Closed lukmanulhakeem97 closed 3 months ago
Hi, you can pass the batch
argument in train: https://github.com/KindXiaoming/pykan/blob/master/kan/KAN.py#L761
thanks @KindXiaoming, got it 👍
@KindXiaoming can you explain how does the batch argument works in case of KANs. It seems only training and testing on the batch size specified per step. Or i got it wrong?
@KindXiaoming
Any thoughts appreciated
I am still having problems OutOfMemoryError with large datasets. Is it possible to modify the code to use Dataloader and pass train_loader and test_loader to train instead of dataset?
So edit this line
def train(self, dataset, opt="LBFGS", steps=100, ......
@sparcycram please try the speed mode model = model.speed()
. tutorial: https://github.com/KindXiaoming/pykan/blob/master/tutorials/Example_2_speed_up.ipynb. You need to update to the most recent version 0.2.0
While training KAN using entire samples say 10000, end up in memory overflow issue. Instead of giving entire data at once on "dataset" dict, can we train KAN on small batches of data like normally do on ANN?