Hugging Face integration basics in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a pretrained Hugging Face model.

PyTorch

from transformers import AutoModel
model = AutoModel.[1]('bert-base-uncased')

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

Acreate

Bfrom_pretrained

Cinit_model

Dload_model

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'load_model' which is not a Hugging Face method.

Using 'init_model' which does not exist in transformers.

Using 'create' which is not a valid method.

The method to load a pretrained model from Hugging Face is 'from_pretrained'.

2fill in blank

medium

Complete the code to tokenize input text using a Hugging Face tokenizer.

PyTorch

from transformers import AutoTokenizer
tokenizer = AutoTokenizer.[1]('bert-base-uncased')
tokens = tokenizer('Hello world!')

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

Aload

Binit

Cfrom_pretrained

Dcreate_tokenizer

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'load' which is not a valid method.

Using 'init' which does not exist.

Using 'create_tokenizer' which is not a transformers method.

Tokenizers are loaded with 'from_pretrained' method to match the model.

3fill in blank

hard

Fix the error in the code to get model outputs from tokenized inputs.

PyTorch

outputs = model([1])

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

Ainputs

Btokenizer

Ctokens

Dtokenizer('Hello world!')

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the tokenizer object instead of tokenized inputs.

Passing raw tokens without converting to tensors.

Passing the tokenizer call directly without saving to a variable.

The model expects a dictionary of inputs, usually named 'inputs' after tokenization with return_tensors.

4fill in blank

hard

Fill both blanks to tokenize text with padding and convert to tensors.

PyTorch

inputs = tokenizer('Hello Hugging Face!', [1]=True, [2]='pt')

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

Apadding

Btruncation

Creturn_tensors

Dmax_length

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'truncation' instead of 'padding' for padding.

Using 'max_length' without padding.

Using 'pt' as a boolean instead of a string for return_tensors.

To pad sequences and get PyTorch tensors, use padding=True and return_tensors='pt'.

5fill in blank

hard

Fill all three blanks to extract the last hidden state from model outputs.

PyTorch

outputs = model(inputs)
last_hidden_state = outputs.[1]
print(last_hidden_state.[2])
print(last_hidden_state.shape[[3]])

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

Alast_hidden_state

Bdetach()

C0

Dhidden_states

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'hidden_states' which is not directly returned unless configured.

Printing tensor without detach causing gradient tracking.

Using wrong shape index.

The model output has 'last_hidden_state' attribute. Use detach() to get tensor without gradient. Shape index 0 is batch size.