Docs for Local Jina AI with PEFT and GPU acceleration

docs/embeddings/jinaaipeft.md

@@ -0,0 +1,60 @@
+---
+---
+
+# Local Jina AI with PEFT and GPU acceleration
+
+This component provides a convenient wrapper around JinaAI's alpine base embeddings model or with optional PEFT config. This embedding function runs completely local, and does not require an API key.
+
+Detailed Guide for LoRA PEFT for semantic search can be found [here](https://huggingface.co/docs/peft/task_guides/semantic-similarity-lora).
+
+<Tabs queryString groupId="lang" className="hideTabSwitcher">
+<TabItem value="py" label="Python">
+
+This embedding function relies on the `peft`, `torch` & `transformers` python package, which you can install with `pip install transformers torch peft`.
+## CPU inference with Alpine Base Model
+```python
+chroma_embedding_fnc = embedding_functions.JinaAIEmbeddingsFunction(device = "cpu")
+```
+## GPU inference with PEFT
+```python
+path_to_lora_config_directory = your_lora_path
+chroma_embedding_fnc = embedding_functions.JinaAIEmbeddingsFunction(device = "cuda", adapters_path = path_to_lora_config_directory)
+chroma_collection = chroma_client.get_or_create_collection(
+    name="test_collection",
+    embedding_function=chroma_embedding_fnc,
+    metadata={
+        # https://github.com/nmslib/hnswlib
+        "hnsw:space": "cosine",  # "l2", "ip, "or "cosine". The default is "l2".
+        "hnsw:construction_ef": 1024,  
+        "hnsw:M": 64,
+    },
+)
+chroma_collection.add(
+    documents=[
+        "Lisa Su: DoB 06/07/1987 -- President and Chief Executive Officer Yeah. Sure, Joe. So, we've been engaging broadly with the customer set. I think in the last earnings call, we said that our engagements had increased seven times.",
+        "Jensen Huang: SSN 456-345-234-345 -- President and Chief Executive Officer The world has something along the lines of about $1 trillion worth of data centers installed in the cloud and enterprise and otherwise.",
+        "Jensen Huang: DoB 12/05/67 -- President and Chief Executive Officer The world has something along the lines of about $1 trillion worth of data centers installed in the cloud and enterprise and otherwise.",
+        "Elon Musk: Date of birth 12th of July '56 is the CEO of Tesla and SpaceX and one of the co-founders of PayPal.",
+    ],
+    metadatas=[
+        {"chapter": "3", "verse": "16"},
+        {"chapter": "3", "verse": "5"},
+        {"chapter": "29", "verse": "11"},
+        {"chapter": "29", "verse": "11"},
+    ],
+    ids=["id1", "id2", "id3", "id4"],
+)
+query_result = chroma_collection.query(
+    query_texts=[
+        "Lisa Su'S Date of Birth",
+        "Elon Musk's Date of Birth",
+        "Jensen Huang's Date of Birth",
+    ],
+    n_results=10,
+    include=["distances"]
+    # where={"metadata_field": "is_equal_to_this"},
+    # where_document={"$contains":"search_string"}
+)
+print(query_result)
+```
+You can pass in an optional `model_name` argument, which lets you choose which Jina model to use. By default, Chroma uses `jina-embedding-v2-base-en`.