Code for paper: [ICLR2025 Oral] FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).