Is there any way around model.fit(x_train, y_train)?

Because x_train is too large to be uploaded on GPU (RTX 3090, 24GB), calling model.fit (x_train, y_train) cannot train the model.

Is there any way around instead of using model.fit (x_train, y_train) ? (e.g. using conventional Pytorch training such as DataLoader with a small batch ?)

Thank you in advance:)

The issue names model.fit(x_train, y_train) and a conventional PyTorch DataLoader with small batches, but no files or tests. Start by locating the model.fit training entry point and its batching documentation; done means establishing and documenting a supported way to train when the full dataset does not fit on the GPU.

Written by the indexing model from the issue text.