added embedding function for the Instructor models.

chromadb/utils/embedding_functions.py

@@ -76,3 +76,19 @@ def __init__(self, api_key: str, model_name: str = "sentence-transformers/all-Mi
     def __call__(self, texts: Documents) -> Embeddings:
         # Call HuggingFace Embedding API for each document
         return self._session.post(self._api_url, json={"inputs": texts, "options":{"wait_for_model":True}}).json()
+
+
+class InstructorEmbeddingFunction(EmbeddingFunction):
+    # If you have a GPU with at least 6GB try model_name = "hkunlp/instructor-xl" and device = "cuda"
+    # for a full list of options: https://github.com/HKUNLP/instructor-embedding#model-list
+    def __init__(self, model_name: str = "hkunlp/instructor-base", device= "cpu"): 
+        try:
+            from InstructorEmbedding import INSTRUCTOR
+        except ImportError:
+            raise ValueError(
+                "The InstructorEmbedding python package is not installed. Please install it with `pip install InstructorEmbedding`"
+            )
+        self._model = INSTRUCTOR(model_name, device=device)
+
+    def __call__(self, texts: Documents) -> Embeddings:
+        return self._model.encode(texts).tolist()