How do I further train already trained model

Hi, I’m training effiicient net b0 on custom data set,
I already have weights for models which was trained with 30 epochs.
now what I did is, I loaded the model,

pth = 'model.pth'
model.load_state_dict(torch.load(pth))

and started training for 5 epochs.
My question is, will this model be considered for trained for total of 35 epochs or this will overwrite previous ones and be trained for 5 epochs?

if it will overwrite, then kindly suggest me how can I train it further

Loading a pretrained model for training will further train the model, for example to fine-tune the model (Finetuning from a pretrained model).

Hi indeed it is training and can be seen through the loss and accuracy that it is further improving.
but on submitting the solution, the accuracy on hidden data is dropped.
on straight 30 epochs it was 87 but now after training for 5 more epochs it is dropped to 83.
is there any parameter I am missing ?