SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training

The efficiency of attention is important due to its quadratic time complexity. We enhance the efficiency of attention through two key contributions: First, we leverage the new FP4 Tensor Cores in Blackwell GPUs to accelerate attention computation. Our implementation achieves 1038 TOPS on RTX5090, which is a 5× speedup over the fastest FlashAttention on RTX5090. Experiments show that our FP4 attention can accelerate inference of various models in a plug-and-play way. Second, we pioneer low-bit attention to training tasks. Existing low-bit attention works like FlashAttention3 and SageAttention focus only on inference. However, the efficiency of training large models is also important. To explore whether low-bit attention can be effectively applied to training tasks, we design an accurate and efficient 8-bit attention for both forward and backward propagation. Experiments indicate that 8-bit attention achieves lossless performance in fine-tuning tasks but exhibits slower convergence in pretraining tasks. The code is available at https://github.com/thu-ml/SageAttention. 5x 212 1038 3x 490s 164s FlashAttention2 (End-to-End Time: 490s on RTX5090) SageAttention3 (End-to-End Time: 164s on RTX5090) Figure 1: The upper left figure shows the kernel speedup on RTX5090. The other two figures show the end-to-end inference speedup of generating a video using HunyuanVideo on RTX5090. Note that FlashAttention3 can only run on Hopper GPUs, so FlashAttention2 is already the fastest on RTX5090.