ray-embedding 0.11.5__tar.gz → 0.11.7__tar.gz

{ray_embedding-0.11.5 → ray_embedding-0.11.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ray-embedding
-Version: 0.11.5
+Version: 0.11.7
 Summary: Deploy SentenceTransformers embedding models to a ray cluster
 Author: Crispin Almodovar
 Author-email: 

{ray_embedding-0.11.5 → ray_embedding-0.11.7}/ray_embedding/deploy.py

@@ -6,12 +6,13 @@ from ray_embedding.dto import AppConfig, ModelDeploymentConfig
 from ray_embedding.embedding_model import EmbeddingModel
 import torch
 
-from ray_embedding.embedding_service import EmbeddingService
+from ray_embedding.model_router import ModelRouter
 
 
 def build_model(model_config: ModelDeploymentConfig) -> DeploymentHandle:
     deployment_name = model_config.deployment
     model = model_config.model
+    served_model_name = model_config.served_model_name
     device = model_config.device
     backend = model_config.backend or "torch"
     matryoshka_dim = model_config.matryoshka_dim
@@ -29,6 +30,7 @@ def build_model(model_config: ModelDeploymentConfig) -> DeploymentHandle:
             raise ValueError(f"Invalid torch_dtype: '{torch_dtype}'")
 
     deployment = EmbeddingModel.options(name=deployment_name).bind(model=model,
+                                                                   served_model_name=served_model_name,
                                                                    device=device,
                                                                    backend=backend,
                                                                    matryoshka_dim=matryoshka_dim,
@@ -41,6 +43,6 @@ def build_app(args: AppConfig) -> Application:
     model_router, models = args.model_router, args.models
     assert model_router and models
 
-    served_models = {model_config.model: build_model(model_config) for model_config in models}
-    app = EmbeddingService.options(name=model_router.deployment).bind(served_models)
-    return app
+    served_models = {model_config.served_model_name: build_model(model_config) for model_config in models}
+    router = ModelRouter.options(name=model_router.deployment).bind(served_models)
+    return router

{ray_embedding-0.11.5 → ray_embedding-0.11.7}/ray_embedding/embedding_model.py

@@ -18,6 +18,7 @@ class EmbeddingModel:
         logging.basicConfig(level=logging.INFO)
         self.logger = logging.getLogger(self.__class__.__name__)
         self.model = model
+        self.served_model_name = served_model_name or os.path.basename(self.model)
         self.init_device = device
         if self.init_device is None or self.init_device == "auto":
             self.init_device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -33,7 +34,6 @@ class EmbeddingModel:
                                                    trust_remote_code=self.trust_remote_code,
                                                    model_kwargs=self.model_kwargs)
 
-        self.served_model_name = served_model_name or os.path.basename(self.model)
         self.logger.info(f"Successfully initialized embedding model {self.model} using device {self.torch_device}")
 
     async def __call__(self, text: Union[str, List[str]], dimensions: Optional[int] = None):

ray_embedding-0.11.5/ray_embedding/embedding_service.py → ray_embedding-0.11.7/ray_embedding/model_router.py

@@ -12,7 +12,7 @@ web_api = FastAPI(title="Ray Embeddings - OpenAI-compatible API")
 
 @serve.deployment
 @serve.ingress(web_api)
-class EmbeddingService:
+class ModelRouter:
     def __init__(self, served_models: Dict[str, DeployedModel]):
         assert served_models, "models cannot be empty"
         logging.basicConfig(level=logging.INFO)
@@ -29,6 +29,7 @@ class EmbeddingService:
 
     async def _compute_embeddings_from_resized_batches(self, model: str, inputs: List[str], dimensions: Optional[int] = None):
         assert model in self.served_models
+        self.logger.info(f"Retrieving model handle for: {model} ")
         model_handle = self.served_models[model].deployment_handle
         self.logger.info(f"Model handle: {model_handle} ")
         batch_size = self.served_models[model].batch_size
@@ -67,7 +68,7 @@ class EmbeddingService:
     async def compute_embeddings(self, request: EmbeddingRequest):
         try:
             inputs = request.input if isinstance(request.input, list) else [request.input]
-            self.logger.info(f"Received input of size {len(inputs)} text chunks")
+            self.logger.info(f"Received input of size {len(inputs)} text chunks; model: {request.model}")
             embeddings = await self._compute_embeddings_from_resized_batches(request.model, inputs, request.dimensions)
             response_data = [
                 {"index": idx, "embedding": emb}

{ray_embedding-0.11.5 → ray_embedding-0.11.7}/ray_embedding.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ray-embedding
-Version: 0.11.5
+Version: 0.11.7
 Summary: Deploy SentenceTransformers embedding models to a ray cluster
 Author: Crispin Almodovar
 Author-email: 

{ray_embedding-0.11.5 → ray_embedding-0.11.7}/ray_embedding.egg-info/SOURCES.txt

@@ -5,7 +5,7 @@ ray_embedding/__init__.py
 ray_embedding/deploy.py
 ray_embedding/dto.py
 ray_embedding/embedding_model.py
-ray_embedding/embedding_service.py
+ray_embedding/model_router.py
 ray_embedding.egg-info/PKG-INFO
 ray_embedding.egg-info/SOURCES.txt
 ray_embedding.egg-info/dependency_links.txt

{ray_embedding-0.11.5 → ray_embedding-0.11.7}/setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = ray-embedding
-version = 0.11.5
+version = 0.11.7
 author = Crispin Almodovar
 author_email = 
 description = Deploy SentenceTransformers embedding models to a ray cluster