..

AI Embeddings - OpenAI, HuggingFace, SentenceTransformers

This session idea was inspired by a blog post on Medium1. You can find the original code on Github2 3 and Google Colab4.

One cool feature in the original Google Colab notebook4 is the ability to build a similarity matrix, which lets you compare dot product scores for all possible sentence combinations.

If you’re building an application and the relevant chunks aren’t picked up, try changing the model1.

Setup

python3 -m venv env-embedding
source env-embedding/bin/activate

# `-I`  ignore the installed packages, overwriting them.
# `-U`  upgrade all specified packages to the newest available version.

pip3 install -U langchain_openai==0.1.8 openai==1.30.5 langchain==0.2.1 langchain_community==0.2.1 sentence_transformers==3.0.0

pip3 install --upgrade --force-reinstall langchain_openai openai langchain langchain_community sentence_transformers

pip3 show langchain_openai openai langchain langchain_community sentence_transformers

pip3 index versions langchain_openai openai langchain langchain_community sentence_transformers

Code Sample

embeddings.py 
from pprint import pprint
# check the dimensionality of the returned vectors.
import numpy as np

sentences_A = [
  "Best travel neck pillow for long flights",
  "Lightweight backpack for hiking and travel",
  "Waterproof duffel bag for outdoor adventures",
  "Stainless steel cookware set for induction cooktops",
  "High-quality chef's knife set",
  "High-performance stand mixer for baking",
  "New releases in fiction literature",
  "Inspirational biographies and memoirs",
  "Top self-help books for personal growth",
]

sentences_B = [
  "I'm afraid of flights",
  "I like flying",
  "aerophobia",
  "Scary pictures of airplanes",
  "Do planes fear thunder?",
]

sentences = sentences_A

# OPENAI
from langchain_openai import OpenAIEmbeddings
openai_embedding = OpenAIEmbeddings()

# or

from openai import OpenAI
client = OpenAI()

def get_embedding(text, model="text-embedding-3-small"):
  text = text.replace("\n", " ")
  return client.embeddings.create(input = [text], model=model).data[0].embedding

_list = [get_embedding(s) for s in sentences]
# print(_list)
# check the dimensionality of the returned vectors.
pprint(np.array(_list).shape)
print('---')

# DIRECTLY FROM HUGGINGFACE
# https://docs.llamaindex.ai/en/stable/examples/embeddings/huggingface/
from langchain_community.embeddings import HuggingFaceEmbeddings
mpnet_embeddings = HuggingFaceEmbeddings(
  model_name="sentence-transformers/all-mpnet-base-v2",
)
print('---')

_list = [mpnet_embeddings.embed_query(s) for s in sentences]
# print(_list)
# check the dimensionality of the returned vectors.
pprint(np.array(_list).shape)
print('---')

# LOAD AS SENTENCETRANSFORMERS
from sentence_transformers import SentenceTransformer
gist_embedding = SentenceTransformer("avsolatorio/GIST-Embedding-v0")

_list = gist_embedding.encode(sentences, convert_to_tensor=True)
# print(_list)
# check the dimensionality of the returned vectors.
pprint(np.array(_list.cpu().numpy()).shape)
"""
TypeError: can't convert mps:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
"""
print('---')

OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx \
  python3 embeddings.py

  1. LangChain 101: Part 3b. Talking to Documents: Embeddings and Vectorstores ↩ ↩2

  2. Public code of Dr. Ivan Reznikov used in posts, articles, conferences ↩

  3. Notebook hosted on Github by Dr. Ivan Reznikov ↩

  4. Notebook hosted on Google Colab by Dr. Ivan Reznikov ↩ ↩2