Explain the role of Tensor Cores and mixed-precision training (e.g., using FP16 or bfloat16) in accelerating the training of large generative AI models.

• A.Tensor Cores perform specialized matrix multiplications optimized for lower-precision data types, enabling faster computation and reduced memory footprint.
• B.Mixed-precision training allows using lower precision for forward and backward passes but keeps weights and gradients in higher precision to maintain stability.
• C.Tensor Cores are only useful for inference, not training.
• D.Mixed-precision training guarantees the same convergence behavior as full-precision training.
• E.A and B.

Comment: *

Name: *

Email: *

Verification: *

Consider the following Python code snippet utilizing the Hugging Face Transformers library for multimodal processing. The objective is to perform visual question answering (VQA). Assume 'image' is a PIL Image object and 'question' is a string. However, the code is incomplete. Choose the options to complete the code.

• A.
• B.
• C.
• D.
• E.

Comment: *

Name: *

Email: *

Verification: *

In ML applications, which machine learning algorithm is commonly used for creating new data based on existing data?

• A.Decision tree
• B.Support vector machine (SVM)
• C.K-means clustering
• D.Generative adversarial network (GAN)

Comment: *

Name: *

Email: *

Verification: *

Consider a scenario where you are developing a multimodal A1 system to translate sign language videos into text. The system utilizes a CNN for processing video frames and an RNN for generating the text sequence. During evaluation, you observe that the system struggles to accurately translate signs that involve complex hand movements or subtle facial expressions. What are the MOST effective strategies to improve performance in this specific scenario? (Select TWO)

• A.Reduce the frame rate of the input video to decrease computational load.
• B.Increase the depth of the CNN to capture more fine-grained visual features.
• C.Incorporate a 3D CNN architecture to better capture spatial-temporal information from the video.
• D.Replace the RNN with a simpler feed forward neural network for faster processing.
• E.Augment the training data with variations of the sign language videos, including different camera angles, lighting conditions, and signers.

Comment: *

Name: *

Email: *

Verification: *

You are tasked with building a system that can generate captions for images. You want to use a transformer-based model. During inference, you notice that the model tends to generate repetitive captions. Which of the following decoding strategies could you use to mitigate this issue?

• A.Greedy decoding
• B.Beam search with a high beam width
• C.Beam search with a length penalty
• D.Random sampling
• E.Top-k sampling

Comment: *

Name: *

Email: *

Verification: *