Merge branch 'main' of https://github.com/oracle/accelerated-data-science into update_langchain_doc

Author: lu-ohai

MANIFEST.in

@@ -1,30 +1,9 @@
-include LICENSE.txt
-include THIRD_PARTY_LICENSES.txt
-include ads/common/*.json
-include ads/feature_engineering/*.json
-include ads/templates/*.jinja2
-include ads/common/function/*.yaml
-include ads/common/artifact/*.*
-include ads/model/runtime/schemas/*.yaml
-include ads/jobs/schema/**
-include ads/opctl/index.yaml
-include ads/opctl/operators/**/*.py
-include ads/opctl/docker/**
-include ads/opctl/conda/manifest_template.yaml
-include ads/opctl/conda/config.yaml
-include ads/opctl/config/diagnostics/**/*.yaml
-include ads/opctl/templates/*.jinja2
-include pyproject.toml
-global-exclude tests/**
-global-exclude notebooks/**
-exclude tests/**
-exclude notebooks/**
-exclude build/lib/tests/**
-exclude build/lib/notebooks/**
-exclude benchmark/**
-include ads/ads
-include ads/model/common/*.*
-include ads/operator/**/*.md
-include ads/operator/**/*.yaml
-include ads/operator/**/*.whl
-include ads/operator/**/MLoperator
+# Manifest.in file used by setuptools, which we are not using anymore.
+# Do not add anything to this file, unless revert back to setuptools, which is not recommended. See why here:
+# https://blog.ganssle.io/articles/2021/10/setup-py-deprecated.html.
+
+# Instead flit-core used as backend to build ADS - see section [build-system] in pyproject.toml.
+# For flit-core we can specify files and folders, which needs to be included or excluded in build using
+# [tool.flit.sdist] section of pyproject.toml. See: https://flit.pypa.io/en/latest/pyproject_toml.html#sdist-section
+# If build-system library will change, make changed to list of included/excluded to build files accordingly
+# as it will be suggested by new build-system library.

Makefile

@@ -1,36 +1,3 @@
-RELEASE_BRANCH := release/ads
-DOCS_RELEASE_BRANCH := release
-CLONE_DIR := /tmp/advanced-ds
-DOCS_CLONE_DIR := /tmp/ads-docs
-COPY_INVENTORY := CONTRIBUTING.md LICENSE.txt MANIFEST.in README-development.md README.md SECURITY.md THIRD_PARTY_LICENSES.txt
-
-prepare-release-branch: clean
-	@git checkout master
-	@git clean -xdf
-	@git pull
-	git checkout -b release/$(RELEASE_VERSION)
-
-prepare-ads:
-	@echo "Started advanced-ds clone at $$(date)"
-	@git clone ssh://git@bitbucket.oci.oraclecorp.com:7999/odsc/advanced-ds.git --branch $(RELEASE_BRANCH)  --depth 1 $(CLONE_DIR)
-	@echo "Finished cloning at $$(date)"
-	cp -r $(CLONE_DIR)/ads .
-	$(foreach var,$(COPY_INVENTORY),cp $(CLONE_DIR)/$(var) .;)
-
-prepare-docs:
-	@echo "Started ads_docs clone at $$(date)"
-	@git clone ssh://git@bitbucket.oci.oraclecorp.com:7999/odsc/ads_docs.git --branch $(DOCS_RELEASE_BRANCH)  --depth 1 $(DOCS_CLONE_DIR)
-	@echo "Finished cloning at $$(date)"
-	cp -r $(DOCS_CLONE_DIR)/source docs/ && cp $(DOCS_CLONE_DIR)/requirements.txt docs
-
-prepare: prepare-release-branch prepare-ads prepare-docs
-
-push: clean
-	@bash -c 'if [[ $$(git branch | grep \*) == "* release/$(RELEASE_VERSION)" ]];then echo "Version matching current branch"; else echo "Set proper value to RELEASE_VERSION";exit 1 ; fi'
-	@git add .
-	@git commit -m "Release version: $(RELEASE_VERSION)"
-	@git push --set-upstream origin release/$(RELEASE_VERSION)
-
 dist: clean
 	@python3 -m build
 
@@ -43,5 +10,4 @@ clean:
 	@find ./ -name '*.pyc' -exec rm -f {} \;
 	@find ./ -name 'Thumbs.db' -exec rm -f {} \;
 	@find ./ -name '*~' -exec rm -f {} \;
-	@rm -rf $(CLONE_DIR)
-	@rm -rf $(DOCS_CLONE_DIR)
+	@find ./ -name '.DS_Store' -exec rm -f {} \;

THIRD_PARTY_LICENSES.txt

@@ -54,6 +54,12 @@ docker
 * Source code: https://github.com/docker
 * Project home: https://www.docker.com/
 
+evaluate
+* Copyright 2023 HuggingFace Inc.
+* License: Apache-2.0 license
+* Source code: https://github.com/huggingface/evaluate
+* Project home: https://huggingface.co/docs/evaluate/index
+
 fastavro
 * Copyright (c) 2011 Miki Tebeka
 * License: MIT License
@@ -133,6 +139,12 @@ jinja2
 * Source code: https://github.com/pallets/jinja/
 * Project home: https://palletsprojects.com/p/jinja/
 
