REVIEW 8 cited by
SageAttention2++: A More Efficient Implementation of SageAttention2
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
SageAttention2++: A More Efficient Implementation of SageAttention2
read the original abstract
The efficiency of attention is critical because its time complexity grows quadratically with sequence length. SageAttention2 addresses this by utilizing quantization to accelerate matrix multiplications (Matmul) in attention. To further accelerate SageAttention2, we propose to utilize the faster instruction of FP8 Matmul accumulated in FP16. The instruction is 2x faster than the FP8 Matmul used in SageAttention2. Our experiments show that SageAttention2++ achieves a 3.9x speedup over FlashAttention while maintaining the same attention accuracy as SageAttention2. This means SageAttention2++ effectively accelerates various models, including those for language, image, and video generation, with negligible end-to-end metrics loss. The code will be available at https://github.com/thu-ml/SageAttention.
Forward citations
Cited by 8 Pith papers
-
Speculative Decoding for Autoregressive Video Generation
A training-free speculative decoding method for block-based autoregressive video diffusion uses a quality router on worst-frame ImageReward scores to accept drafter proposals, achieving up to 2.09x speedup at 95.7% qu...
-
MXAttention: Data-Free Optimal Scaling and Pre-Normalization Quantization for MXFP4 Attention
A closed-form MXFP4 scale boundary Qmax=7.25 plus pre-normalization softmax quantization recovers nearly all FP16 video quality lost by naive MXFP4 attention.
-
CoSA: Accelerating Long-Context Inference via Proxy-Kernel Co-Designed Sparse Attention
CoSA couples an order-aware sparse-mask proxy with a skip-capable attention kernel, achieving higher long-context accuracy at lower compute budgets than existing sparse-attention baselines.
-
Vidu S1: A Real-Time Interactive Video Generation Model
Vidu S1 generates voice-controlled interactive avatar video in real time at 540p/42 FPS with claimed infinite stable streams and top reported quality metrics.
-
Vidu S1: A Real-Time Interactive Video Generation Model
Vidu S1 generates 540p interactive avatars at up to 42 FPS with voice control, using diffusion forcing, DMD distillation, and a serving stack.
-
Video Generation Models as World Models: Efficient Paradigms, Architectures and Algorithms
Video generation models can function as world simulators if efficiency gaps in spatiotemporal modeling are bridged via organized paradigms, architectures, and algorithms.
-
P-Cast Precision in FP8 Attention: Sink-Induced Collapse and the Optimality of S=2^8
Forward KV iteration in FP8 attention produces P-collapse under attention sink; reverse iteration with S=256 removes it and is optimal among bit-exact scales.
-
Video Generation Models as World Models: Efficient Paradigms, Architectures and Algorithms
In twisted bilayer nodal d-wave superconductors, interlayer hopping creates nodes on the C2 axis and Bogoliubov flat bands when the single-layer Berry connection is parallel to that axis.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.