Introduction – Quiz

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

menu

alarm  

14m 33s

(Submitted)

Q1 of 4outlined_flag

Which of the following sections in OpenAI helps in experimenting and iterating with different custom use cases?

Playground check

Resources

Examples

Documentation

Option Playground is correct

Explanation :

Yes!! You got the right answer!

Q2 of 4outlined_flag

OpenAI playground supports users to experiment with various preset comprising of predefined training text and respective settings.

True check

False

Option True is correct

Explanation :

Yes!! You got the right answer!

Q3 of 4outlined_flag

Users can interact with the OpenAI API’s only using the Playgrounds

True

False check

Option False is correct

Explanation :

Yes!! You got the right answer!

Q4 of 4outlined_flag

 Which of the following are the basic constructs of OpenAI API?

prompt

completion

tokens

All the given options check

Option All the given options is correct

Explanation :

Yes!! You got the right answer!

Development using GPT-3 -Quiz

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

menu

alarm  

7m

(Submitted)

Q1 of 3outlined_flag

The most capable engine (GPT-3 model) of all different flavors of the GPT-3 model provided in openAI API is:

ada

babbage

curie

davinci check

Option davinci is correct

Explanation :

Yes!! You got the right answer!

Q2 of 3outlined_flag

Lower values for temperature parameter in the parameter settings in GPT-3 API makes the model

deterministic check

creative

Random

None of the above

Option deterministic is correct

Explanation :

Yes!! You got the right answer!

Q3 of 3outlined_flag

The parameter used to control the response length in GPT-3 API from openAI is:

temperature

max_tokens check

frequency_penalty

logprobs

Option max_tokens is correct

Explanation :

Yes!! You got the right answer!

Language model - Quiz

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

menu

alarm  

6m 59s

(Submitted)

Q1 of 3outlined_flag

Which of following techniques is an efficient way to encode words in the language model?

Label encoding

Bag of words

TFIDF vectorizer

Word2Vec check

Option Word2Vec is correct

Explanation :

Yes!! You got the correct answer!

Q2 of 3outlined_flag

Shashi wants to translate an English sentence to French and she has some training data for English French translation.

Which of the following techniques she can use to perform this ML task?

RNNs

GRUs

LSTMs

Transformers

All the above check

Option All the above is correct

Explanation :

Yes!! You got the correct answer!

Q3 of 3outlined_flag

What are the advantages of using transfer learning technique in model building for an NLP task?

Much less data required

Reduces the training time

Speeds up the model building process

All the above check

Option All the above is correct

Explanation :

Yes!! You got the correct answer!

Input embeddings and Positional encoding- Quiz

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

menu

alarm  

14m 2s

(Submitted)

Q1 of 6outlined_flag

What should be the characteristics of a numerical representation of textual data ?

It should be semantically meaningful

Should provide informative input representation

Should have a massive impact on overall model performance

All of the above check

Option All of the above is correct

Explanation :

All the three options are the characteristics of numerical representation of textual data

Q2 of 6outlined_flag

Why don’t we use one hot encoding to represent words as vectors ?

With every increase of ‘n’ in the vocabulary, the number of features will also increase by ‘n’. check

With every increase of ‘n’ in the vocabulary, the number of features will also increase by ‘n/2’.

It is not possible to create vectors for more than 10,000 words.

None of the above

Option With every increase of ‘n’ in the vocabulary, the number of features will also increase by ‘n’. is correct

Explanation :

Correct!

Q3 of 6outlined_flag

 State True or False-

We can think of word embedding algorithms like word2vec as unsupervised feature extractors for words.

True check

False

Option True is correct

Explanation :

Correct!

Q4 of 6outlined_flag

State true or false-

RNNs (Recurrent neural networks) do not parse the sentences word by word in a sequential manner.

True

False check

Option False is correct

Explanation :

Correct!

Q5 of 6outlined_flag

Fill in the blanks-

For every odd index on the input vector, we create a vector using the ________ function. For every even index, create a vector using the ______ function.

sine, cosine

cosine, sine check

cosine, tan

sigmoid, relu

Option cosine, sine is correct

Explanation :

Correct!

Q6 of 6outlined_flag

Which of the following is the core idea behind word embeddings ?

words that used in different contexts will have similar representations.

words that used in similar contexts will have similar representations. check

words that used in similar contexts will have different representations.

words that used in different contexts will have different representations.

Option words that used in similar contexts will have similar representations. is correct

Explanation :

Correct!

Transformer Model Quiz

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

menu

alarm  

6m 49s

(Submitted)

Q1 of 4outlined_flag

Which of following machine learning technique can be trained faster with less amount of training data?

N-Grams approach

Sequence-to-sequence modeling using LSTM

Transformer Architecture check

Word Sense disambiguation

Option Transformer Architecture is correct

Explanation :

Yes!! You got the correct answer!

Q2 of 4outlined_flag

What do you mean by Transformation based learning?

The technique in which rules are manually induced from data

The technique in which rules are automatically induced from data check

The technique in which rules are transformed into another data

The technique in which rules are not used.

Option The technique in which rules are automatically induced from data  is correct

Explanation :

Yes!! You got the correct answer!

Q3 of 4outlined_flag

What is most the important concepts which makes the transformer model most powerful?

Encoder-decoder concept

Self-attention check

Artificial Neural Network

All of the above

Option Self-attention is correct

Explanation :

Yes!! You got the correct answer!

Q4 of 4outlined_flag

Which type of attention is used to reduce the information leak while training a model?

Encoder-decoder attention

Self-attention in input sequence

Self-attention in output sequence check

Both a and b

Option Self-attention in output sequence is correct

Explanation :

Yes!! You got the correct answer!

Problem Statement

Problem: Generating SQL commands from English text

Steps:

·  Load the key stored in json file

·  Create an instance of gpt model

·  Add few examples to make gpt-3 understand what we are trying to accomplish.

·  Test whether gpt-3 can generate SQL commands for a given English text.

import openai

import json

# Load the API key from the JSON file

with open('GPT-3-SECRET-KEY.json') as f:

    key = json.load(f)

openai.api_key = key["OPEN_API_KEY"]

class GPTSQL:

    def __init__(self, engine="davinci", temperature=0.5, max_tokens=100):

        self.engine = engine

        self.temperature = temperature

        self.max_tokens = max_tokens

        self.examples = []

    def add_example(self, example):

        self.examples.append(example)

    def generate_sql(self, prompt):

        # Combine examples into a formatted string

        examples_str = "\n".join(

            f"###\n{example['english']}\n{example['sql']}" for example in self.examples

        )

        # Construct the final prompt with examples and the new query

        full_prompt = f"{examples_str}\n###\n{prompt}\n"

        response = openai.Completion.create(

            engine=self.engine,

            prompt=full_prompt,

            temperature=self.temperature,

            max_tokens=self.max_tokens,

            top_p=1,

            frequency_penalty=0,

            presence_penalty=0,

            stop=["###"]

        )

        # Extract the generated SQL command

        return response.choices[0].text.strip()

# Example data class (for readability, structuring)

class Example:

    def __init__(self, english, sql):

        self.english = english

        self.sql = sql

    def to_dict(self):

        return {

            "english": self.english,

            "sql": self.sql

        }

# Initialize the GPTSQL object

gpt_sql = GPTSQL(engine="davinci", temperature=0.5, max_tokens=100)

# Add examples to the model

gpt_sql.add_example(Example(

    'Find unique values of DEPARTMENT from Employee table.', 

    'SELECT DISTINCT DEPARTMENT FROM Employee;'

).to_dict())

gpt_sql.add_example(Example(

    "Find the count of employees in department ECE.", 

    "SELECT COUNT(*) FROM Employee WHERE DEPARTMENT = 'ECE';"

).to_dict())

# Function to test the model with a new English query

def test_query(query):

    sql_command = gpt_sql.generate_sql(query)

    print(f"English Query: {query}")

    print(f"Generated SQL: {sql_command}")

# Testing the model with different queries

test_query("What is the total number of employees?")

test_query("List all the departments in the company.")

test_query("Show the names of all employees in the HR department.")

