In the Model properties dioalogue several advanced properties of an OpenAI model can be set:
Model (display only):
- The name of the referenced OpenAI Model. Note, that with every training this name changes
Model is tunable / Model is tuned (display only):
- general capabilities of the underlying Model. Note that only selected OpenAI Models can be tuned (the model is selected when adding an OpenAI language-resource)
Use default system-message when translating:
- defines, if the default system-message of a training (the topmost message in the training-window) is used when pretranslating with the model
Use user defined system-messages when translating:
- defines, if the other user-defined system-messages of a training are used when pretranslating with the model
Generation Sensitivity / Temperature:
- Temperature is a parameter that governs the randomness and thus the creativity of the responses. It is always a number between 0 and 1. A temperature of 0 means the responses will be very straightforward, almost deterministic (meaning you almost always get the same response to a given prompt) A temperature of 1 means the responses can vary wildly. It’s advisable to adjust either the temperature or top_p, but not both.
Probability Threshold / Top P:
The "top P" parameter, also known as nucleus sampling, is a nuanced alternative to temperature-based sampling. It is a "spotlight" that shines on the most probable words. At a default value of 1.0, the model considers all words. This parameter can help control the distribution of word choices, keeping the generated content relevant and coherent. It’s advisable to adjust either the temperature or top_p, but not both.
Presence Penalty:
Frequency Penalty:
Max. target tokens (% of source tokens):