I have fine tuned gpt 4.1 nano with 1050 datasets for a classification specially something sentimental analysis. I set the batch size to 16, lr-mulitplier to 0.1 and epoch to 3.
My results were actually solid but still on evaluation dataset my model does not perform well. 43-44 % accuracy. I u
I want to improve my fine tuned model with more data around 2500 examples.
My question is how would you change the parameter of my model right now? To slowly improve it ?
This is right now my current results:
My idea was it to finetuen my output model to the new data with batch size = 16, lr-multiplier = 0.03 and epoch = 7.
What do you think about this?
thank you very much!
Best regards