Conversation with Merlin [email protected] · Fri Jan 05 2024

Choose the answer that correctly fill in the blanks.

prediction request, prediction response

tunable request, completion

prompt, fine-tuned LLM

prompt, completion

The correct answer to fill in the blanks is "prompt, completion."

Text summarization

Information Retrieval

Translation

Invoke actions from text

The task that supports the use case of converting code comments into executable code is "Invoke actions from text."

A measure of how well a model can understand and generate human-like language.

A mechanism that allows a model to focus on different parts of the input sequence during computation.

The ability of the transformer to analyze its own performance and make adjustments accordingly.

A technique used to improve the generalization capabilities of a model by training it on diverse datasets.

The correct answer is: "A mechanism that allows a model to focus on different parts of the input sequence during computation

Defining the problem and identifying relevant datasets.

Manipulating the model to align with specific project needs.

Performing regularization

Selecting a candidate model and potentially pre-training a custom model.

Deploying the model into the infrastructure and integrating it with the application.

The stages that are part of the generative AI model lifecycle mentioned in the course are:

  • Defining the problem and identifying relevant datasets.
  • Performing regularization.
  • Selecting a candidate model and potentially pre-training a custom model.
  • Deploying the model into the infrastructure and integrating it with the application.

Is this true or false?

False.

Autoencoder

Sequence-to-sequence

Autoregressive

The transformer-based model architecture with the objective of guessing a masked token based on the previous sequence of tokens is "Autoregressive."

Autoencoder

Sequence-to-sequence

Autoregressive

The transformer-based model architecture well-suited to the task of text translation is "Sequence-to-sequence".

True

False

False. Increasing the model size is not always necessary to improve its performance. Other factors such as data quality, training duration, and optimization methods can also significantly impact model performance.

Model size: Number of parameters

Batch size: Number of samples per iteration

Compute budget: Compute constraints

Dataset size: Number of tokens

The alternatives that should be considered for scaling when performing model pre-training are:

  • Model size: Number of parameters
  • Batch size: Number of samples per iteration
  • Compute budget: Compute constraints
  • Dataset size: Number of tokens

Is this true or false?

True