This video shows the difference between a tuned and non-tuned Gemini AI model with a very simple example.
I hope you like this video. For any questions, suggestions or appreciation please contact us at: https://programmerworld.co/contact/ or email at: programmerworld1990@gmail.com
Steps:
- Create a sample data
- Go to https://aistudio.google.com/app for Tuning the model
Excerpts:
Tuning vs prompting in AI?
Prompting: Concept: Involves crafting specific instructions or prompts that guide the LLM towards the desired response.
Fine-tuning: Retraining the model on a new dataset specific to the desired task.
Prompting is ideal for:
- Scenarios where access to computational resources or large amounts of data is limited.
Fine-tuning is preferred for:
- When you have a large amount of task-specific data available.
Data file used in this Demo:
A snippet of data set and respective prompts are below:
prompt output
1+8 17
9+10 29
4+8 20
2+4 10
4+6 16
2+5 12
6+5 16
2+6 14
8+7 22
3+5 13
5+9 23
9+2 13
5+3 11
7+10 27
5+5 15
5+7 19
4+6 16
10+8 26
3+9 21
4+6 16
4+9 22
8+1 10
5+5 15
9+3 15
6+9 24
2+7 16
1+5 11
4+5 14
6+5 16
6+1 8
2+10 22
6+7 20
6+4 14
3+6 15
10+6 22
10+9 28
10+8 26
10+8 26
4+2 8
6+2 10
4+7 18
9+1 11
4+5 14
1+6 13
6+4 14
5+8 21
5+4 13
10+2 14
4+6 16
Screenshots:
No fine tuning done in the above.
Above is the output from the fine tuned model.