gpt-fast
                                
                                 gpt-fast copied to clipboard
                                
                                    gpt-fast copied to clipboard
                            
                            
                            
                        Update to use torch.nn.attention.sdpa_kernel
The main reason I didn't do this previously is worry that this'll cause the code to hard break on older versions. When was this new API added?
It added two months ago (https://github.com/pytorch/pytorch/commit/4e29f01bf2edfe40d6fb8b38228199185645b6af). I understand the concern, don't have strong opinion on changing this.