RAG architecture overview in Prompt Engineering / GenAI - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to define the main components of a RAG model.

Prompt Engineering / GenAI

from transformers import RagTokenizer, RagRetriever, RagSequenceForGeneration

tokenizer = RagTokenizer.from_pretrained('facebook/rag-token-nq')
retriever = RagRetriever.from_pretrained('facebook/rag-token-nq')
model = RagSequenceForGeneration.from_pretrained('facebook/rag-token-nq', retriever=[1])

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

Aretriever

BNone

Cmodel

Dtokenizer

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the tokenizer instead of the retriever to the model.

Passing None or forgetting to pass the retriever.

The retriever component must be passed to the RAG model to enable document retrieval during generation.

2fill in blank

medium

Complete the code to tokenize the input question for the RAG model.

Prompt Engineering / GenAI

question = "What is RAG architecture?"
inputs = tokenizer(question, return_tensors=[1])

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

Anp

Btf

Cpt

Dpd

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'tf' which returns TensorFlow tensors instead of PyTorch tensors.

Using unsupported return_tensors values like 'np' or 'pd'.

The tokenizer returns PyTorch tensors when 'pt' is specified as the return_tensors argument.

3fill in blank

hard

Fix the error in generating answers with the RAG model.

Prompt Engineering / GenAI

outputs = model.generate(input_ids=inputs['input_ids'], [1]=inputs['attention_mask'])

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

Ainput_mask

Battention_mask

Cmask

Dattn_mask

Attempts:

3 left

💡 Hint

Common Mistakes

Using incorrect argument names like 'mask' or 'input_mask' causes errors.

Omitting the attention mask leads to wrong model behavior.

The generate method expects the argument 'attention_mask' to specify which tokens to attend to.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that maps retrieved document titles to their scores.

Prompt Engineering / GenAI

doc_scores = {doc['title']: doc[1] for doc in retrieved_docs if doc[1] [2] 0.5}

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

A['score']

B>

C<

D['id']

Attempts:

3 left

💡 Hint

Common Mistakes

Using wrong keys like 'id' instead of 'score'.

Using incorrect comparison operators like '<' instead of '>'.

We access the 'score' key for each document and filter documents with score greater than 0.5.

5fill in blank

hard

Fill all three blanks to create a dictionary comprehension that maps document IDs to their text if the text length is greater than 100.

Prompt Engineering / GenAI

filtered_docs = {doc[1]: doc[2] for doc in docs if len(doc[3]) > 100}

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

A['id']

B['text']

D['title']

Attempts:

3 left

💡 Hint

Common Mistakes

Using wrong keys like 'title' instead of 'text' or 'id'.

Not using len() to check text length.

We map document IDs to their text and filter documents where the text length is greater than 100 characters.