Consider the sentence:
"Data science is fun"
In a bigram language model, which probability expression correctly represents the probability of
the word "is"?
A.
P(is | Data, science)
B.
P(is | science)
C.
P(is | fun)
D.
P(is | science, fun)
A published solution is not available for this question yet.
We are given the Q, K, V matrices to compute the scaled dot product attention matrix for a
transformer.
Keeping everything else same we double every value in the V matrix.
How will this change the scaled dot product attention scores that are computed?
A.
There won't be any change to the attention scores.
B.
Attention scores will double.
C.
Attention scores will get halved.
D.
Attention scores will change but we can't determine how much they will
change without checking the actual values of the V matrix.
A published solution is not available for this question yet.
Consider an attention mechanism defined as:
Attention(Q, K, V) = softmax(QK\(^{T}\)) V
Which of the following is the most likely consequence of removing the scaling factor in this
attention computation?
A.
The attention scores become smaller, leading to uniform attention weights.
B.
The dot product values grow large, causing the softmax to produce extremely
peaked distributions and unstable gradients.
C.
The model becomes invariant to the dimensionality of key vectors.
D.
The attention mechanism works as usual.
A published solution is not available for this question yet.
You are training an RNN with batch_size=32 on a dataset of 1050 sequences. You notice an error
occurs at the end of every epoch. Which DataLoader fix resolves this?
A.
Set shuffle=False
B.
Increase batch_size to 64
C.
Set drop_last=True
D.
Set num_workers=0
A published solution is not available for this question yet.
A transformer model has the following embedding layer:
• Vocabulary size = 50,000
• Embedding dimension = 768
The embeddings are currently stored in **float32** format.
You decide to optimize memory as follows:
1. Quantize **70%** of the embedding vectors to **int8**
2. Keep the remaining **30%** in **float32** (for high-frequency tokens)
What is the **overall percentage reduction in memory** for the embedding matrix?
(Round to the nearest integer)
A.
52%
B.
60%
C.
53%
D.
65%
A published solution is not available for this question yet.
Consider the following optimization algorithms used for training neural networks:
• SGD with Momentum
• RMSProp
• Adam
Which of the following statements is/are correct?
A.
SGD with Momentum accumulates an exponentially decaying moving average
of past gradients to accelerate convergence along consistent directions.
B.
Momentum methods adapt individual learning rates for each parameter
based on past squared gradients.
C.
RMSProp maintains a running average of squared gradients to scale learning
rate.
D.
Adam combines ideas from Momentum and RMSProp by using both first and
second moment estimates of gradients.
E.
Adam requires computing second-order derivatives of the loss function.
F.
Adaptive methods such as RMSProp and Adam can use different effective
learning rates for different parameters.
A published solution is not available for this question yet.
Consider the evaluation metrics commonly used for generative models:
**Fréchet Inception Distance (FID)**, **Inception Score (IS)**, and **CLIP Score**.
Which of the following statements are correct?
A.
FID compares the mean and covariance of real and generated feature
distributions extracted from a pretrained Inception network.
B.
Inception Score directly compares generated images with real images.
C.
A lower FID score indicates that generated samples are closer to the real data
distribution.
D.
Inception Score rewards images that are both classifiable (low entropy
conditional distribution) and diverse (high entropy marginal distribution).
E.
CLIP Score requires real images to compute the metric.
F.
CLIP Score measures alignment between generated images and their
corresponding text prompts.
G.
FID and Inception Score both explicitly evaluate text-image alignment.
A published solution is not available for this question yet.
A text-to-image generative model is evaluated on three metrics with the following results:
• **FID = 8.5**
• **Inception Score (IS) = 2.1**
• **CLIP Score = 0.34**
For comparison, a strong baseline model on the same dataset achieves:
• **FID = 15.2**
• **IS = 6.8**
• **CLIP Score = 0.31**
Which of the following interpretations are correct?
A.
The new model generates images that are statistically closer to the real data
distribution than the baseline.
B.
The new model produces more class-diverse samples than the baseline.
C.
The new model likely generates less classifiable or less confident object
predictions compared to the baseline.
D.
The new model aligns slightly better with the provided text prompts than the
baseline.
E.
The higher CLIP score guarantees better perceptual image quality.
F.
The baseline model likely generates more diverse categories of images than
the new model.
A published solution is not available for this question yet.
You are told that the total number of learnable parameters in this RNN layer is
. Assume the
hidden size is H = 20.
What are the total number of learnable parameters in the below layers if the same configuration
(same input size, same hidden size H = 20,
,
) is used?
1. A Bidirectional GRU
2. A Bidirectional LSTM
A.B.C.D.
A published solution is not available for this question yet.
A model is fine-tuned with the following configuration:
• dataset_size = 2048
• batch_size = 16
• epochs = 2
Assuming the optimizer performs exactly one update after a fixed number of gradient
accumulation steps and 64 updates are performed, how many gradient accumulation steps were
used during training?
A published solution is not available for this question yet.