community[minor]: Add support for modle2vec embeddings (#28507)

This PR add an embeddings integration for model2vec, the
`Model2vecEmbeddings` class.

- **Description**: [Model2Vec](https://github.com/MinishLab/model2vec)
lets you turn any sentence transformer into a really small static model
and makes running the model faster.
- **Issue**:
- **Dependencies**: model2vec
([pypi](https://pypi.org/project/model2vec/))
- **Twitter handle:**:

- [x] **Add tests and docs**: 
-
[Test](https://github.com/blacksmithop/langchain/blob/model2vec_embeddings/libs/community/langchain_community/embeddings/model2vec.py),
[docs](https://github.com/blacksmithop/langchain/blob/model2vec_embeddings/docs/docs/integrations/text_embedding/model2vec.ipynb)

- [x] **Lint and test**:

---------

Co-authored-by: Abhinav KM <abhinav.m@zerone-consulting.com>
Co-authored-by: Bagatur <baskaryan@gmail.com>

docs/docs/integrations/text_embedding/model2vec.ipynb

@@ -0,0 +1,201 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "e8712110",
+   "metadata": {},
+   "source": [
+    "## Overview\n",
+    "\n",
+    "Model2Vec is a technique to turn any sentence transformer into a really small static model\n",
+    "[model2vec](https://github.com/MinishLab/model2vec) can be used to generate embeddings."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "266dd424",
+   "metadata": {},
+   "source": [
+    "## Setup\n",
+    "\n",
+    "```bash\n",
+    "pip install -U langchain-community\n",
+    "```\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "78ab91a6",
+   "metadata": {},
+   "source": [
+    "## Instantiation"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d06e7719",
+   "metadata": {},
+   "source": [
+    "Ensure that `model2vec` is installed\n",
+    "\n",
+    "```bash\n",
+    "pip install -U model2vec\n",
+    "```"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f8ea1ed5",
+   "metadata": {},
+   "source": [
+    "## Indexing and Retrieval"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "d25dc22d-b656-46c6-a42d-eace958590cd",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-05-24T15:13:17.176956Z",
+     "start_time": "2023-05-24T15:13:15.399076Z"
+    },
+    "execution": {
+     "iopub.execute_input": "2024-03-29T15:39:19.252281Z",
+     "iopub.status.busy": "2024-03-29T15:39:19.252101Z",
+     "iopub.status.idle": "2024-03-29T15:39:19.339106Z",
+     "shell.execute_reply": "2024-03-29T15:39:19.338614Z",
+     "shell.execute_reply.started": "2024-03-29T15:39:19.252260Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from langchain_community.embeddings import Model2vecEmbeddings"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "8397b91f-a1f9-4be6-a699-fedaada7c37a",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-05-24T15:13:17.193751Z",
+     "start_time": "2023-05-24T15:13:17.182053Z"
+    },
+    "execution": {
+     "iopub.execute_input": "2024-03-29T15:39:19.901573Z",
+     "iopub.status.busy": "2024-03-29T15:39:19.900935Z",
+     "iopub.status.idle": "2024-03-29T15:39:19.906540Z",
+     "shell.execute_reply": "2024-03-29T15:39:19.905345Z",
+     "shell.execute_reply.started": "2024-03-29T15:39:19.901529Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "embeddings = Model2vecEmbeddings(\"minishlab/potion-base-8M\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "abcf98b7-424c-4691-a1cd-862c3d53be11",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-05-24T15:13:17.844903Z",
+     "start_time": "2023-05-24T15:13:17.198751Z"
+    },
+    "execution": {
+     "iopub.execute_input": "2024-03-29T15:39:20.434581Z",
+     "iopub.status.busy": "2024-03-29T15:39:20.433117Z",
+     "iopub.status.idle": "2024-03-29T15:39:22.178650Z",
+     "shell.execute_reply": "2024-03-29T15:39:22.176058Z",
+     "shell.execute_reply.started": "2024-03-29T15:39:20.434501Z"
+    },
+    "scrolled": true
+   },
+   "outputs": [],
+   "source": [
+    "query_text = \"This is a test query.\"\n",
+    "query_result = embeddings.embed_query(query_text)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "98897454-b280-4ee1-bbb9-2c6c15342f87",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-05-24T15:13:18.605339Z",
+     "start_time": "2023-05-24T15:13:17.845906Z"
+    },
+    "execution": {
+     "iopub.execute_input": "2024-03-29T15:39:28.164009Z",
+     "iopub.status.busy": "2024-03-29T15:39:28.161759Z",
+     "iopub.status.idle": "2024-03-29T15:39:30.217232Z",
+     "shell.execute_reply": "2024-03-29T15:39:30.215348Z",
+     "shell.execute_reply.started": "2024-03-29T15:39:28.163876Z"
+    },
+    "scrolled": true
+   },
+   "outputs": [],
+   "source": [
+    "document_text = \"This is a test document.\"\n",
+    "document_result = embeddings.embed_documents([document_text])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "11bac134",
+   "metadata": {},
+   "source": [
+    "## Direct Usage\n",
+    "\n",
+    "Here's how you would directly make use of `model2vec`\n",
+    "\n",
+    "```python\n",
+    "from model2vec import StaticModel\n",
+    "\n",
+    "# Load a model from the HuggingFace hub (in this case the potion-base-8M model)\n",
+    "model = StaticModel.from_pretrained(\"minishlab/potion-base-8M\")\n",
+    "\n",
+    "# Make embeddings\n",
+    "embeddings = model.encode([\"It's dangerous to go alone!\", \"It's a secret to everybody.\"])\n",
+    "\n",
+    "# Make sequences of token embeddings\n",
+    "token_embeddings = model.encode_as_sequence([\"It's dangerous to go alone!\", \"It's a secret to everybody.\"])\n",
+    "```"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d81e21aa",
+   "metadata": {},
+   "source": [
+    "## API Reference\n",
+    "\n",
+    "For more information check out the model2vec github [repo](https://github.com/MinishLab/model2vec)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

