Model Settings
This guide provides detailed explanations of the available settings to help you customize your AI experience. Each setting includes descriptions, default values, recommended values.
Last updated
This guide provides detailed explanations of the available settings to help you customize your AI experience. Each setting includes descriptions, default values, recommended values.
Last updated
Description: Temperature is a hyperparameter that adjusts the randomness of the model's predictions. It controls the creativity and variability of the AI's responses.
Low Temperature (0.0 - 0.5): The model becomes more deterministic, favoring the most likely next tokens. This setting is ideal for tasks requiring precise and factual responses, as it reduces variability in outputs.
High Temperature (1.0 - 2.0): The model generates more diverse and creative outputs by allowing less probable tokens to be selected more frequently. This setting is beneficial for creative tasks like storytelling or poetry, where variety is desired.
Values:
Option
Value Range
Description
Default Value
0.8
Balances creativity and coherence, suitable for general use cases.
Recommended Value
0.7 - 1.2
Provides a good mix of creativity and accuracy for most applications.
Disable
0
Description: Repetition penalty is a technique used to discourage the model from producing repetitive or redundant output. It reduces the likelihood of generating tokens that have appeared recently in the generated text, promoting more diverse and varied responses.
Values:
Option
Value Range
Description
Default Value
1.08
Slightly discourages repetition, maintaining a balance between diversity and coherence.
Recommended Value
1 - 1.4
Enhances diversity by more strongly penalizing repeated phrases or tokens.
Disable
1.0
Description: Min-P sampling sets a minimum threshold for the probability of selecting a token during generation. Tokens with probabilities below this threshold are excluded from selection, helping to maintain coherence while allowing some degree of randomness.
Values:
Option
Value Range
Description
Default Value
0.06
Provides a baseline threshold that balances coherence and creativity.
Recommended Value
0.04 - 0.1
Adjusts the strictness of token selection to fine-tune the desired level of randomness and coherence.
Disable
0
Removes the minimum probability threshold, allowing the model to consider all tokens regardless of their probability.
1. Experimentation is Key
Don't hesitate to try different combinations of settings to see how they affect your outputs. Each task may require different configurations.
2. Use Gradual Adjustments
Make small changes to one parameter at a time, so you can clearly observe its impact on the generated text.
3. Reset to Default Settings
If you feel that customizing the settings has made the responses worse, you can always revert to the default settings by clicking the "Reset to Default Settings" button.
4. Review and Iterate
After generating text, review it critically. If it doesn't meet your expectations, adjust your parameters accordingly and regenerate.
5. Document Your Findings
Keep track of the settings that work well for specific tasks so you can replicate successful configurations in the future.
Setting
Description
Default Value
Recommended Value
Disable Option
Temperature
Adjusts the randomness of the model's predictions.
0.8
0.7 - 1.2
0
Repetition Penalty
Discourages repetitive or redundant output.
1.08
1 - 1.4
1.0 (Disabled)
Min-P
Sets a minimum threshold for the probability of selecting a token.
0.06
0.04 - 0.1
0 (No Threshold)
By understanding and utilizing these settings effectively, you can tailor the AI's behavior to better suit your specific needs and preferences. Happy customizing! For more help you can join our discord:
