together 1.5.35__py3-none-any.whl → 2.0.0a6__py3-none-any.whl

together/resources/models.py

@@ -1,252 +1,282 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
 from __future__ import annotations
 
-from typing import List
+from typing_extensions import Literal
+
+import httpx
 
-from together.abstract import api_requestor
-from together.together_response import TogetherResponse
-from together.types import (
-    ModelObject,
-    ModelUploadRequest,
-    ModelUploadResponse,
-    TogetherClient,
-    TogetherRequest,
+from ..types import model_upload_params
+from .._types import Body, Omit, Query, Headers, NotGiven, omit, not_given
+from .._utils import maybe_transform, async_maybe_transform
+from .._compat import cached_property
+from .._resource import SyncAPIResource, AsyncAPIResource
+from .._response import (
+    to_raw_response_wrapper,
+    to_streamed_response_wrapper,
+    async_to_raw_response_wrapper,
+    async_to_streamed_response_wrapper,
 )
+from .._base_client import make_request_options
+from ..types.model_list_response import ModelListResponse
+from ..types.model_upload_response import ModelUploadResponse
 
+__all__ = ["ModelsResource", "AsyncModelsResource"]
 
-class ModelsBase:
-    def __init__(self, client: TogetherClient) -> None:
-        self._client = client
 
-    def _filter_dedicated_models(
-        self, models: List[ModelObject], dedicated_response: TogetherResponse
-    ) -> List[ModelObject]:
+class ModelsResource(SyncAPIResource):
+    @cached_property
+    def with_raw_response(self) -> ModelsResourceWithRawResponse:
         """
-        Filter models based on dedicated model response.
+        This property can be used as a prefix for any HTTP method call to return
+        the raw response object instead of the parsed content.
 
-        Args:
-            models (List[ModelObject]): List of all models
-            dedicated_response (TogetherResponse): Response from autoscale models endpoint
-
-        Returns:
-            List[ModelObject]: Filtered list of models
+        For more information, see https://www.github.com/togethercomputer/together-py#accessing-raw-response-data-eg-headers
         """
-        assert isinstance(dedicated_response.data, list)
-
-        # Create a set of dedicated model names for efficient lookup
-        dedicated_model_names = {model["name"] for model in dedicated_response.data}
+        return ModelsResourceWithRawResponse(self)
 
-        # Filter models to only include those in dedicated_model_names
-        # Note: The model.id from ModelObject matches the name field in the autoscale response
-        return [model for model in models if model.id in dedicated_model_names]
-
-
-class Models(ModelsBase):
-    def list(
-        self,
-        dedicated: bool = False,
-    ) -> List[ModelObject]:
+    @cached_property
+    def with_streaming_response(self) -> ModelsResourceWithStreamingResponse:
         """
-        Method to return list of models on the API
+        An alternative to `.with_raw_response` that doesn't eagerly read the response body.
 
-        Args:
-            dedicated (bool, optional): If True, returns only dedicated models. Defaults to False.
-
-        Returns:
-            List[ModelObject]: List of model objects
+        For more information, see https://www.github.com/togethercomputer/together-py#with_streaming_response
         """
-        requestor = api_requestor.APIRequestor(
-            client=self._client,
-        )
+        return ModelsResourceWithStreamingResponse(self)
 
-        response, _, _ = requestor.request(
-            options=TogetherRequest(
-                method="GET",
-                url="models",
+    def list(
+        self,
+        *,
+        # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
+        # The extra values given here take precedence over values defined on the client or passed to this method.
+        extra_headers: Headers | None = None,
+        extra_query: Query | None = None,
+        extra_body: Body | None = None,
+        timeout: float | httpx.Timeout | None | NotGiven = not_given,
+    ) -> ModelListResponse:
+        """Lists all of Together's open-source models"""
+        return self._get(
+            "/models",
+            options=make_request_options(
+                extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
             ),
-            stream=False,
+            cast_to=ModelListResponse,
         )
 
-        assert isinstance(response, TogetherResponse)
-        assert isinstance(response.data, list)
-
-        models = [ModelObject(**model) for model in response.data]
-
-        if dedicated:
-            # Get dedicated models
-            dedicated_response, _, _ = requestor.request(
-                options=TogetherRequest(
-                    method="GET",
-                    url="autoscale/models",
-                ),
-                stream=False,
-            )
-
-            models = self._filter_dedicated_models(models, dedicated_response)
-
-        models.sort(key=lambda x: x.id.lower())
-
-        return models
-
     def upload(
         self,
         *,
         model_name: str,
         model_source: str,
-        model_type: str = "model",
-        hf_token: str | None = None,
-        description: str | None = None,
-        base_model: str | None = None,
-        lora_model: str | None = None,
+        base_model: str | Omit = omit,
+        description: str | Omit = omit,
+        hf_token: str | Omit = omit,
+        lora_model: str | Omit = omit,
+        model_type: Literal["model", "adapter"] | Omit = omit,
+        # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
+        # The extra values given here take precedence over values defined on the client or passed to this method.
+        extra_headers: Headers | None = None,
+        extra_query: Query | None = None,
+        extra_body: Body | None = None,
+        timeout: float | httpx.Timeout | None | NotGiven = not_given,
     ) -> ModelUploadResponse:
         """
-        Upload a custom model or adapter from Hugging Face or S3.
+        Upload a custom model or adapter from Hugging Face or S3
 
         Args:
-            model_name (str): The name to give to your uploaded model
-            model_source (str): The source location of the model (Hugging Face repo or S3 path)
-            model_type (str, optional): Whether the model is a full model or an adapter. Defaults to "model".
-            hf_token (str, optional): Hugging Face token (if uploading from Hugging Face)
-            description (str, optional): A description of your model
-            base_model (str, optional): The base model to use for an adapter if setting it to run against a serverless pool. Only used for model_type "adapter".
-            lora_model (str, optional): The lora pool to use for an adapter if setting it to run against, say, a dedicated pool. Only used for model_type "adapter".
-
-        Returns:
-            ModelUploadResponse: Object containing upload job information
-        """
-        requestor = api_requestor.APIRequestor(
-            client=self._client,
-        )
+          model_name: The name to give to your uploaded model
 
-        data = {
-            "model_name": model_name,
-            "model_source": model_source,
-            "model_type": model_type,
-        }
-
-        if hf_token is not None:
-            data["hf_token"] = hf_token
-        if description is not None:
-            data["description"] = description
-        if base_model is not None:
-            data["base_model"] = base_model
-        if lora_model is not None:
-            data["lora_model"] = lora_model
-
-        response, _, _ = requestor.request(
-            options=TogetherRequest(
-                method="POST",
-                url="models",
-                params=data,
-            ),
-            stream=False,
-        )
+          model_source: The source location of the model (Hugging Face repo or S3 path)
 
-        assert isinstance(response, TogetherResponse)
+          base_model: The base model to use for an adapter if setting it to run against a serverless
+              pool. Only used for model_type `adapter`.
 
-        return ModelUploadResponse.from_api_response(response.data)
+          description: A description of your model
 
+          hf_token: Hugging Face token (if uploading from Hugging Face)
 
-class AsyncModels(ModelsBase):
-    async def list(
-        self,
-        dedicated: bool = False,
-    ) -> List[ModelObject]:
-        """
-        Async method to return list of models on API
+          lora_model: The lora pool to use for an adapter if setting it to run against, say, a
+              dedicated pool. Only used for model_type `adapter`.
 
-        Args:
-            dedicated (bool, optional): If True, returns only dedicated models. Defaults to False.
+          model_type: Whether the model is a full model or an adapter
 
-        Returns:
-            List[ModelObject]: List of model objects
-        """
-        requestor = api_requestor.APIRequestor(
-            client=self._client,
-        )
+          extra_headers: Send extra headers
+
+          extra_query: Add additional query parameters to the request
+
+          extra_body: Add additional JSON properties to the request
 
-        response, _, _ = await requestor.arequest(
-            options=TogetherRequest(
-                method="GET",
-                url="models",
+          timeout: Override the client-level default timeout for this request, in seconds
+        """
+        return self._post(
+            "/models",
+            body=maybe_transform(
+                {
+                    "model_name": model_name,
+                    "model_source": model_source,
+                    "base_model": base_model,
+                    "description": description,
+                    "hf_token": hf_token,
+                    "lora_model": lora_model,
+                    "model_type": model_type,
+                },
+                model_upload_params.ModelUploadParams,
+            ),
+            options=make_request_options(
+                extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
             ),
-            stream=False,
+            cast_to=ModelUploadResponse,
         )
 
-        assert isinstance(response, TogetherResponse)
-        assert isinstance(response.data, list)
 
-        models = [ModelObject(**model) for model in response.data]
+class AsyncModelsResource(AsyncAPIResource):
+    @cached_property
+    def with_raw_response(self) -> AsyncModelsResourceWithRawResponse:
+        """
+        This property can be used as a prefix for any HTTP method call to return
+        the raw response object instead of the parsed content.
 
-        if dedicated:
-            # Get dedicated models
-            dedicated_response, _, _ = await requestor.arequest(
-                options=TogetherRequest(
-                    method="GET",
-                    url="autoscale/models",
-                ),
-                stream=False,
-            )
+        For more information, see https://www.github.com/togethercomputer/together-py#accessing-raw-response-data-eg-headers
+        """
+        return AsyncModelsResourceWithRawResponse(self)
 
-            models = self._filter_dedicated_models(models, dedicated_response)
+    @cached_property
+    def with_streaming_response(self) -> AsyncModelsResourceWithStreamingResponse:
+        """
+        An alternative to `.with_raw_response` that doesn't eagerly read the response body.
 
-        models.sort(key=lambda x: x.id.lower())
+        For more information, see https://www.github.com/togethercomputer/together-py#with_streaming_response
+        """
+        return AsyncModelsResourceWithStreamingResponse(self)
 
-        return models
+    async def list(
+        self,
+        *,
+        # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
+        # The extra values given here take precedence over values defined on the client or passed to this method.
+        extra_headers: Headers | None = None,
+        extra_query: Query | None = None,
+        extra_body: Body | None = None,
+        timeout: float | httpx.Timeout | None | NotGiven = not_given,
+    ) -> ModelListResponse:
+        """Lists all of Together's open-source models"""
+        return await self._get(
+            "/models",
+            options=make_request_options(
+                extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
+            ),
+            cast_to=ModelListResponse,
+        )
 
     async def upload(
         self,
         *,
         model_name: str,
         model_source: str,
-        model_type: str = "model",
-        hf_token: str | None = None,
-        description: str | None = None,
-        base_model: str | None = None,
-        lora_model: str | None = None,
+        base_model: str | Omit = omit,
+        description: str | Omit = omit,
+        hf_token: str | Omit = omit,
+        lora_model: str | Omit = omit,
+        model_type: Literal["model", "adapter"] | Omit = omit,
+        # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
+        # The extra values given here take precedence over values defined on the client or passed to this method.
+        extra_headers: Headers | None = None,
+        extra_query: Query | None = None,
+        extra_body: Body | None = None,
+        timeout: float | httpx.Timeout | None | NotGiven = not_given,
     ) -> ModelUploadResponse:
         """
-        Upload a custom model or adapter from Hugging Face or S3.
+        Upload a custom model or adapter from Hugging Face or S3
 
         Args:
-            model_name (str): The name to give to your uploaded model
-            model_source (str): The source location of the model (Hugging Face repo or S3 path)
-            model_type (str, optional): Whether the model is a full model or an adapter. Defaults to "model".
-            hf_token (str, optional): Hugging Face token (if uploading from Hugging Face)
-            description (str, optional): A description of your model
-            base_model (str, optional): The base model to use for an adapter if setting it to run against a serverless pool. Only used for model_type "adapter".
-            lora_model (str, optional): The lora pool to use for an adapter if setting it to run against, say, a dedicated pool. Only used for model_type "adapter".
-
-        Returns:
-            ModelUploadResponse: Object containing upload job information
+          model_name: The name to give to your uploaded model
+
+          model_source: The source location of the model (Hugging Face repo or S3 path)
+
+          base_model: The base model to use for an adapter if setting it to run against a serverless
+              pool. Only used for model_type `adapter`.
+
+          description: A description of your model
+
+          hf_token: Hugging Face token (if uploading from Hugging Face)
+
+          lora_model: The lora pool to use for an adapter if setting it to run against, say, a
+              dedicated pool. Only used for model_type `adapter`.
+
+          model_type: Whether the model is a full model or an adapter
+
+          extra_headers: Send extra headers
+
+          extra_query: Add additional query parameters to the request
+
+          extra_body: Add additional JSON properties to the request
+
+          timeout: Override the client-level default timeout for this request, in seconds
         """
-        requestor = api_requestor.APIRequestor(
-            client=self._client,
+        return await self._post(
+            "/models",
+            body=await async_maybe_transform(
+                {
+                    "model_name": model_name,
+                    "model_source": model_source,
+                    "base_model": base_model,
+                    "description": description,
+                    "hf_token": hf_token,
+                    "lora_model": lora_model,
+                    "model_type": model_type,
+                },
+                model_upload_params.ModelUploadParams,
+            ),
+            options=make_request_options(
+                extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
+            ),
+            cast_to=ModelUploadResponse,
         )
 
-        data = {
-            "model_name": model_name,
-            "model_source": model_source,
-            "model_type": model_type,
-        }
-
-        if hf_token is not None:
-            data["hf_token"] = hf_token
-        if description is not None:
-            data["description"] = description
-        if base_model is not None:
-            data["base_model"] = base_model
-        if lora_model is not None:
-            data["lora_model"] = lora_model
-
-        response, _, _ = await requestor.arequest(
-            options=TogetherRequest(
-                method="POST",
-                url="models",
-                params=data,
-            ),
-            stream=False,
+
+class ModelsResourceWithRawResponse:
+    def __init__(self, models: ModelsResource) -> None:
+        self._models = models
+
+        self.list = to_raw_response_wrapper(
+            models.list,
+        )
+        self.upload = to_raw_response_wrapper(
+            models.upload,
+        )
+
+
+class AsyncModelsResourceWithRawResponse:
+    def __init__(self, models: AsyncModelsResource) -> None:
+        self._models = models
+
+        self.list = async_to_raw_response_wrapper(
+            models.list,
         )
+        self.upload = async_to_raw_response_wrapper(
+            models.upload,
+        )
+
+
+class ModelsResourceWithStreamingResponse:
+    def __init__(self, models: ModelsResource) -> None:
+        self._models = models
 
-        assert isinstance(response, TogetherResponse)
+        self.list = to_streamed_response_wrapper(
+            models.list,
+        )
+        self.upload = to_streamed_response_wrapper(
+            models.upload,
+        )
 
-        return ModelUploadResponse.from_api_response(response.data)
+
+class AsyncModelsResourceWithStreamingResponse:
+    def __init__(self, models: AsyncModelsResource) -> None:
+        self._models = models
+
+        self.list = async_to_streamed_response_wrapper(
+            models.list,
+        )
+        self.upload = async_to_streamed_response_wrapper(
+            models.upload,
+        )