, ChatCompletionRequest):
|
||||
input_dict["prompt"] = await chat_completion_request_to_prompt(
|
||||
request, llama_model
|
||||
)
|
||||
input_dict["prompt"] = await chat_completion_request_to_prompt(request, llama_model)
|
||||
else:
|
||||
assert (
|
||||
not media_present
|
||||
), "Together does not support media for Completion requests"
|
||||
assert not media_present, "Together does not support media for Completion requests"
|
||||
input_dict["prompt"] = await completion_request_to_prompt(request)
|
||||
if request.sampling_params:
|
||||
if request.sampling_params.strategy:
|
||||
input_dict["params"][
|
||||
GenParams.DECODING_METHOD
|
||||
] = request.sampling_params.strategy.type
|
||||
input_dict["params"][GenParams.DECODING_METHOD] = request.sampling_params.strategy.type
|
||||
if request.sampling_params.max_tokens:
|
||||
input_dict["params"][
|
||||
GenParams.MAX_NEW_TOKENS
|
||||
] = request.sampling_params.max_tokens
|
||||
input_dict["params"][GenParams.MAX_NEW_TOKENS] = request.sampling_params.max_tokens
|
||||
if request.sampling_params.repetition_penalty:
|
||||
input_dict["params"][
|
||||
GenParams.REPETITION_PENALTY
|
||||
] = request.sampling_params.repetition_penalty
|
||||
input_dict["params"][GenParams.REPETITION_PENALTY] = request.sampling_params.repetition_penalty
|
||||
|
||||
if isinstance(request.sampling_params.strategy, TopPSamplingStrategy):
|
||||
input_dict["params"][
|
||||
GenParams.TOP_P
|
||||
] = request.sampling_params.strategy.top_p
|
||||
input_dict["params"][
|
||||
GenParams.TEMPERATURE
|
||||
] = request.sampling_params.strategy.temperature
|
||||
input_dict["params"][GenParams.TOP_P] = request.sampling_params.strategy.top_p
|
||||
input_dict["params"][GenParams.TEMPERATURE] = request.sampling_params.strategy.temperature
|
||||
if isinstance(request.sampling_params.strategy, TopKSamplingStrategy):
|
||||
input_dict["params"][
|
||||
GenParams.TOP_K
|
||||
] = request.sampling_params.strategy.top_k
|
||||
input_dict["params"][GenParams.TOP_K] = request.sampling_params.strategy.top_k
|
||||
if isinstance(request.sampling_params.strategy, GreedySamplingStrategy):
|
||||
input_dict["params"][GenParams.TEMPERATURE] = 0.0
|
||||
|
||||
|
|
|
|||
|
|
@ -15,9 +15,17 @@ from typing import Any
|
|||
from openai import AsyncStream
|
||||
from openai.types.chat import (
|
||||
ChatCompletionAssistantMessageParam as OpenAIChatCompletionAssistantMessage,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionChunk as OpenAIChatCompletionChunk,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionContentPartImageParam as OpenAIChatCompletionContentPartImageParam,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionContentPartParam as OpenAIChatCompletionContentPartParam,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionContentPartTextParam as OpenAIChatCompletionContentPartTextParam,
|
||||
)
|
||||
|
||||
|
|
@ -29,15 +37,56 @@ except ImportError:
|
|||
from openai.types.chat.chat_completion_message_tool_call import (
|
||||
ChatCompletionMessageToolCall as OpenAIChatCompletionMessageFunctionToolCall,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionMessageParam as OpenAIChatCompletionMessage,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionMessageToolCall,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionSystemMessageParam as OpenAIChatCompletionSystemMessage,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionToolMessageParam as OpenAIChatCompletionToolMessage,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionUserMessageParam as OpenAIChatCompletionUserMessage,