Problem 2 : Information extraction

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

Problem Statement

Steps:

·      Load the key stored in json file

·      Input the passage to the prompt and provide some instructions to gpt-3 in natural language.

·      prompt="Someone asked me to extract information from this passage:\n\"\"\"\nGenerative Pre-trained Transformer 3 (GPT-3) is an autoregressive language model that uses deep learning to produce human-like text. It is the third-generation language prediction model in the GPT-n series (and the successor to GPT-2) created by OpenAI, a San Francisco-based artificial intelligence research laboratory.[2] GPT-3's full version has a capacity of 175 billion machine learning parameters. GPT-3, which was introduced in May 2020, and was in beta testing as of July 2020,[3] is part of a trend in natural language processing (NLP) systems of pre-trained language representations.[1] Before the release of GPT-3, the largest language model was Microsoft's Turing NLG, introduced in February 2020, with a capacity of 17 billion parameters or less than a tenth of GPT-3s.[4]\n\"\"\"\nI noted down few points after reading the passage:\n\"\"\"\n" 

Test whether gpt-3 can extract information from a given passage.

import openai

import json

# Load the API key from the JSON file

with open('GPT-3-SECRET-KEY.json') as f:

    key = json.load(f)

openai.api_key = key["OPEN_API_KEY"]

# Define the passage

passage = """

Generative Pre-trained Transformer 3 (GPT-3) is an autoregressive language model that uses deep learning to produce human-like text. It is the third-generation language prediction model in the GPT-n series (and the successor to GPT-2) created by OpenAI, a San Francisco-based artificial intelligence research laboratory.[2] GPT-3's full version has a capacity of 175 billion machine learning parameters. GPT-3, which was introduced in May 2020, and was in beta testing as of July 2020,[3] is part of a trend in natural language processing (NLP) systems of pre-trained language representations.[1] Before the release of GPT-3, the largest language model was Microsoft's Turing NLG, introduced in February 2020, with a capacity of 17 billion parameters or less than a tenth of GPT-3s.[4]

"""

# Define the prompt with instructions

prompt = f"Someone asked me to extract information from this passage:\n\"\"\"\n{passage}\n\"\"\"\nI noted down few points after reading the passage:\n\"\"\"\n"

# Make the API call

response = openai.Completion.create(

    engine="davinci",

    prompt=prompt,

    temperature=0.3,

    max_tokens=150,

    top_p=1.0,

    frequency_penalty=0.0,

    presence_penalty=0.0,

    stop=["\"\"\""]

)

# Extract and print the response

extracted_info = response.choices[0].text.strip()

print("Extracted Information:")

print(extracted_info)

Problem 3: Recipe Generator

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

Problem Statement

Problem: Create a recipe from list of ingredients.

Steps:

1. Import the necessary packages (openai, json).

2. Store your private API key in a JSON file and load it in ‘openai.api_key’.

3. Inside the prompt type we have to provide the list of ingredients list on which recipe will be generated.
Here, for instance, we have taken the example of the Caramel Custard.

4. To view the result of your prompt, you need to drill down to the “text” key.

import openai

import json

# Import and load the secret key from a JSON file

with open('GPT-3-SECRET-KEY.json') as f:

    key = json.load(f)

openai.api_key = key["OPEN_API_KEY"]

# Define the ingredients and the prompt

ingredients_list = """

Caramel Custard

Ingredients:

- Milk

- Custard Powder

- Sugar

- Milkmaid

Steps to be followed:

"""

# Make the API call to generate the recipe

response = openai.Completion.create(

    engine="davinci-instruct-beta",

    prompt=f"Write a recipe based on these ingredients and instructions:\n{ingredients_list}",

    temperature=0,

    max_tokens=150,

    top_p=1.0,

    frequency_penalty=0.5,

    presence_penalty=0.5

)

# Print the generated recipe

print("Generated Recipe:")

print(response['choices'][0]["text"].strip())

Problem 4: Essay outline completion

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

Problem Statement

Steps:

1. Import the necessary packages (openai, json).

2. Store your private API key in a JSON file and load it in ‘openai.api_key’.

3. Inside the prompt type in an essay topic on which you want your outline to be generated.
Here, for instance, we have taken the example of the American legendary Metal band Metallica.

