lesson1Title

lesson2Title

lesson3Title

lesson4Title

lesson5Title

lesson6Title

lesson7Title

lesson8Title

lesson9Title

lesson10Title

aiFineTuningBasicsChapter3Title

lesson11Title

lesson12Title

lesson13Title

aiFineTuningBasicsChapter1Title

aiFineTuningBasicsChapter2Title

### What is the most appropriate word to fill in the blank?

# Comparing Hyperparameters Across Fine-tuning Models

Let's summarize what we've learned so far and examine how hyperparameter settings impact model training and performance.

 

## Overview of Hyperparameters

| Hyperparameter | When Low | When High |
|-------------------|-------------------------------------------------------------|-------------------------------------------------------------|
| Learning Rate | *Advantage:* Stable learning *Disadvantage:* Slow progress | *Advantage:* Fast learning *Disadvantage:* May overshoot optimal weights |
| Batch Size | *Advantage:* Frequent weight updates, less computational resources *Disadvantage:* Can slow down training | *Advantage:* Faster training *Disadvantage:* Increased overfitting risk, higher computational resources |
| Number of Epochs | *Advantage:* Can prevent overfitting *Disadvantage:* May lead to underfitting and low performance | *Advantage:* Improved performance through more training *Disadvantage:* Risk of overfitting, poor performance on new data |

 

## Break: Experience Fine-tuning Models

The two models are AI models fine-tuned with Southern and Northern American English contexts.

Compare the responses of each model.

 

## Practice

1. *Copying*: Click the `Copy` button at the top-right corner of the prompt below to copy it.

```text title="Example Prompt"
Hello there!
```

2. *Entering the Prompt*: Paste the copied prompt into the input box and press `Enter`.

3. *Compare the Results*: Review the responses of both AI models.

Hyperparameter	When Low	When High
Learning Rate	Advantage: Stable learning Disadvantage: Slow progress	Advantage: Fast learning Disadvantage: May overshoot optimal weights
Batch Size	Advantage: Frequent weight updates, less computational resources Disadvantage: Can slow down training	Advantage: Faster training Disadvantage: Increased overfitting risk, higher computational resources
Number of Epochs	Advantage: Can prevent overfitting Disadvantage: May lead to underfitting and low performance	Advantage: Improved performance through more training Disadvantage: Risk of overfitting, poor performance on new data