embeddings() examples | Teradata Package for Generative AI - Examples: How to use embeddings() - Teradata VantageCloud Lake

Example setup: Import required packages and set up input data

>>> import os
>>> import teradatagenai
 >>> from teradatagenai import TeradataAI, TextAnalyticsAI, load_data
>>> from teradataml import DataFrame
>>> load_data('employee', 'employee_data')
>>> data = DataFrame('employee_data')
>>> df_articles = data.select(["employee_id", "employee_name", "articles"])

Complete the example setup requirements.

Example 1: Generate the embeddings for employee reviews from the 'reviews' column of a teradataml DataFrame using hugging face model 'all-MiniLM-L6-v2'

Create an LLM endpoint.

>>> model_name = 'sentence-transformers/all-MiniLM-L6-v2'
>>> model_args = {'transformer_class': 'AutoModel',
                  'task' : 'token-classification'}
>>> llm = TeradataAI(api_type = "hugging_face",
                      model_name = model_name,
                      model_args = model_args)

Create a TextAnalyticsAI object.

>>> obj = TextAnalyticsAI(llm=llm)

Generate embeddings for input text column.

>>> obj.embeddings(column = "articles",
                    data = df_articles,
                    libs = 'sentence_transformers',
                    delimiter = '#')

Output:

Example 2: Extend Example 1 to use user defined script as input

>>> base_dir = os.path.dirname(teradatagenai.__file__)
>>> embeddings_script = os.path.join(base_dir,
                                     'example-data',
                                     'embeddings.py')

Construct returns argument based on the user defined script.

>>> returns = OrderedDict([('text', VARCHAR(512))])

>>> _ = [returns.update({"v{}".format(i+1): VARCHAR(1000)}) for i in range(384)]
>>> obj.embeddings(column = "articles",
                   data = df_articles,
                   script = embeddings_script,
                   returns = returns,
                   libs = 'sentence_transformers',
                   delimiter = '#')

Output:

df_reviews = df_reviews.set_index('employee_id')
df_filtered_review = df_reviews.filter(items=[2], axis=0)
obj_nim.embeddings(column="articles",data=data,accumulate='articles',output_format='VARCHAR')

Output:

Import the modules and create a teradataml DataFrame.

>>> from teradatagenai import TeradataAI, TextAnalyticsAI, load_data
>>> from teradataml import DataFrame
>>> load_data("byom", "amazon_reviews_25")
>>> amazon_reviews_25 = DataFrame.from_table("amazon_reviews_25")

Set up a TeradataAI onnx endpoint.

>>> llm_onnx= TeradataAI(api_type = "onnx",
                         model_name = "bge-small-en-v1.5",
                         model_id = "td-bge-small",
                         model_path = "/path/to/onnx/model",
                         tokenizer_path = "/path/to/onnx/tokenizer,
                         table_name = "onnx_models")

Create a TextAnalyticsAI object and generate embeddings for 'rev_text' column in amazon_reviews_25 teradataml DataFrame.

>>> obj = TextAnalyticsAI(llm=llm_onnx)
>>> obj.embeddings(data=amazon_reviews_25,
                   column = "rev_text", 
                   accumulate= "rev_id",
                   model_output_tensor = "sentence_embedding")

Output: