lesson1Title

lesson2Title

lesson3Title

lesson4Title

lesson5Title

lesson6Title

lesson7Title

lesson8Title

lesson9Title

aiFineTuningApplicationChapter3Title

lesson10Title

lesson11Title

lesson12Title

aiFineTuningApplicationChapter1Title

aiFineTuningApplicationChapter2Title

# Setting Hyperparameters

Let's set hyperparameters, which are parameters that significantly affect the performance of training.

![thumbnail-600](https://academy.codefriends.net/assets/ai/fine-tuning/application/hyperparameter-demo.png)

<br />

## Practice

1. **Set Options**: Adjust the slider on the right in the `STEP 2` section or enter numbers directly.

2. Use the table below as a reference to set appropriate hyperparameters.

| Hyperparameter | When Low                            | When High                              |
|----------------|-------------------------------------|---------------------------------------|
| Number of Epochs| Model may not learn sufficiently    | Increased risk of overfitting         |
|                | Faster training time                | Longer training time                  |
| Batch Size     | Lower memory requirements           | Higher memory requirements            |
|                | Unstable learning curve             | Stable learning curve                 |
| Learning Rate  | Slow learning speed                 | Fast learning speed                   |
|                | Increased chance of overfitting     | Increased chance of underfitting      |

When the learning rate is too high, the model may try to optimize too quickly, making it difficult to find the optimal values. This situation is more likely to lead to underfitting.

### Generally, if the learning rate is too high, the probability of overfitting increases.

Setting Hyperparameters

Practice

Generally, if the learning rate is too high, the probability of overfitting increases.

Create a fine-tuned model

JSONL Data

Configure hyperparameters

Hyperparameter	When Low	When High
Number of Epochs	Model may not learn sufficiently	Increased risk of overfitting
	Faster training time	Longer training time
Batch Size	Lower memory requirements	Higher memory requirements
	Unstable learning curve	Stable learning curve
Learning Rate	Slow learning speed	Fast learning speed
	Increased chance of overfitting	Increased chance of underfitting