Prompt Engineering Flashcards
What is prompt engineering?
Developing, designing, and optimizing prompts to enhance the output of your foundation model.
What are the four main prompt engineering techniques?
Instructions
Context
Input Data
Output Indicator
What is negative prompting?
It helps to avoid unwanted content, maintains focus and clarity. It explicitly states what NOT to include.
How is text generated in an LLM?
By weights or probabilities on the next word.
What is the Model System Prompt configuration?
It dictates how the model should behave and reply
What does the Model Temperature configuration do?
It sets the creativity of the model’s output. Low values are more conservative and repetitive. High values are more diverse and creative, but can be unpredictable and maybe less coherent. 0 is low and 1 is high
What does the Model Top P configuration do?
It sets the word selection of the model. Low values like .25 consider 25% of the most likely words. High values like .99 consider a broader range, possibly more creative and diverse.
What does the Model Top K configuration do?
Limits the number of probably words. Low K values are more coherent. High K more probable words, more diverse and creative.
What does the Model Length configuration do?
Maximum length of the answer
What does the Model Stop sequences configuration do?
Tokens that signal the model to stop generating output.
What is prompt llatency?
How fast the model responds
What impacts prompt latency?
Model size
Model type
Number of tokens in the input
Number of tokens in the ouput
Can Top P, Top K, or Temperature impact model latency?
No.
What is zero-shot prompting?
It presents a task to the model without providing examples or explicit training for that specific task.
What is few-shot prompting?
You provide examples of a task to the model to guide the output.