4. To view the result of your prompt, you need to drill down to the “text” key.

import openai

import json

# Load the API key from a JSON file

with open('secret_key.json') as f:

    key = json.load(f)

openai.api_key = key["secret_key"]

# Define the essay topic and prompt

prompt = "Create an outline for an essay about Metallica and their contribution to metal music:\n\nI: Introduction"

# Make the API call to generate the essay outline

response = openai.Completion.create(

    engine="davinci",

    prompt=prompt,

    temperature=0.7,

    max_tokens=150,

    top_p=1.0,

    frequency_penalty=0.0,

    presence_penalty=0.0

)

# Print the generated outline

print("Generated Essay Outline:")

print(response['choices'][0]["text"].strip())

Problem 5 : Product ad Generator

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

Problem Statement

Turn a product description into ad copy

Steps:

1. Import the necessary packages (openai, json). 

2. Store your private API key in a JSON file and load it in ‘openai.api_key’. 

3. Inside the prompt, type in the description of the product you want to generate an advertisement for
   Here, for instance, we have taken the example of an offroad motorbike XR350. We will let the system design an advertisement for this bike. 

4. To view the result of your prompt, you need to drill down to the “text” key.

import openai

import json

# Load the API key from a JSON file

with open(r'secret_key.json') as f:

    key = json.load(f)

openai.api_key = key["secret_key"]

# Define the product description and the prompt

product_description = """

XR350 motorcycle for offroad and mountain riders. High performance 349cc engine. Available in 3 color segments. Customization available.

"""

prompt = f"""Write a creative ad for the following product to run on Instagram:

\"\"\"\"\"\"

{product_description}

\"\"\"\"\"\"

This is the ad I wrote for Instagram aimed at motorbiking lovers:

\"\"\"\"\"\"

"""

# Make the API call to generate the ad copy

response = openai.Completion.create(

    engine="davinci-instruct-beta",

    prompt=prompt,

    temperature=0.8,

    max_tokens=150,

    top_p=1.0,

    frequency_penalty=0.0,

    presence_penalty=0.0,

    stop=["\"\"\"\"\"\""]

)

# Print the generated ad copy

print("Generated Ad Copy:")

print(response['choices'][0]["text"].strip())

Problem 6 : Interview question generation

more_vert

fullscreen

read_more

arrow_back_ios

arrow_forward_ios

Problem Statement

Steps:

1. Import the necessary packages (openai, json).

2. Store your private API key in a JSON file and load it in ‘openai.api_key’.

3. Inside the prompt type in an essay topic on which you want your outline to be generated.

4. To view the result of your prompt, you need to drill down to the “text” key.

import openai

import json

# Load the API key from a JSON file

with open('secret_key.json') as f:

    key = json.load(f)

openai.api_key = key["secret_key"]

# Define the prompt for generating interview questions

prompt = "Create a list of questions for my interview with a data scientist:\n\n1."

# Make the API call to generate the list of interview questions

response = openai.Completion.create(

    engine="davinci-instruct-beta",

    prompt=prompt,

    temperature=0.8,

    max_tokens=200,

    top_p=1.0,

    frequency_penalty=0.0,

    presence_penalty=0.0,

    stop=["\n\n"]

)

# Print the generated questions

print("Generated Interview Questions:")

print(response['choices'][0]["text"].strip())

The range of values logit_bias parameter can take is given by: -1 to +1 -100 to +100 -5 to +5 None of the above Reset

ChatGPT said:

ChatGPT

The correct answer is -100 to +100. The logit_bias parameter can take values in this range, where higher values significantly increase or decrease the probability of specific tokens being generated. If you need more information about how this parameter works, feel free to ask!

