Hello. I recently finetuned gpt-4o-2024-08-06 with my custom visual question answering dataset: 1000 images for the training, 200 images for the validation, and five questions for each image.
I just have a question about the validation loss not being plotted anymore at a certain step, something like this.
In case you cannot figure out, the green plot represents the train loss, the purple for the validation loss, and the red dots mean full validation loss.
Does the validation loss stop being shown because of the valid set size, or it is just not showing everything?
Also, I am wondering if the stopping of the plotting is where the overfitting begins, or if the full validation loss dots represent the steps of overfitting.
Thank you.