diff --git a/llama_stack/providers/remote/inference/runpod/runpod.py b/llama_stack/providers/remote/inference/runpod/runpod.py index 82252b04d..3294d8eab 100644 --- a/llama_stack/providers/remote/inference/runpod/runpod.py +++ b/llama_stack/providers/remote/inference/runpod/runpod.py @@ -3,52 +3,29 @@ # # 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 AsyncGenerator - from openai import OpenAI - -from llama_stack.apis.inference import * # noqa: F403 +from llama_stack.apis.inference import * from llama_stack.apis.inference import OpenAIEmbeddingsResponse - -# from llama_stack.providers.datatypes import ModelsProtocolPrivate -from llama_stack.providers.utils.inference.model_registry import ModelRegistryHelper, build_hf_repo_model_entry +from llama_stack.apis.models import Model, ModelType +from llama_stack.providers.utils.inference.model_registry import ModelRegistryHelper from llama_stack.providers.utils.inference.openai_compat import ( OpenAIChatCompletionToLlamaStackMixin, OpenAICompletionToLlamaStackMixin, get_sampling_options, process_chat_completion_response, process_chat_completion_stream_response, + process_completion_response, + process_completion_stream_response, ) from llama_stack.providers.utils.inference.prompt_adapter import ( - chat_completion_request_to_prompt, + completion_request_to_prompt, + interleaved_content_as_str, ) - from .config import RunpodImplConfig -# https://docs.runpod.io/serverless/vllm/overview#compatible-models -# https://github.com/runpod-workers/worker-vllm/blob/main/README.md#compatible-model-architectures -RUNPOD_SUPPORTED_MODELS = { - "Llama3.1-8B": "meta-llama/Llama-3.1-8B", - "Llama3.1-70B": "meta-llama/Llama-3.1-70B", - "Llama3.1-405B:bf16-mp8": "meta-llama/Llama-3.1-405B", - "Llama3.1-405B": "meta-llama/Llama-3.1-405B-FP8", - "Llama3.1-405B:bf16-mp16": "meta-llama/Llama-3.1-405B", - "Llama3.1-8B-Instruct": "meta-llama/Llama-3.1-8B-Instruct", - "Llama3.1-70B-Instruct": "meta-llama/Llama-3.1-70B-Instruct", - "Llama3.1-405B-Instruct:bf16-mp8": "meta-llama/Llama-3.1-405B-Instruct", - "Llama3.1-405B-Instruct": "meta-llama/Llama-3.1-405B-Instruct-FP8", - "Llama3.1-405B-Instruct:bf16-mp16": "meta-llama/Llama-3.1-405B-Instruct", - "Llama3.2-1B": "meta-llama/Llama-3.2-1B", - "Llama3.2-3B": "meta-llama/Llama-3.2-3B", -} - -SAFETY_MODELS_ENTRIES = [] - -# Create MODEL_ENTRIES from RUNPOD_SUPPORTED_MODELS for compatibility with starter template -MODEL_ENTRIES = [ - build_hf_repo_model_entry(provider_model_id, model_descriptor) - for provider_model_id, model_descriptor in RUNPOD_SUPPORTED_MODELS.items() -] + SAFETY_MODELS_ENTRIES +MODEL_ENTRIES = [] class RunpodInferenceAdapter( @@ -57,30 +34,83 @@ class RunpodInferenceAdapter( OpenAIChatCompletionToLlamaStackMixin, OpenAICompletionToLlamaStackMixin, ): + """ + Adapter for RunPod's OpenAI-compatible API endpoints. + Supports VLLM for serverless endpoint self-hosted or public endpoints. + Can work with any runpod endpoints that support OpenAI-compatible API + """ + def __init__(self, config: RunpodImplConfig) -> None: - ModelRegistryHelper.__init__(self, stack_to_provider_models_map=RUNPOD_SUPPORTED_MODELS) + ModelRegistryHelper.__init__(self, MODEL_ENTRIES) self.config = config async def initialize(self) -> None: - return + pass async def shutdown(self) -> None: pass + async def register_model(self, model: Model) -> Model: + """ + Register any model with the runpod provider_id. + + Pass-through registration - accepts any model string that the RunPod endpoint serves. + No static model validation since RunPod endpoints can serve arbitrary vLLM models. + + YAML Configuration Example: + models: + - metadata: {} + model_id: runpod/qwen/qwen3-8b + model_type: llm + provider_id: runpod + provider_model_id: qwen/qwen3-8b + - metadata: {} + model_id: runpod/deepcogito/cogito-v2-preview-llama-70B + model_type: llm + provider_id: runpod + provider_model_id: deepcogito/cogito-v2-preview-llama-70B + + The provider strips 'runpod/' prefix before API calls: + "runpod/qwen/qwen3-8b" -> "qwen/qwen3-8b" + """ + if model.provider_id == "runpod": + logger.info( + f"Registering model: {model.identifier} -> {model.provider_resource_id}" + ) + return model + return await super().register_model(model) + async def completion( self, - model: str, + model_id: str, content: InterleavedContent, sampling_params: SamplingParams | None = None, response_format: ResponseFormat | None = None, stream: bool | None = False, logprobs: LogProbConfig | None = None, ) -> AsyncGenerator: - raise NotImplementedError() + if sampling_params is None: + sampling_params = SamplingParams() + + request = CompletionRequest( + model=model_id, + content=content, + sampling_params=sampling_params, + response_format=response_format, + stream=stream, + logprobs=logprobs, + ) + + client = OpenAI(base_url=self.config.url, api_key=self.config.api_token) + + if stream: + return self._stream_completion(request, client) + else: + return await self._nonstream_completion(request, client) async def chat_completion( self, - model: str, + model_id: str, messages: list[Message], sampling_params: SamplingParams | None = None, response_format: ResponseFormat | None = None, @@ -91,10 +121,12 @@ class RunpodInferenceAdapter( logprobs: LogProbConfig | None = None, tool_config: ToolConfig | None = None, ) -> AsyncGenerator: + """Process chat completion requests using RunPod's OpenAI-compatible API.""" if sampling_params is None: sampling_params = SamplingParams() + request = ChatCompletionRequest( - model=model, + model=model_id, messages=messages, sampling_params=sampling_params, tools=tools or [], @@ -104,6 +136,7 @@ class RunpodInferenceAdapter( ) client = OpenAI(base_url=self.config.url, api_key=self.config.api_token) + if stream: return self._stream_chat_completion(request, client) else: @@ -112,15 +145,17 @@ class RunpodInferenceAdapter( async def _nonstream_chat_completion( self, request: ChatCompletionRequest, client: OpenAI ) -> ChatCompletionResponse: - params = self._get_params(request) - r = client.completions.create(**params) + params = await self._get_chat_params(request) + r = client.chat.completions.create(**params) return process_chat_completion_response(r, request) - async def _stream_chat_completion(self, request: ChatCompletionRequest, client: OpenAI) -> AsyncGenerator: - params = self._get_params(request) + async def _stream_chat_completion( + self, request: ChatCompletionRequest, client: OpenAI + ) -> AsyncGenerator: + params = await self._get_chat_params(request) async def _to_async_generator(): - s = client.completions.create(**params) + s = client.chat.completions.create(**params) for chunk in s: yield chunk @@ -128,14 +163,102 @@ class RunpodInferenceAdapter( async for chunk in process_chat_completion_stream_response(stream, request): yield chunk - def _get_params(self, request: ChatCompletionRequest) -> dict: - return { - "model": self.map_to_provider_model(request.model), - "prompt": chat_completion_request_to_prompt(request), + async def _get_chat_params(self, request: ChatCompletionRequest) -> dict: + """Convert Llama Stack request to RunPod API parameters.""" + messages = [ + {"role": msg.role, "content": msg.content} for msg in request.messages + ] + + # Resolve model_id to provider_resource_id + model_obj = await self.model_store.get_model(request.model) + model = model_obj.provider_resource_id or request.model + + if model.startswith("runpod/"): + model = model.replace("runpod/", "", 1) + + params = { + "model": model, + "messages": messages, "stream": request.stream, **get_sampling_options(request.sampling_params), } + if request.stream: + params["stream_options"] = {"include_usage": True} + + return params + + async def _nonstream_completion( + self, request: CompletionRequest, client: OpenAI + ) -> CompletionResponse: + params = await self._get_completion_params(request) + r = client.completions.create(**params) + return process_completion_response(r) + + async def _stream_completion( + self, request: CompletionRequest, client: OpenAI + ) -> AsyncGenerator: + params = await self._get_completion_params(request) + + async def _to_async_generator(): + s = client.completions.create(**params) + for chunk in s: + yield chunk + + stream = _to_async_generator() + async for chunk in process_completion_stream_response(stream): + yield chunk + + async def _get_completion_params(self, request: CompletionRequest) -> dict: + # Resolve model_id to provider_resource_id + model_obj = await self.model_store.get_model(request.model) + model = model_obj.provider_resource_id or request.model + + if model.startswith("runpod/"): + model = model.replace("runpod/", "", 1) + + params = { + "model": model, + "prompt": completion_request_to_prompt(request), + "stream": request.stream, + **get_sampling_options(request.sampling_params), + } + + if request.stream: + params["stream_options"] = {"include_usage": True} + + return params + + async def embeddings( + self, + model_id: str, + contents: list[str] | list[InterleavedContentItem], + text_truncation: TextTruncation | None = TextTruncation.none, + output_dimension: int | None = None, + task_type: EmbeddingTaskType | None = None, + ) -> EmbeddingsResponse: + # Resolve model_id to provider_resource_id + model_obj = await self.model_store.get_model(model_id) + model = model_obj.provider_resource_id or model_id + + if model.startswith("runpod/"): + model = model.replace("runpod/", "", 1) + + client = OpenAI(base_url=self.config.url, api_key=self.config.api_token) + + kwargs = {} + if output_dimension: + kwargs["dimensions"] = output_dimension + + response = client.embeddings.create( + model=model, + input=[interleaved_content_as_str(content) for content in contents], + **kwargs, + ) + + embeddings = [data.embedding for data in response.data] + return EmbeddingsResponse(embeddings=embeddings) + async def openai_embeddings( self, model: str, @@ -144,4 +267,21 @@ class RunpodInferenceAdapter( dimensions: int | None = None, user: str | None = None, ) -> OpenAIEmbeddingsResponse: - raise NotImplementedError() + # Resolve model_id to provider_resource_id + model_obj = await self.model_store.get_model(model) + model_stripped = model_obj.provider_resource_id or model + + if model_stripped.startswith("runpod/"): + model_stripped = model_stripped.replace("runpod/", "", 1) + + client = OpenAI(base_url=self.config.url, api_key=self.config.api_token) + + response = client.embeddings.create( + model=model_stripped, + input=input, + encoding_format=encoding_format, + dimensions=dimensions, + user=user, + ) + + return response \ No newline at end of file