+langchain
+* Copyright (c) 2023 LangChain, Inc.
+* License: MIT license
+* Source code: https://github.com/langchain-ai/langchain
+* Project home: https://www.langchain.com/
+
 lightgbm
 * Copyright (c) 2023 Microsoft Corporation
 * License: MIT license

ads/llm/serialize.py

@@ -15,7 +15,6 @@
 from langchain.chains import RetrievalQA
 from langchain.chains.loading import load_chain_from_config
 from langchain.llms import loading
-from langchain.load import dumpd
 from langchain.load.load import Reviver
 from langchain.load.serializable import Serializable
 from langchain.schema.runnable import RunnableParallel

ads/llm/serializers/retrieval_qa.py

@@ -1,3 +1,9 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*--
+
+# Copyright (c) 2023 Oracle and/or its affiliates.
+# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
+
 import base64
 import json
 import os
@@ -30,7 +36,7 @@ def load(config: dict, **kwargs):
                 os.environ.get("OCI_OPENSEARCH_PASSWORD", None),
             ),
             verify_certs=True
-            if os.environ.get("OCI_OPENSEARCH_VERIFY_CERTS", None).lower() == "true"
+            if os.environ.get("OCI_OPENSEARCH_VERIFY_CERTS", None) == "True"
             else False,
             ca_certs=os.environ.get("OCI_OPENSEARCH_CA_CERTS", None),
         )
@@ -44,18 +50,18 @@ def save(obj):
         serialized["type"] = "constructor"
         serialized["_type"] = OpenSearchVectorDBSerializer.type()
         kwargs = {}
-        for key, val in obj.__dict__.items():
-            if key == "client":
-                if isinstance(val, OpenSearch):
-                    client_info = val.transport.hosts[0]
+        for component_name, component in obj.__dict__.items():
+            if component_name == "client":
+                if isinstance(component, OpenSearch):
+                    client_info = component.transport.hosts[0]
                     opensearch_url = (
                         f"https://{client_info['host']}:{client_info['port']}"
                     )
                     kwargs.update({"opensearch_url": opensearch_url})
                 else:
                     raise NotImplementedError("Only support OpenSearch client.")
                 continue
-            kwargs[key] = dump(val)
+            kwargs[component_name] = dump(component)
         serialized["kwargs"] = kwargs
         return serialized
 

ads/secrets/secrets.py

@@ -324,9 +324,7 @@ def __enter__(self, *args, **kwargs):
                     )
             os.environ.update(export_dict)
             self.exported_keys = list(export_dict.keys())
-            logger.debug(
-                f"Exporting credentials to environment. Keys: {self.exported_keys}"
-            )
+            logger.debug(f"Exporting credentials to environment.")
         else:
             logger.debug(f"Skipping Export credentials to environment.")
 

docs/source/user_guide/large_language_model/index.rst

@@ -4,9 +4,37 @@
 Large Language Model
 ####################
 
+Oracle Cloud Infrastructure (OCI) provides fully managed infrastructure to work with Large Language Model (LLM). 
+
+Train and Deploy LLM
+********************
+You can train LLM at scale with multi-node and multi-GPU using `Data Science Jobs (Jobs) <https://docs.oracle.com/en-us/iaas/data-science/using/jobs-about.htm>`_, and deploy it with `Data Science Model Deployment (Model Deployments) <https://docs.oracle.com/en-us/iaas/data-science/using/model-dep-about.htm>`_. The following blog posts show examples training and deploying Llama2 models:
+
+* `Multi-GPU multinode fine-tuning Llama2 on OCI Data Science <https://blogs.oracle.com/ai-and-datascience/post/multi-gpu-multi-node-finetuning-llama2-oci>`_
+* `Deploy Llama 2 in OCI Data Science <https://blogs.oracle.com/ai-and-datascience/post/llama2-oci-data-science-cloud-platform>`_
+* `Quantize and deploy Llama 2 70B on cost-effective NVIDIA A10 Tensor Core GPUs in OCI Data Science <https://blogs.oracle.com/ai-and-datascience/post/quantize-deploy-llama2-70b-costeffective-a10s-oci>`_
+
+
+Integration with LangChain
+**************************
+ADS is designed to work with LangChain, enabling developers to incorporate various LangChain components and models deployed on OCI seamlessly into their applications. Additionally, ADS can package LangChain applications and deploy it as a REST API endpoint using OCI Data Science Model Deployment.
+
+
+.. admonition:: Installation
+  :class: note
+
+  Install ADS and other dependencies for LLM integrations.
+
+  .. code-block:: bash
+
+    $ python3 -m pip install "oracle-ads[llm]"
+
+
 
 .. toctree::
     :hidden:
     :maxdepth: 2
 
+    training_llm
     deploy_langchain_application
+    retrieval