|
||||
)
|
||||
from openai.types.chat.chat_completion import (
|
||||
Choice as OpenAIChoice,
|
||||
)
|
||||
from openai.types.chat.chat_completion import (
|
||||
ChoiceLogprobs as OpenAIChoiceLogprobs, # same as chat_completion_chunk ChoiceLogprobs
|
||||
)
|
||||
from openai.types.chat.chat_completion_chunk import (
|
||||
Choice as OpenAIChatCompletionChunkChoice,
|
||||
)
|
||||
from openai.types.chat.chat_completion_chunk import (
|
||||
ChoiceDelta as OpenAIChoiceDelta,
|
||||
)
|
||||
from openai.types.chat.chat_completion_chunk import (
|
||||
ChoiceDeltaToolCall as OpenAIChoiceDeltaToolCall,
|
||||
)
|
||||
from openai.types.chat.chat_completion_chunk import (
|
||||
ChoiceDeltaToolCallFunction as OpenAIChoiceDeltaToolCallFunction,
|
||||
)
|
||||
from openai.types.chat.chat_completion_content_part_image_param import (
|
||||
ImageURL as OpenAIImageURL,
|
||||
)
|
||||
from openai.types.chat.chat_completion_message_tool_call import (
|
||||
Function as OpenAIFunction,
|
||||
)
|
||||
from pydantic import BaseModel
|
||||
|
||||
from llama_stack.apis.common.content_types import (
|
||||
_URLOrData,
|
||||
URL,
|
||||
ImageContentItem,
|
||||
InterleavedContent,
|
||||
TextContentItem,
|
||||
TextDelta,
|
||||
ToolCallDelta,
|
||||
ToolCallParseStatus,
|
||||
URL,
|
||||
_URLOrData,
|
||||
)
|
||||
from llama_stack.apis.inference import (
|
||||
ChatCompletionRequest,
|
||||
|
|
@ -74,30 +123,6 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
|
|||
convert_image_content_to_url,
|
||||
decode_assistant_message,
|
||||
)
|
||||
from openai.types.chat import (
|
||||
ChatCompletionMessageParam as OpenAIChatCompletionMessage,
|
||||
ChatCompletionMessageToolCall,
|
||||
ChatCompletionSystemMessageParam as OpenAIChatCompletionSystemMessage,
|
||||
ChatCompletionToolMessageParam as OpenAIChatCompletionToolMessage,
|
||||
ChatCompletionUserMessageParam as OpenAIChatCompletionUserMessage,
|
||||
)
|
||||
from openai.types.chat.chat_completion import (
|
||||
Choice as OpenAIChoice,
|
||||
ChoiceLogprobs as OpenAIChoiceLogprobs, # same as chat_completion_chunk ChoiceLogprobs
|
||||
)
|
||||
from openai.types.chat.chat_completion_chunk import (
|
||||
Choice as OpenAIChatCompletionChunkChoice,
|
||||
ChoiceDelta as OpenAIChoiceDelta,
|
||||
ChoiceDeltaToolCall as OpenAIChoiceDeltaToolCall,
|
||||
ChoiceDeltaToolCallFunction as OpenAIChoiceDeltaToolCallFunction,
|
||||
)
|
||||
from openai.types.chat.chat_completion_content_part_image_param import (
|
||||
ImageURL as OpenAIImageURL,
|
||||
)
|
||||
from openai.types.chat.chat_completion_message_tool_call import (
|
||||
Function as OpenAIFunction,
|
||||
)
|
||||
from pydantic import BaseModel
|
||||
|
||||
logger = get_logger(name=__name__, category="providers::utils")
|
||||
|
||||
|
|
@ -196,16 +221,12 @@ def convert_openai_completion_logprobs(
|
|||
if logprobs.tokens and logprobs.token_logprobs:
|
||||
return [
|
||||
TokenLogProbs(logprobs_by_token={token: token_lp})
|
||||
for token, token_lp in zip(
|
||||
logprobs.tokens, logprobs.token_logprobs, strict=False
|
||||
)
|
||||
for token, token_lp in zip(logprobs.tokens, logprobs.token_logprobs, strict=False)
|
||||
]
|
||||
return None
|
||||
|
||||
|
||||
def convert_openai_completion_logprobs_stream(
|
||||
text: str, logprobs: float | OpenAICompatLogprobs | None
|
||||
):
|
||||
def convert_openai_completion_logprobs_stream(text: str, logprobs: float | OpenAICompatLogprobs | None):
|
||||
if logprobs is None:
|
||||
return None
|
||||
if isinstance(logprobs, float):
|
||||
|
|
@ -250,9 +271,7 @@ def process_chat_completion_response(
|
|||
if not choice.message or not choice.message.tool_calls:
|
||||
raise ValueError("Tool calls are not present in the response")
|
||||
|
||||
tool_calls = [
|
||||
convert_tool_call(tool_call) for tool_call in choice.message.tool_calls
|
||||
]
|
||||
tool_calls = [convert_tool_call(tool_call) for tool_call in choice.message.tool_calls]
|
||||
if any(isinstance(tool_call, UnparseableToolCall) for tool_call in tool_calls):
|
||||
# If we couldn't parse a tool call, jsonify the tool calls and return them
|
||||
return ChatCompletionResponse(
|
||||
|
|
@ -276,9 +295,7 @@ def process_chat_completion_response(
|
|||
|
||||
# TODO: This does not work well with tool calls for vLLM remote provider
|
||||
# Ref: https://github.com/meta-llama/llama-stack/issues/1058
|
||||
raw_message = decode_assistant_message(
|
||||
text_from_choice(choice), get_stop_reason(choice.finish_reason)
|
||||
)
|
||||
raw_message = decode_assistant_message(text_from_choice(choice), get_stop_reason(choice.finish_reason))
|
||||
|
||||
# NOTE: If we do not set tools in chat-completion request, we should not
|
||||
# expect the ToolCall in the response. Instead, we should return the raw
|
||||
|
|
@ -479,17 +496,13 @@ async def process_chat_completion_stream_response(
|
|||
)
|
||||
|
||||
|
||||
async def convert_message_to_openai_dict(
|
||||
message: Message, download: bool = False
|
||||
) -> dict:
|
||||
async def convert_message_to_openai_dict(message: Message, download: bool = False) -> dict:
|
||||
async def _convert_content(content) -> dict:
|
||||
if isinstance(content, ImageContentItem):
|
||||
return {
|
||||
"type": "image_url",
|
||||
"image_url": {
|
||||
"url": await convert_image_content_to_url(
|
||||
content, download=download
|
||||
),
|
||||
"url": await convert_image_content_to_url(content, download=download),
|
||||
},
|
||||
}
|
||||
else:
|
||||
|
|
@ -574,11 +587,7 @@ async def convert_message_to_openai_dict_new(
|
|||
) -> str | Iterable[OpenAIChatCompletionContentPartParam]:
|
||||
async def impl(
|
||||
content_: InterleavedContent,
|
||||
) -> (
|
||||
str
|
||||
| OpenAIChatCompletionContentPartParam
|
||||
| list[OpenAIChatCompletionContentPartParam]
|
||||
):
|
||||
) -> str | OpenAIChatCompletionContentPartParam | list[OpenAIChatCompletionContentPartParam]:
|
||||
# Llama Stack and OpenAI spec match for str and text input
|
||||
if isinstance(content_, str):
|
||||
return content_
|
||||
|
|
@ -591,9 +600,7 @@ async def convert_message_to_openai_dict_new(
|
|||
return OpenAIChatCompletionContentPartImageParam(
|
||||
type="image_url",
|
||||
image_url=OpenAIImageURL(
|
||||
url=await convert_image_content_to_url(
|
||||
content_, download=download_images
|
||||
)
|
||||
url=await convert_image_content_to_url(content_, download=download_images)
|
||||
),
|
||||
)
|
||||
elif isinstance(content_, list):
|
||||
|
|
@ -620,11 +627,7 @@ async def convert_message_to_openai_dict_new(
|
|||
OpenAIChatCompletionMessageFunctionToolCall(
|
||||
id=tool.call_id,
|
||||
function=OpenAIFunction(
|
||||
name=(
|
||||
tool.tool_name
|
||||
if not isinstance(tool.tool_name, BuiltinTool)
|
||||
else tool.tool_name.value
|
||||
),
|
||||
name=(tool.tool_name if not isinstance(tool.tool_name, BuiltinTool) else tool.tool_name.value),
|
||||
