hyper-models 0.3.3__tar.gz → 0.3.4__tar.gz

{hyper_models-0.3.3 → hyper_models-0.3.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: hyper-models
-Version: 0.3.3
+Version: 0.3.4
 Summary: A model zoo for non-Euclidean embedding models (hyperbolic, spherical, product manifolds)
 Project-URL: Homepage, https://github.com/Hyper3Labs/hyper-models
 Project-URL: Repository, https://github.com/Hyper3Labs/hyper-models
@@ -96,7 +96,7 @@ hyper_models.list_loaders()
 # Load model (auto-downloads from Hugging Face Hub)
 model = hyper_models.load("hycoclip-vit-s")
 model.geometry  # 'hyperboloid'
-model.dim       # 513
+model.dim  # 513
 
 # Encode PIL images
 images = [Image.open("image.jpg")]
@@ -104,9 +104,9 @@ embeddings = model.encode_images(images)  # (1, 513) ndarray
 
 # Get model info
 info = hyper_models.get_model_info("hycoclip-vit-s")
-info.hub_id     # 'mnm-matin/hyperbolic-clip'
-info.loader     # 'onnx'
-info.license    # 'CC-BY-NC'
+info.hub_id  # 'mnm-matin/hyperbolic-clip'
+info.loader  # 'onnx'
+info.license  # 'CC-BY-NC'
 
 # Low-level: preprocess images yourself
 batch = hyper_models.preprocess_images(images)  # (B, 3, 224, 224)
@@ -149,10 +149,10 @@ HyperView auto-detects `hyper-models` names and routes them to the `hyper-models
 import hyperview as hv
 
 dataset = hv.Dataset.from_huggingface(
-  name="demo",
-  hf_dataset="uoft-cs/cifar10",
-  split="train",
-  image_key="img",
+    name="demo",
+    hf_dataset="uoft-cs/cifar10",
+    split="train",
+    image_key="img",
 )
 
 # Uses provider='hyper-models' automatically.

{hyper_models-0.3.3 → hyper_models-0.3.4}/README.md

@@ -55,7 +55,7 @@ hyper_models.list_loaders()
 # Load model (auto-downloads from Hugging Face Hub)
 model = hyper_models.load("hycoclip-vit-s")
 model.geometry  # 'hyperboloid'
-model.dim       # 513
+model.dim  # 513
 
 # Encode PIL images
 images = [Image.open("image.jpg")]
@@ -63,9 +63,9 @@ embeddings = model.encode_images(images)  # (1, 513) ndarray
 
 # Get model info
 info = hyper_models.get_model_info("hycoclip-vit-s")
-info.hub_id     # 'mnm-matin/hyperbolic-clip'
-info.loader     # 'onnx'
-info.license    # 'CC-BY-NC'
+info.hub_id  # 'mnm-matin/hyperbolic-clip'
+info.loader  # 'onnx'
+info.license  # 'CC-BY-NC'
 
 # Low-level: preprocess images yourself
 batch = hyper_models.preprocess_images(images)  # (B, 3, 224, 224)
@@ -108,10 +108,10 @@ HyperView auto-detects `hyper-models` names and routes them to the `hyper-models
 import hyperview as hv
 
 dataset = hv.Dataset.from_huggingface(
-  name="demo",
-  hf_dataset="uoft-cs/cifar10",
-  split="train",
-  image_key="img",
+    name="demo",
+    hf_dataset="uoft-cs/cifar10",
+    split="train",
+    image_key="img",
 )
 
 # Uses provider='hyper-models' automatically.

{hyper_models-0.3.3 → hyper_models-0.3.4}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "hyper-models"
-version = "0.3.3"
+version = "0.3.4"
 description = "A model zoo for non-Euclidean embedding models (hyperbolic, spherical, product manifolds)"
 readme = "README.md"
 license = { text = "MIT" }

{hyper_models-0.3.3 → hyper_models-0.3.4}/src/hyper_models/models.py

@@ -3,6 +3,8 @@
 from __future__ import annotations
 
 from pathlib import Path
+from shutil import copyfile
+from tempfile import TemporaryDirectory
 
 import numpy as np
 from PIL import Image
@@ -32,14 +34,26 @@ class ONNXModel:
         self._output_name = output_name
         self._image_config = image_config or ImageConfig()
         self._session = None
+        self._artifact_directory = None
 
     def _ensure_session(self) -> None:
         if self._session is None:
             import onnxruntime as ort
 
-            self._session = ort.InferenceSession(
-                str(self._path), providers=["CPUExecutionProvider"]
-            )
+            path = self._path
+            # Hub snapshots link the graph and external weights to different
+            # blob directories. New ONNX Runtime versions correctly reject
+            # external data outside the resolved graph directory. Materialize
+            # our catalog's graph and sidecars together, keeping validation on.
+            artifacts = [path, *path.parent.glob(f"{path.name}.*")]
+            if any(artifact.is_symlink() for artifact in artifacts):
+                self._artifact_directory = TemporaryDirectory(prefix="hyper-models-onnx-")
+                directory = Path(self._artifact_directory.name)
+                for artifact in artifacts:
+                    if artifact.is_file():
+                        copyfile(artifact, directory / artifact.name)
+                path = directory / path.name
+            self._session = ort.InferenceSession(str(path), providers=["CPUExecutionProvider"])
 
     def encode(self, inputs: np.ndarray) -> np.ndarray:
         """Encode preprocessed inputs (B, C, H, W) to embeddings (B, D)."""