FlexTensor is a tensor offloading and management library for PyTorch that enables running large models on limited GPU memory by intelligently offloading tensors between GPU and CPU memory.
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).