arguments=tool.arguments, # Already a JSON string, don't double-encode
|
||||
),
|
||||
type="function",
|
||||
|
|
@ -804,9 +807,7 @@ def _convert_openai_finish_reason(finish_reason: str) -> StopReason:
|
|||
}.get(finish_reason, StopReason.end_of_turn)
|
||||
|
||||
|
||||
def _convert_openai_request_tool_config(
|
||||
tool_choice: str | dict[str, Any] | None = None
|
||||
) -> ToolConfig:
|
||||
def _convert_openai_request_tool_config(tool_choice: str | dict[str, Any] | None = None) -> ToolConfig:
|
||||
tool_config = ToolConfig()
|
||||
if tool_choice:
|
||||
try:
|
||||
|
|
@ -817,9 +818,7 @@ def _convert_openai_request_tool_config(
|
|||
return tool_config
|
||||
|
||||
|
||||
def _convert_openai_request_tools(
|
||||
tools: list[dict[str, Any]] | None = None
|
||||
) -> list[ToolDefinition]:
|
||||
def _convert_openai_request_tools(tools: list[dict[str, Any]] | None = None) -> list[ToolDefinition]:
|
||||
lls_tools = []
|
||||
if not tools:
|
||||
return lls_tools
|
||||
|
|
@ -918,11 +917,7 @@ def _convert_openai_logprobs(
|
|||
return None
|
||||
|
||||
return [
|
||||
TokenLogProbs(
|
||||
logprobs_by_token={
|
||||
logprobs.token: logprobs.logprob for logprobs in content.top_logprobs
|
||||
}
|
||||
)
|
||||
TokenLogProbs(logprobs_by_token={logprobs.token: logprobs.logprob for logprobs in content.top_logprobs})
|
||||
for content in logprobs.content
|
||||
]
|
||||
|
||||
|
|
@ -961,13 +956,9 @@ def openai_messages_to_messages(
|
|||
converted_messages = []
|
||||
for message in messages:
|
||||
if message.role == "system":
|
||||
converted_message = SystemMessage(
|
||||
content=openai_content_to_content(message.content)
|
||||
)
|
||||
converted_message = SystemMessage(content=openai_content_to_content(message.content))
|
||||
elif message.role == "user":
|
||||
converted_message = UserMessage(
|
||||
content=openai_content_to_content(message.content)
|
||||
)
|
||||
converted_message = UserMessage(content=openai_content_to_content(message.content))
|
||||
elif message.role == "assistant":
|
||||
converted_message = CompletionMessage(
|
||||
content=openai_content_to_content(message.content),
|
||||
|
|
@ -999,9 +990,7 @@ def openai_content_to_content(
|
|||
if content.type == "text":
|
||||
return TextContentItem(type="text", text=content.text)
|
||||
elif content.type == "image_url":
|
||||
return ImageContentItem(
|
||||
type="image", image=_URLOrData(url=URL(uri=content.image_url.url))
|
||||
)
|
||||
return ImageContentItem(type="image", image=_URLOrData(url=URL(uri=content.image_url.url)))
|
||||
else:
|
||||
raise ValueError(f"Unknown content type: {content.type}")
|
||||
else:
|
||||
|
|
@ -1041,17 +1030,14 @@ def convert_openai_chat_completion_choice(
|
|||
end_of_message = "end_of_message"
|
||||
out_of_tokens = "out_of_tokens"
|
||||
"""
|
||||
assert (
|
||||
hasattr(choice, "message") and choice.message
|
||||
), "error in server response: message not found"
|
||||
assert (
|
||||
hasattr(choice, "finish_reason") and choice.finish_reason
|
||||
), "error in server response: finish_reason not found"
|
||||
assert hasattr(choice, "message") and choice.message, "error in server response: message not found"
|
||||
assert hasattr(choice, "finish_reason") and choice.finish_reason, (
|
||||
"error in server response: finish_reason not found"
|
||||
)
|
||||
|
||||
return ChatCompletionResponse(
|
||||
completion_message=CompletionMessage(
|
||||
content=choice.message.content
|
||||
or "", # CompletionMessage content is not optional