4o

You said:

How many tokens are currently supported by Open Al for the combined text prompt and generated completion? 1024 2048 4096 Unlimited Reset

ChatGPT said:

ChatGPT

The current maximum number of tokens supported by OpenAI for the combined text prompt and generated completion is 4096 for models like GPT-3.5 Turbo. Some models like GPT-4 have larger limits, up to 8192 or 32,768 tokens, depending on the model variant.

4o

You said:

Choose the correct order for steps for positional encoding- 1. For every even index, create a vector using the sin function. 2. For every odd index on the input vector, create a vector using the cos function. 3. Add these vectors to their corresponding input embeddings. Ο 1,23 2,1,3 3.1.2 1,3,2 Reset

ChatGPT said:

ChatGPT

The correct order for steps for positional encoding is:

1, 2, 3

• 1: For every even index, create a vector using the sin function.

• 2: For every odd index on the input vector, create a vector using the cos function.

• 3: Add these vectors to their corresponding input embeddings.

Therefore, the answer is 1, 2, 3.

4o

You said:

What characteristics should this encoding have? Questio it should output a unique encoding for each time-step (word's position in a sentence) Distance between any two time-steps should be consistent across sentences with different lengths. Our model should generalize to longer sentences without any efforts. Its values should be bounded. it must be deterministic All of the above Reset

ChatGPT said:

ChatGPT

The correct answer is All of the above. Positional encoding in models like the Transformer should have the following characteristics:

1. It should output a unique encoding for each time-step (word's position in a sentence).

2. The distance between any two time-steps should be consistent across sentences with different lengths.

3. Our model should generalize to longer sentences without any efforts. Its values should be bounded.

4. It must be deterministic.

These characteristics ensure that the model effectively captures the relative positions of tokens in the input sequence and can generalize well to sequences of varying lengths.

4o

You said:

State True/False In Adidases to the playground Oberal supports users to teract with the AP through HTTP requisits from different programming languages like Python, CN Crystal Dart Go, μια μενα επpuNode, Ruby and Unity True Fate Save

ChatGPT said:

ChatGPT

The statement is True. OpenAI supports users to interact with the API through HTTP requests from different programming languages like Python, C#, Crystal, Dart, Go, Java, Node, Ruby, and Unity, among others. This allows developers to integrate AI capabilities into their applications using various programming languages and frameworks.

4o

You said:

What is the use of softmax function in decoder part of model? It helps to map the sequences of different length to each other It helps to compute the context vector. It helps to find out the highest probability word from the vocabulary All of the above Reset Jave

ChatGPT said:

ChatGPT

The correct answer is:

It helps to find out the highest probability word from the vocabulary.

In the decoder part of the model, the softmax function is used to convert the output logits into a probability distribution over the vocabulary. This allows the model to select the word with the highest probability as the next word in the sequence.

4o

You said:

Complete the following sentence - In positional encoding, 8 12 Owe enhance the model's input to inject the order of words. words that are used in similar ways will be placed close together. it is not possible to create positions for more than 10,000 words. None of the above Reset Save

ChatGPT said:

ChatGPT

The best completion for the sentence is:

Owe enhance the model's input to inject the order of words.

In positional encoding, the technique is used to provide information about the relative or absolute position of words in a sequence, thereby enhancing the model's ability to understand the order of words.

4o

You said:

Which of the following characteristics of Transformers which makes them more efficient compared to other Seq2Seq techniques are: a. Attention mechanism b. Parallel architecture c. Long short memory d. Transfer learng Only a Both a and b Oabd All of the above Reset

ChatGPT said:

ChatGPT

The correct answer is Both a and b.

Explanation:

• a. Attention mechanism: Allows the model to focus on specific parts of the input sequence, improving efficiency and effectiveness in handling longer sequences.

• b. Parallel architecture: Enables the processing of sequences in parallel rather than sequentially, making Transformers more computationally efficient.

c. Long short memory: This is related to LSTM networks, not Transformers.

d. Transfer learning: While Transformers can benefit from transfer learning, this is a general concept applicable to various types of models, not a specific characteristic of Transformers that makes them more efficient compared to other Seq2Seq techniques.

