Why fine-tuning adapts models to domains in Prompt Engineering / GenAI - Test Your Understanding

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a pre-trained model for fine-tuning.

Prompt Engineering / GenAI

model = load_model('[1]')

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

Apretrained-domain-model

Bempty-model

Crandom-model

Dbase-model

Attempts:

3 left

💡 Hint

Common Mistakes

Choosing a model without prior training like 'random-model'.

We start fine-tuning from a base-model that already learned general knowledge.

2fill in blank

medium

Complete the code to prepare domain-specific data for fine-tuning.

Prompt Engineering / GenAI

domain_data = load_data('[1]')

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

Arandom_text.txt

Bgeneral_corpus.txt

Cdomain_corpus.txt

Dempty_file.txt

Attempts:

3 left

💡 Hint

Common Mistakes

Using general or unrelated data files for fine-tuning.

Fine-tuning uses domain_corpus.txt to adapt the model to the specific domain.

3fill in blank

hard

Fix the error in the fine-tuning loop by completing the missing method call.

Prompt Engineering / GenAI

for batch in domain_data:
    loss = model.[1](batch)
    loss.backward()

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

Aforward

Btrain

Cpredict

Devaluate

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'predict' or 'evaluate' which are for inference, not training.

The forward method computes the output and loss during training.

4fill in blank

hard

Fill both blanks to update the model parameters during fine-tuning.

Prompt Engineering / GenAI

optimizer.[1]()
optimizer.[2]()

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

Azero_grad

Bstep

Cbackward

Dforward

Attempts:

3 left

💡 Hint

Common Mistakes

Calling 'backward' or 'forward' on optimizer which is incorrect.

First, zero_grad() clears old gradients. Then, step() updates model weights.

5fill in blank

hard

Fill all three blanks to create a dictionary of fine-tuned model metrics.

Prompt Engineering / GenAI

metrics = {'loss': [1], 'accuracy': [2], 'epoch': [3]

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

Aloss_value

Bacc_value

Ccurrent_epoch

Dmodel_output

Attempts:

3 left

💡 Hint

Common Mistakes

Using model output instead of metric values.

We store the current loss_value, acc_value, and current_epoch in the metrics dictionary.