Free NVIDIA Generative AI LLMs Practice Questions

A.  

Quantization might help in saving power and reducing heat production.

B.  

It consists of removing a quantity of weights whose values are zero.

C.  

It leads to a substantial loss of model accuracy.

D.  

Helps reduce memory requirements and achieve better cache utilization.

E.  

It only involves reducing the number of bits of the parameters.

A.  

Trained without the need for labeled data.

B.  

Smaller latency, higher throughput.

C.  

It is easier to explain the predictions.

D.  

Cheaper computational costs during inference.

E.  

Single generic model can do more than one task.

A.  

Ensuring that CPUs are running at full clock speed.

B.  

Subsampling datasets to provide rapid but approximate answers.

C.  

Using the GPU for parallel processing of data.

D.  

Enabling data processing to scale to multiple GPUs.

E.  

Providing more memory for data analysis.

A.  

It ensures that AI systems are transparent in their decision-making processes.

B.  

It requires AI systems to be developed with an ethical consideration for societal impacts.

C.  

It involves verifying that AI models are fit for their intended purpose according to regional or industry-specific standards.

D.  

It mandates that AI models comply with relevant laws and regulations specific to their deployment region and industry.

A.  

The model weights are synced across all processes/devices only at the end of every epoch.

B.  

A master-worker method for syncing the weights across different processes is desirable due to its scalability.

C.  

A ring all-reduce is an efficient algorithm for syncing the weights across different processes/devices.

D.  

The model weights are kept independent for as long as possible increasing the model exploration.

A.  

Automatic parallelization of Python code across CPU cores.

B.  

GPU-accelerated columnar data processing with zero-copy memory access.

C.  

Integration with cloud-based storage for distributed data access.

D.  

Conversion of Pandas DataFrames to SQL tables for faster querying.

A.  

Inspecting, cleansing, and transforming data

B.  

Performing GPU-accelerated machine learning tasks

C.  

Training deep learning models

D.  

Data visualization and analysis

A.  

Model size

B.  

Accuracy on a validation set

C.  

Training duration

D.  

Number of layers

A.  

Q, K, and V are randomly initialized weight matrices used for positional encoding.

B.  

K is responsible for computing the attention scores between the query and key vectors.

C.  

Q represents the query vector used to retrieve relevant information from the input sequence.

D.  

V is used to calculate the positional embeddings for each token in the input sequence.

A.  

To optimize the model’s inference speed for production deployment.

B.  

To identify and mitigate potential biases, safety risks, and harmful outputs.

C.  

To increase the model’s parameter count for better performance.

D.  

To automate the collection of training data for fine-tuning.