diff --git a/llama_stack/distribution/templates/local-sambanova-build.yaml b/llama_stack/distribution/templates/local-sambanova-build.yaml new file mode 100644 index 000000000..ea0bb1766 --- /dev/null +++ b/llama_stack/distribution/templates/local-sambanova-build.yaml @@ -0,0 +1,10 @@ +name: local-sambanova +distribution_spec: + description: Use SambaNova Cloud API for running LLM inference + providers: + inference: remote::sambanova + memory: meta-reference + safety: meta-reference + agents: meta-reference + telemetry: meta-reference +image_type: conda diff --git a/llama_stack/providers/adapters/inference/sambanova/__init__.py b/llama_stack/providers/adapters/inference/sambanova/__init__.py new file mode 100644 index 000000000..bd9747480 --- /dev/null +++ b/llama_stack/providers/adapters/inference/sambanova/__init__.py @@ -0,0 +1,16 @@ +# 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 .config import SambaNovaImplConfig +from .sambanova import SambaNovaInferenceAdapter + +async def get_adapter_impl(config: SambaNovaImplConfig, _deps): + assert isinstance( + config, SambaNovaImplConfig + ), f"Unexpected config type: {type(config)}" + impl = SambaNovaInferenceAdapter(config) + await impl.initialize() + return impl diff --git a/llama_stack/providers/adapters/inference/sambanova/config.py b/llama_stack/providers/adapters/inference/sambanova/config.py new file mode 100644 index 000000000..e2db47dbb --- /dev/null +++ b/llama_stack/providers/adapters/inference/sambanova/config.py @@ -0,0 +1,20 @@ +# 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 llama_models.schema_utils import json_schema_type +from pydantic import BaseModel, Field + + +@json_schema_type +class SambaNovaImplConfig(BaseModel): + url: str = Field( + default="https://api.sambanova.ai/v1", + description="The URL for the SambaNova Cloud AI server", + ) + api_key: str = Field( + default="", + description="The SambaNova AI API Key", + ) diff --git a/llama_stack/providers/adapters/inference/sambanova/sambanova.py b/llama_stack/providers/adapters/inference/sambanova/sambanova.py new file mode 100644 index 000000000..cf4e00a44 --- /dev/null +++ b/llama_stack/providers/adapters/inference/sambanova/sambanova.py @@ -0,0 +1,254 @@ +# 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 typing import AsyncGenerator + +import openai +from openai import OpenAI + +from llama_models.llama3.api.chat_format import ChatFormat + +from llama_models.llama3.api.datatypes import Message, StopReason +from llama_models.llama3.api.tokenizer import Tokenizer +from llama_models.sku_list import resolve_model + +from llama_stack.apis.inference import * # noqa: F403 +from llama_stack.providers.utils.inference.augment_messages import ( + augment_messages_for_tools, +) + +from .config import SambaNovaImplConfig + +SAMBANOVA_SUPPORTED_MODELS = { + "Llama3.1-8B-Instruct": "Meta-Llama-3.1-8B-Instruct", + "Llama3.1-70B-Instruct": "Meta-Llama-3.1-70B-Instruct", + "Llama3.1-405B-Instruct": "Meta-Llama-3.1-405B-Instruct", +} + + +class SambaNovaInferenceAdapter(Inference): + def __init__(self, config: SambaNovaImplConfig) -> None: + self.config = config + tokenizer = Tokenizer.get_instance() + self.formatter = ChatFormat(tokenizer) + + @property + def client(self) -> OpenAI: + return OpenAI( + base_url=self.config.url, + api_key=self.config.api_key + ) + + async def initialize(self) -> None: + return + + async def shutdown(self) -> None: + pass + + async def completion(self, request: CompletionRequest) -> AsyncGenerator: + raise NotImplementedError() + + def _messages_to_sambanova_messages(self, messages: list[Message]) -> list: + sambanova_messages = [] + for message in messages: + if message.role == "ipython": + role = "tool" + else: + role = message.role + sambanova_messages.append({"role": role, "content": message.content}) + + return sambanova_messages + + def resolve_sambanova_model(self, model_name: str) -> str: + model = resolve_model(model_name) + assert ( + model is not None + and model.descriptor(shorten_default_variant=True) + in SAMBANOVA_SUPPORTED_MODELS + ), f"Unsupported model: {model_name}, use one of the supported models: {','.join(SAMBANOVA_SUPPORTED_MODELS.keys())}" + + return SAMBANOVA_SUPPORTED_MODELS.get( + model.descriptor(shorten_default_variant=True) + ) + + def get_sambanova_chat_options(self, request: ChatCompletionRequest) -> dict: + options = {} + if request.sampling_params is not None: + for attr in {"temperature", "top_p", "top_k", "max_tokens"}: + if getattr(request.sampling_params, attr): + options[attr] = getattr(request.sampling_params, attr) + + return options + + async def chat_completion( + self, + model: str, + messages: List[Message], + sampling_params: Optional[SamplingParams] = SamplingParams(), + tools: Optional[List[ToolDefinition]] = None, + tool_choice: Optional[ToolChoice] = ToolChoice.auto, + tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json, + stream: Optional[bool] = False, + logprobs: Optional[LogProbConfig] = None, + ) -> AsyncGenerator: + request = ChatCompletionRequest( + model=model, + messages=messages, + sampling_params=sampling_params, + tools=tools or [], + tool_choice=tool_choice, + tool_prompt_format=tool_prompt_format, + stream=stream, + logprobs=logprobs, + ) + + messages = augment_messages_for_tools(request) + options = self.get_sambanova_chat_options(request) + sambanova_model = self.resolve_sambanova_model(request.model) + + if not request.stream: + + r = self.client.chat.completions.create( + model=sambanova_model, + messages=self._messages_to_sambanova_messages(messages), + stream=False, + **options, + ) + + stop_reason = None + if r.choices[0].finish_reason: + if r.choices[0].finish_reason == "stop": + stop_reason = StopReason.end_of_turn + elif r.choices[0].finish_reason == "length": + stop_reason = StopReason.out_of_tokens + + completion_message = self.formatter.decode_assistant_message_from_content( + r.choices[0].message.content, stop_reason + ) + yield ChatCompletionResponse( + completion_message=completion_message, + logprobs=None, + ) + else: + yield ChatCompletionResponseStreamChunk( + event=ChatCompletionResponseEvent( + event_type=ChatCompletionResponseEventType.start, + delta="", + ) + ) + + buffer = "" + ipython = False + stop_reason = None + + for chunk in self.client.chat.completions.create( + model=sambanova_model, + messages=self._messages_to_sambanova_messages(messages), + stream=True, + **options, + ): + if chunk.choices[0].finish_reason: + if ( + stop_reason is None + and chunk.choices[0].finish_reason == "stop" + ): + stop_reason = StopReason.end_of_turn + elif ( + stop_reason is None + and chunk.choices[0].finish_reason == "length" + ): + stop_reason = StopReason.out_of_tokens + break + + text = chunk.choices[0].delta.content + + if text is None: + continue + + # check if its a tool call ( aka starts with <|python_tag|> ) + if not ipython and text.startswith("<|python_tag|>"): + ipython = True + yield ChatCompletionResponseStreamChunk( + event=ChatCompletionResponseEvent( + event_type=ChatCompletionResponseEventType.progress, + delta=ToolCallDelta( + content="", + parse_status=ToolCallParseStatus.started, + ), + ) + ) + buffer += text + continue + + if ipython: + if text == "<|eot_id|>": + stop_reason = StopReason.end_of_turn + text = "" + continue + elif text == "<|eom_id|>": + stop_reason = StopReason.end_of_message + text = "" + continue + + buffer += text + delta = ToolCallDelta( + content=text, + parse_status=ToolCallParseStatus.in_progress, + ) + + yield ChatCompletionResponseStreamChunk( + event=ChatCompletionResponseEvent( + event_type=ChatCompletionResponseEventType.progress, + delta=delta, + stop_reason=stop_reason, + ) + ) + else: + buffer += text + yield ChatCompletionResponseStreamChunk( + event=ChatCompletionResponseEvent( + event_type=ChatCompletionResponseEventType.progress, + delta=text, + stop_reason=stop_reason, + ) + ) + + # parse tool calls and report errors + message = self.formatter.decode_assistant_message_from_content( + buffer, stop_reason + ) + parsed_tool_calls = len(message.tool_calls) > 0 + if ipython and not parsed_tool_calls: + yield ChatCompletionResponseStreamChunk( + event=ChatCompletionResponseEvent( + event_type=ChatCompletionResponseEventType.progress, + delta=ToolCallDelta( + content="", + parse_status=ToolCallParseStatus.failure, + ), + stop_reason=stop_reason, + ) + ) + + for tool_call in message.tool_calls: + yield ChatCompletionResponseStreamChunk( + event=ChatCompletionResponseEvent( + event_type=ChatCompletionResponseEventType.progress, + delta=ToolCallDelta( + content=tool_call, + parse_status=ToolCallParseStatus.success, + ), + stop_reason=stop_reason, + ) + ) + + yield ChatCompletionResponseStreamChunk( + event=ChatCompletionResponseEvent( + event_type=ChatCompletionResponseEventType.complete, + delta="", + stop_reason=stop_reason, + ) + ) diff --git a/llama_stack/providers/registry/inference.py b/llama_stack/providers/registry/inference.py index 31b3e2c2d..e60f37a82 100644 --- a/llama_stack/providers/registry/inference.py +++ b/llama_stack/providers/registry/inference.py @@ -94,4 +94,15 @@ def available_providers() -> List[ProviderSpec]: header_extractor_class="llama_stack.providers.adapters.inference.together.TogetherHeaderExtractor", ), ), + remote_provider_spec( + api=Api.inference, + adapter=AdapterSpec( + adapter_id="sambanova", + pip_packages=[ + "openai", + ], + module="llama_stack.providers.adapters.inference.sambanova", + config_class="llama_stack.providers.adapters.inference.sambanova.SambaNovaImplConfig", + ), + ), ]