|
||||
content=choice.message.content or "", # CompletionMessage content is not optional
|
||||
stop_reason=_convert_openai_finish_reason(choice.finish_reason),
|
||||
tool_calls=_convert_openai_tool_calls(choice.message.tool_calls),
|
||||
),
|
||||
|
|
@ -1291,9 +1277,7 @@ class OpenAIChatCompletionToLlamaStackMixin:
|
|||
outstanding_responses.append(response)
|
||||
|
||||
if stream:
|
||||
return OpenAIChatCompletionToLlamaStackMixin._process_stream_response(
|
||||
self, model, outstanding_responses
|
||||
)
|
||||
return OpenAIChatCompletionToLlamaStackMixin._process_stream_response(self, model, outstanding_responses)
|
||||
|
||||
return await OpenAIChatCompletionToLlamaStackMixin._process_non_stream_response(
|
||||
self, model, outstanding_responses
|
||||
|
|
@ -1302,29 +1286,21 @@ class OpenAIChatCompletionToLlamaStackMixin:
|
|||
async def _process_stream_response(
|
||||
self,
|
||||
model: str,
|
||||
outstanding_responses: list[
|
||||
Awaitable[AsyncIterator[ChatCompletionResponseStreamChunk]]
|
||||
],
|
||||
outstanding_responses: list[Awaitable[AsyncIterator[ChatCompletionResponseStreamChunk]]],
|
||||
):
|
||||
id = f"chatcmpl-{uuid.uuid4()}"
|
||||
for i, outstanding_response in enumerate(outstanding_responses):
|
||||
response = await outstanding_response
|
||||
async for chunk in response:
|
||||
event = chunk.event
|
||||
finish_reason = _convert_stop_reason_to_openai_finish_reason(
|
||||
event.stop_reason
|
||||
)
|
||||
finish_reason = _convert_stop_reason_to_openai_finish_reason(event.stop_reason)
|
||||
|
||||
if isinstance(event.delta, TextDelta):
|
||||
text_delta = event.delta.text
|
||||
delta = OpenAIChoiceDelta(content=text_delta)
|
||||
yield OpenAIChatCompletionChunk(
|
||||
id=id,
|
||||
choices=[
|
||||
OpenAIChatCompletionChunkChoice(
|
||||
index=i, finish_reason=finish_reason, delta=delta
|
||||
)
|
||||
],
|
||||
choices=[OpenAIChatCompletionChunkChoice(index=i, finish_reason=finish_reason, delta=delta)],
|
||||
created=int(time.time()),
|
||||
model=model,
|
||||
object="chat.completion.chunk",
|
||||
|
|
@ -1346,9 +1322,7 @@ class OpenAIChatCompletionToLlamaStackMixin:
|
|||
yield OpenAIChatCompletionChunk(
|
||||
id=id,
|
||||
choices=[
|
||||
OpenAIChatCompletionChunkChoice(
|
||||
index=i, finish_reason=finish_reason, delta=delta
|
||||
)
|
||||
OpenAIChatCompletionChunkChoice(index=i, finish_reason=finish_reason, delta=delta)
|
||||
],
|
||||
created=int(time.time()),
|
||||
model=model,
|
||||
|
|
@ -1365,9 +1339,7 @@ class OpenAIChatCompletionToLlamaStackMixin:
|
|||
yield OpenAIChatCompletionChunk(
|
||||
id=id,
|
||||
choices=[
|
||||
OpenAIChatCompletionChunkChoice(
|
||||
index=i, finish_reason=finish_reason, delta=delta
|
||||
)
|
||||
OpenAIChatCompletionChunkChoice(index=i, finish_reason=finish_reason, delta=delta)
|
||||
],
|
||||
created=int(time.time()),
|
||||
model=model,
|
||||
|
|
@ -1382,9 +1354,7 @@ class OpenAIChatCompletionToLlamaStackMixin:
|
|||
response = await outstanding_response
|
||||
completion_message = response.completion_message
|
||||
message = await convert_message_to_openai_dict_new(completion_message)
|
||||
finish_reason = _convert_stop_reason_to_openai_finish_reason(
|
||||
completion_message.stop_reason
|
||||
)
|
||||
finish_reason = _convert_stop_reason_to_openai_finish_reason(completion_message.stop_reason)
|
||||
|
||||
choice = OpenAIChatCompletionChoice(
|
||||
index=len(choices),
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue