Translation in Prompt Engineering / GenAI - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load the translation model using Hugging Face Transformers.

Prompt Engineering / GenAI

from transformers import MarianMTModel, MarianTokenizer
model_name = 'Helsinki-NLP/opus-mt-en-de'
tokenizer = MarianTokenizer.from_pretrained([1])

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

A'en-de'

B'Helsinki-NLP/opus-mt-en-de'

Cmodel_name

DMarianTokenizer

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the model name as a string literal instead of using the variable.

Passing the class name instead of the model name.

The variable model_name holds the model identifier string, so it should be passed to from_pretrained() to load the tokenizer.

2fill in blank

medium

Complete the code to tokenize the input text for translation.

Prompt Engineering / GenAI

text = 'Hello, how are you?'
inputs = tokenizer([1], return_tensors='pt', padding=True)

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

Atext

B'text'

C['Hello']

D['Hello, how are you?']

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the string 'text' instead of the variable text.

Passing a list when a string is expected.

The variable text contains the input string and should be passed directly to the tokenizer.

3fill in blank

hard

Fix the error in generating the translated tokens from the model output.

Prompt Engineering / GenAI

translated_tokens = model.generate([1])

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

Ainputs['input_ids']

Binputs['attention_mask']

Cinputs

Dinputs['token_type_ids']

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the whole inputs dictionary instead of just the input IDs.

Passing attention mask or token type IDs instead of input IDs.

The generate method requires the input IDs tensor, which is accessed by inputs['input_ids'].

4fill in blank

hard

Fill both blanks to decode the translated tokens into a readable string.

Prompt Engineering / GenAI

translated_text = tokenizer.decode([1][0], skip_special_tokens=[2])

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

Atranslated_tokens

BTrue

CFalse

Dinputs

Attempts:

3 left

💡 Hint

Common Mistakes

Using the inputs variable instead of translated tokens.

Setting skip_special_tokens to False causing special tokens to appear.

The decode method takes the first element of translated_tokens and skips special tokens by setting skip_special_tokens=True.

5fill in blank

hard

Fill all three blanks to create a function that translates English text to German.

Prompt Engineering / GenAI

def translate_en_to_de(text):
    tokenizer = MarianTokenizer.from_pretrained([1])
    model = MarianMTModel.from_pretrained([2])
    inputs = tokenizer(text, return_tensors=[3], padding=True)
    translated_tokens = model.generate(inputs['input_ids'])
    return tokenizer.decode(translated_tokens[0], skip_special_tokens=True)

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

A'Helsinki-NLP/opus-mt-en-de'

C'pt'

D'tf'

Attempts:

3 left

💡 Hint

Common Mistakes

Using different model names for tokenizer and model.

Using 'tf' instead of 'pt' for return_tensors causing errors.

The function loads the tokenizer and model for English to German translation using the same model name string. The tokenizer returns PyTorch tensors, so return_tensors='pt' is used.

Practice

1. What is the main purpose of a translation model in AI?

easy

A. To change text from one language to another automatically

B. To generate images from text descriptions

C. To recognize faces in photos

D. To sort numbers in a list

Solution

1. Step 1: Understand the function of translation models

   Translation models convert text from one language to another automatically.

2. Step 2: Compare with other AI tasks

   Other options describe different AI tasks like image generation or face recognition, not translation.

3. Final Answer:

   To change text from one language to another automatically -> Option A

4. Quick Check:

   Translation = language conversion [OK]

Hint: Translation means changing languages automatically [OK]

Common Mistakes:

• Confusing translation with image generation
• Thinking translation sorts data
• Mixing translation with face recognition

2. Which of the following is the correct way to call a pre-trained translation model in Python using a library like Hugging Face Transformers?

easy

A. model = pipeline('image-classification')

B. model = pipeline('speech-recognition')

C. model = pipeline('text-generation')

D. model = pipeline('translation_en_to_fr')

Solution

1. Step 1: Identify the pipeline for translation

   The correct pipeline for English to French translation is 'translation_en_to_fr'.

2. Step 2: Check other pipeline types

   Other options are for different tasks like image classification, text generation, or speech recognition, not translation.

3. Final Answer:

   model = pipeline('translation_en_to_fr') -> Option D

4. Quick Check:

   Translation pipeline = 'translation_en_to_fr' [OK]

Hint: Use 'translation_en_to_fr' for English to French translation [OK]

Common Mistakes:

• Using wrong pipeline name
• Confusing translation with image tasks
• Calling text generation instead of translation

3. Given the following Python code using a translation model, what will be the output?

from transformers import pipeline
translator = pipeline('translation_en_to_de')
result = translator('Hello, how are you?')
print(result[0]['translation_text'])

medium

A. Ciao, come stai?

B. Bonjour, comment ça va?

C. Hallo, wie geht es dir?

D. Hola, ¿cómo estás?

Solution

1. Step 1: Identify the translation direction

   The pipeline is 'translation_en_to_de', which means English to German translation.

2. Step 2: Translate the input text

   'Hello, how are you?' translates to 'Hallo, wie geht es dir?' in German.

3. Final Answer:

   Hallo, wie geht es dir? -> Option C

4. Quick Check:

   English to German translation = Hallo, wie geht es dir? [OK]

Hint: Check language codes: en_to_de means English to German [OK]

Common Mistakes:

• Choosing French or Spanish output
• Ignoring language direction
• Assuming output is same as input

4. You wrote this code to translate English to Spanish but get an error:

from transformers import pipeline
translator = pipeline('translation_en_to_es')
result = translator('Good morning')
print(result['translation_text'])

What is the error and how to fix it?

medium

A. Accessing result as dict instead of list; use result[0]['translation_text']

B. Wrong pipeline name; should be 'translation_en_to_fr'

C. Missing model download; add download=True parameter

D. print statement syntax error; use print result['translation_text']

Solution

1. Step 1: Understand the output format of pipeline

   The pipeline returns a list of dicts, so result is a list, not a dict.

2. Step 2: Correct the access to translation text

   Access the first element with result[0], then get 'translation_text' key.

3. Final Answer:

   Accessing result as dict instead of list; use result[0]['translation_text'] -> Option A

4. Quick Check:

   Pipeline output is list of dicts [OK]

Hint: Pipeline returns list; access first item before keys [OK]

Common Mistakes:

• Treating output as dict directly
• Using wrong pipeline name
• Incorrect print syntax

5. You want to build a program that translates a list of English sentences to French and then back to English to check accuracy. Which approach is best?

hard

A. Translate sentences manually without AI models

B. Use two pipelines: 'translation_en_to_fr' then 'translation_fr_to_en' on each sentence

C. Use 'translation_en_to_de' pipeline followed by 'translation_de_to_en'

D. Use only 'translation_en_to_fr' pipeline twice on each sentence

Solution

1. Step 1: Identify correct translation directions

   To translate English to French and back, use 'translation_en_to_fr' then 'translation_fr_to_en'.

2. Step 2: Avoid wrong language pairs

   Using German pipelines or repeating the same pipeline won't give correct back translation.

3. Step 3: Manual translation is inefficient and error-prone

   AI pipelines automate and improve accuracy checking.

4. Final Answer:

   Use two pipelines: 'translation_en_to_fr' then 'translation_fr_to_en' on each sentence -> Option B

5. Quick Check:

   Back translation needs correct language pairs [OK]

Hint: Use matching forward and backward pipelines for accuracy check [OK]

Common Mistakes:

• Using wrong language pairs
• Repeating same pipeline twice
• Ignoring AI automation