libs/community/langchain_community/embeddings/__init__.py

@@ -145,6 +145,9 @@ if TYPE_CHECKING:
     from langchain_community.embeddings.mlflow_gateway import (
         MlflowAIGatewayEmbeddings,
     )
+    from langchain_community.embeddings.model2vec import (
+        Model2vecEmbeddings,
+    )
     from langchain_community.embeddings.modelscope_hub import (
         ModelScopeEmbeddings,
     )
@@ -289,6 +292,7 @@ __all__ = [
     "MlflowAIGatewayEmbeddings",
     "MlflowCohereEmbeddings",
     "MlflowEmbeddings",
+    "Model2vecEmbeddings",
     "ModelScopeEmbeddings",
     "MosaicMLInstructorEmbeddings",
     "NLPCloudEmbeddings",
@@ -372,6 +376,7 @@ _module_lookup = {
     "MlflowAIGatewayEmbeddings": "langchain_community.embeddings.mlflow_gateway",
     "MlflowCohereEmbeddings": "langchain_community.embeddings.mlflow",
     "MlflowEmbeddings": "langchain_community.embeddings.mlflow",
+    "Model2vecEmbeddings": "langchain_community.embeddings.model2vec",
     "ModelScopeEmbeddings": "langchain_community.embeddings.modelscope_hub",
     "MosaicMLInstructorEmbeddings": "langchain_community.embeddings.mosaicml",
     "NLPCloudEmbeddings": "langchain_community.embeddings.nlpcloud",

libs/community/langchain_community/embeddings/model2vec.py

@@ -0,0 +1,66 @@
+"""Wrapper around model2vec embedding models."""
+
+from typing import List
+
+from langchain_core.embeddings import Embeddings
+
+
+class Model2vecEmbeddings(Embeddings):
+    """model2v embedding models.
+
+    Install model2vec first, run 'pip install -U model2vec'.
+    The github repository for model2vec is : https://github.com/MinishLab/model2vec
+
+    Example:
+        .. code-block:: python
+
+            from langchain_community.embeddings import Model2vecEmbeddings
+
+            embedding = Model2vecEmbeddings("minishlab/potion-base-8M")
+            embedding.embed_documents([
+                "It's dangerous to go alone!",
+                "It's a secret to everybody.",
+            ])
+            embedding.embed_query(
+                "Take this with you."
+            )
+    """
+
+    def __init__(self, model: str):
+        """Initialize embeddings.
+
+        Args:
+            model: Model name.
+        """
+        try:
+            from model2vec import StaticModel
+        except ImportError as e:
+            raise ImportError(
+                "Unable to import model2vec, please install with "
+                "`pip install -U model2vec`."
+            ) from e
+        self._model = StaticModel.from_pretrained(model)
+
+    def embed_documents(self, texts: List[str]) -> List[List[float]]:
+        """Embed documents using the model2vec embeddings model.
+
+        Args:
+            texts: The list of texts to embed.
+
+        Returns:
+            List of embeddings, one for each text.
+        """
+
+        return self._model.encode_as_sequence(texts)
+
+    def embed_query(self, text: str) -> List[float]:
+        """Embed a query using the model2vec embeddings model.
+
+        Args:
+            text: The text to embed.
+
+        Returns:
+            Embeddings for the text.
+        """
+
+        return self._model.encode(text)

libs/community/tests/unit_tests/embeddings/test_imports.py

@@ -26,6 +26,7 @@ EXPECTED_ALL = [
     "MlflowAIGatewayEmbeddings",
     "MlflowEmbeddings",
     "MlflowCohereEmbeddings",
+    "Model2vecEmbeddings",
     "ModelScopeEmbeddings",
     "TensorflowHubEmbeddings",
     "SagemakerEndpointEmbeddings",

libs/community/tests/unit_tests/embeddings/test_model2vec.py

@@ -0,0 +1,11 @@
+from langchain_community.embeddings.model2vec import Model2vecEmbeddings
+
+
+def test_hugginggface_inferenceapi_embedding_documents_init() -> None:
+    """Test model2vec embeddings."""
+    try:
+        embedding = Model2vecEmbeddings("minishlab/potion-base-8M")
+        assert len(embedding.embed_query("hi")) == 256
+    except Exception:
+        # model2vec is not installed
+        assert True