Therefore, the correct characteristics are a and b.

4o

You said:

Which of the following is the official programming language suppo by OpenAI? O Ruby Python O Java Unity Reset

ChatGPT said:

ChatGPT

The official programming language supported by OpenAI is Python.

While OpenAI's API can be accessed using many different programming languages, their official SDKs, libraries, and most documentation are primarily written and provided in Python.

4o

You said:

What are the attention vectors used by self-attention layer in each decoder? 8 12 Key Vector Query Vector Value Vector All of the above Reset

ChatGPT said:

ChatGPT

The correct answer is All of the above.

Explanation:

In the self-attention layer of each decoder in the Transformer model, the following attention vectors are used:

1. Key Vector: Represents the features of each input position and is used to match with the Query vectors.

2. Query Vector: Represents the features of each position where attention is being computed, used to query the Key vectors.

3. Value Vector: Provides the actual information to be attended to, associated with each Key vector.

These three vectors are critical components of the self-attention mechanism in Transformer models.

4o

You said:

The benefit of using sin and cos functions for positional encoding is - by controlling the frequency parameters of these functions, we can manipulate them between [0,1] according to our requirements manipulate them between [0,180] according to our requirements None of the above by controlling the frequency parameters of these functions, we can by controlling the frequency parameters of these functions, we can manipulate them between [-1,1] according to our requirements 11 12 14 15 Reset

ChatGPT said:

ChatGPT

The correct answer is:

by controlling the frequency parameters of these functions, we can manipulate them between [-1,1] according to our requirements.

In positional encoding, sine and cosine functions are used with varying frequencies to generate unique and bounded positional encodings for each position in the sequence. These functions produce values within the range of [−1,1][-1, 1][−1,1], which helps in distinguishing positions while ensuring the values remain within this range.

4o mini

You said:

Which parameter helps in modifying the likelihood of a specific token in the completion text generated by GPT-3 API? logit bias Omax tokens Ologprobs Otemperature Reset

ChatGPT said:

ChatGPT

The parameter that helps in modifying the likelihood of a specific token in the completion text generated by the GPT-3 API is logit bias.

Explanation:

• Logit Bias: Allows you to adjust the probability of specific tokens by applying a bias to their logits, effectively increasing or decreasing their likelihood of being generated.

• Max Tokens: Controls the maximum length of the generated text but does not directly modify token likelihood.

• Logprobs: Provides the log probabilities of tokens but does not modify them.

• Temperature: Controls the randomness of the output by scaling the logits, affecting the probability distribution but not specific token biases directly.

4o mini

You said:

Complete the following sentence with appropriate option-Within the high-dimensional semantic space, Question 13 words that are used in similar ways will be placed close together. the vectors/points representing similar words will cluster together. Othe distance between vector representation of similar words will be low. Any of the above can be used. Reset

ChatGPT said:

ChatGPT

The appropriate completion for the sentence is:

Any of the above can be used.

In a high-dimensional semantic space, the following statements are true:

• Words that are used in similar ways will be placed close together.

• The vectors/points representing similar words will cluster together.

• The distance between vector representations of similar words will be low.

All these statements describe aspects of how semantic similarity is represented in high-dimensional spaces used in word embeddings and similar models.

4o mini

You said:

With the help of sinusoidal function in positional encoding, our model can attend to easily. End positions Relative Positions Top positions None of the above 12 Reset

ChatGPT said:

ChatGPT

The correct answer is:

Relative Positions

Sinusoidal functions in positional encoding help the model to attend to relative positions in the sequence, as they provide unique and continuous representations of each position that allow the model to infer the relative distance between different positions.

4o mini

You said:

Transformer model pays more attention to the most relevant and important word in the sentence. 12 True False Reset

ChatGPT said:

ChatGPT

True

The Transformer model uses the attention mechanism to focus on the most relevant and important words in the sentence. This allows the model to weigh different parts of the input sequence differently based on their relevance, improving its ability to capture contextual relationships and dependencies.

4o mini