Why advanced RAG improves answer quality in Prompt Engineering / GenAI - Test Your Understanding

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the Retriever class used in RAG models.

Prompt Engineering / GenAI

from transformers import [1]

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

ARetriever

BRagRetriever

CTokenizer

DRagModel

Attempts:

3 left

💡 Hint

Common Mistakes

Choosing RagModel instead of RagRetriever

Using Tokenizer which is unrelated to retrieval

Using a generic Retriever class not specific to RAG

The RagRetriever class is used to fetch relevant documents in advanced RAG models, improving answer quality by retrieving context.

2fill in blank

medium

Complete the code to initialize a RAG model with a retriever.

Prompt Engineering / GenAI

from transformers import RagTokenizer, RagSequenceForGeneration, RagRetriever

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

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

Aretriever

Bmodel

Ctokenizer

DNone

Attempts:

3 left

💡 Hint

Common Mistakes

Passing tokenizer instead of retriever

Passing None disables retrieval

Passing model itself causes error

The retriever object is passed to the RAG model to enable it to fetch relevant documents, which improves answer quality.

3fill in blank

hard

Fix the error in this code that generates an answer using the RAG model.

Prompt Engineering / GenAI

input_dict = tokenizer(question, return_tensors='pt')
outputs = model.generate(input_ids=[1], num_beams=5, max_length=50)
answer = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]

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

Ainput_dict['labels']

Binput_dict['attention_mask']

Cinput_dict['token_type_ids']

Dinput_dict['input_ids']

Attempts:

3 left

💡 Hint

Common Mistakes

Using attention_mask instead of input_ids

Using token_type_ids which is optional

Using labels which is for training

The input_ids key contains the token IDs needed for generation. Using other keys causes errors or wrong input.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that filters retrieved documents with score above 0.5.

Prompt Engineering / GenAI

filtered_docs = {doc: score for doc, score in retrieved_docs.items() if score [1] 0.5 and len(doc) [2] 0}

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

A>

B<

C!=

D==

Attempts:

3 left

💡 Hint

Common Mistakes

Using '<' instead of '>' for score comparison

Using '==' instead of '!=' for length check

Using '<' for length which excludes valid docs

We want scores greater than 0.5 and documents with length greater than 0 to keep only relevant, non-empty docs.

5fill in blank

hard

Fill all three blanks to build a dictionary of document texts and their scores filtered by score > 0.7.

Prompt Engineering / GenAI

high_score_docs = [1]: [2] for [3], [2] in docs_with_scores.items() if [2] > 0.7

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

Atext

Bscore

Cdoc

Ditem

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'text' instead of 'doc' for keys

Using 'item' which is undefined

Mixing variable names inconsistently

The dictionary comprehension uses doc as key, score as value, iterating over doc, score pairs.