Upload folder using huggingface_hub

README.md

@@ -0,0 +1,74 @@
+---
+license: agpl-3.0
+library_name: pytorch
+base_model:
+- Qwen/Qwen3-VL-4B-Instruct 
+tags:
+- rgcn
+- embedding
+- onnx
+---
+
+# EduGraph Embed
+
+This model generates embeddings for labels from the
+[EduGraph Ontology](httpss://github.com/christian-bick/edugraph-ontology).
+
+When combined with an [EduGraph Classification Model](httpss://github.com/christian-bick/edugraph-classify-qwen3vl),
+we can determine similarity between any type of learning content covered by the EduGraph ontology. 
+For example, in tandem, the two models can determine whether some content of a math learning app 
+trains the exact same set of skills tested in a paper quiz, by providing nothing else than a screenshot 
+and a photo.
+
+## How it works
+
+The model determines similarity based on the *structure* of the EduGraph Ontology. It respects
+various types of entity relationships to determine similarity, most importantly, parent-child and sibling 
+relationships within the graph in addition to the semantic similarity of their definitions.
+
+For example, the model will reliably place labels like `IntegerAddition` and `FractionAddition` 
+closer together than, say, `ShapeIdentification`.
+
+To accomplish this, the model generates knowledge graph embeddings that
+map the ontology structure into a high-dimensional vector space using a 
+[Relational Graph Convolutional Network (R-GCN)](httpss://arxiv.org/abs/1703.06103).
+
+## Limitations
+
+This model is centered around the EduGraph ontology. The embedding model was trained 
+on the entities and relationships in this ontology. Consequently, it can only embed 
+labels that are defined as entities within this ontology.
+
+## Risks
+
+**Important:** Currently this model is in a research status and has not been evaluated under real-world conditions. 
+
+* **ONLY use this model for research, experimentation and evaluation**
+* **Do NOT use in a classroom environment**
+* **Do NOT use for automations that might impact children**
+
+## Using the Model
+
+### Preparation
+
+1) Download the following files:
+
+- `embed_entities_biased.onnx`
+- `embed_entities.pt`
+
+2) Install the following dependencies:
+
+- `torch`
+- `numpy`
+- `onnxruntime`
+
+### Reference Example
+
+See [entity_embeddings_infer.py](httpss://github.com/christian-bick/edugraph-embed/blob/master/src/edugraph/embed/entity_embeddings_infer.py)
+for reference usage.
+
+## License
+
+This project is licensed under the GNU Affero General Public License. See the [LICENSE](LICENSE) file for details.
+
+If these license terms are not working for you, then contact us, and we can discuss alternative options.

embed_entities.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55b10aa169320b084cfeb83439071a456b1d923681a9a1a5bf3f1ea5ea200fec
+size 516195

embed_entities_biased.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29a346b61acc03373bf9680f7882ab70bac2440db633cf8577534413e3f930e0
+size 2471425

embed_entities_neutral.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cb74338767dacf2f11f46bddbf6cd22106e83312140f2bc5262bc7c5aa1a668
+size 2470723

embed_entities_text.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:293d209283ac9e0faf4a5f846c566071bd4ff12af1162d7c8bd52120b3bb1848
+size 516373