TurboPairFormer: Fast and Stable Protein Folding Model Training with an Optimized Triangle Attention Kernel
cs.LG
Submitted: 2026-10-05
Updated: 2026-10-05
Code: https://github.com/latkins/trifast
Project page: https://pypi.org/project/turbopairformer
Sources
- FastFold: Reducing AlphaFold Training Time from 11 Days to 67 Hours
- Leveraging the bfloat16 Artificial Intelligence Datatype For Higher-Precision Computations
- MegaFold: Efficient Training of Next-Generation 3D Attention Protein Models on Cross-Platform GPUs
- Online normalizer calculation for softmax
- NeuralPLexer3: Accurate Biomolecular Complex Structure Prediction with Flow Models
- Self-attention Does Not Need $O(n^2)$ Memory
- DeepSpeed4Science Initiative: Enabling Large-Scale Scientific Discovery through Sophisticated AI System Technologies
- FlashAttention-4: Algorithm and Kernel Pipelining Co-Design for Asymmetric Hardware Scaling
- Attn-QAT: 4-Bit Attention With Quantization-Aware Training
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks