Embedding generation in Prompt Engineering / GenAI - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to generate an embedding vector from text using a model.

Prompt Engineering / GenAI

embedding = model.[1](text)

Drag options to blanks, or click blank then click option'

Afit

Btrain

Cpredict

Dencode

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'train' or 'fit' instead of 'encode' to get embeddings.

Using 'predict' which is for classification or regression outputs.

The encode method converts text into an embedding vector.

2fill in blank

medium

Complete the code to normalize the embedding vector to unit length.

Prompt Engineering / GenAI

normalized_embedding = embedding / [1](embedding)

Drag options to blanks, or click blank then click option'

Asum

Blen

Cnp.linalg.norm

Dmax

Attempts:

3 left

💡 Hint

Common Mistakes

Using sum or max which do not compute vector length.

Using len which returns number of elements, not magnitude.

We use np.linalg.norm to compute the length (magnitude) of the embedding vector for normalization.

3fill in blank

hard

Fix the error in the code to generate embeddings for a list of texts.

Prompt Engineering / GenAI

embeddings = [model.[1](text) for text in texts]

Drag options to blanks, or click blank then click option'

Aencode

Btrain

Cfit

Dpredict

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'train' or 'fit' which are for model training, not embedding generation.

Using 'predict' which is for output predictions, not embeddings.

The encode method correctly generates embeddings for each text in the list.

4fill in blank

hard

Fill both blanks to create a dictionary of text to embedding length for texts longer than 5 characters.

Prompt Engineering / GenAI

embedding_lengths = {text: len(model.[1](text)) for text in texts if len(text) [2] 5}

Drag options to blanks, or click blank then click option'

Aencode

B>

C<

Dpredict

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'predict' instead of 'encode' for embeddings.

Using '<' instead of '>' which filters shorter texts.

We encode each text to get embeddings and filter texts with length greater than 5.

5fill in blank

hard

Fill all three blanks to compute cosine similarity between two normalized embeddings.

Prompt Engineering / GenAI

cos_sim = np.dot([1], [2]) / (np.linalg.norm([3]) * np.linalg.norm([2]))

Drag options to blanks, or click blank then click option'

Aembedding1

Bembedding2

Dembedding3

Attempts:

3 left

💡 Hint

Common Mistakes

Using a different vector like embedding3 which is undefined.

Mixing up the order of embeddings in dot product or norms.

Cosine similarity is the dot product of two vectors divided by the product of their norms. Here, embedding1 and embedding2 are used correctly.