The paper introduces a typed tensor language designed for shared-state federated computations, distinguishing between client-partitioned data and globally available values through two tensor sorts. It demonstrates that typed one-round programs factor through shared tensors whose shapes are program-dependent but client-independent, and extends this construction to multi-round programs with persistent shared state. The language supports server-side first-order updates, curvature-block updates, and covers federated analytics and FedSGD, while excluding more complex schemes like multi-local-step FedAvg and persistent private client state.
arXiv:2604. 09558v2 Announce Type: replace-cross Abstract: With the widening gap between compute and memory operation latencies, data movement optimizations have become increasingly important for DNN compilation.
arXiv:2512. 22219v2 Announce Type: replace-cross Abstract: We introduce Mirage Persistent Kernel (MPK), the first compiler and runtime system that automatically transforms multi-GPU model inference into a single high-performance mega-kernel.
arXiv:2511. 10480v3 Announce Type: replace-cross Abstract: Optimizing the performance of large language models (LLMs) on large-scale AI training and inference systems requires a scalable and expressive mechanism to model distributed workload execution.
This survey reviews tensor methods applied to large language models, framing them through a seven‑stage lifecycle (tokenization, embeddings, pre‑training, adaptation, compression, inference, interpretability) and a component view (embeddings, attention, feed‑forward networks). It offers unified notation, theoretical foundations, and comparative analyses of tensorization strategies for Transformer components, while highlighting evaluation protocol differences and model scale effects. The paper also introduces a new metric, ρ_gap, to quantify the gap between theoretical memory savings and actual system‑level speedup, and connects tensor techniques to related efficiency and probabilistic methods.
DanLing NestedTensor is a PyTorch tensor abstraction that embeds multi‑ragged structure directly into the tensor, allowing packed values to carry partition information and logical dimension order. This design enables broadcasting, feature transformations, and reductions to automatically respect ragged axes while preserving the same representation through autograd and both eager and compiled execution. Benchmarks on an A100 show significant speedups—up to 3.39× over padding for BERT models and 2.40–4.32× for a Pairformer‑style workload—while dramatically reducing peak memory usage.
Large language models (LLMs) are built from structured high-dimensional objects such as token representations, weights, adaptation updates, caches, and activations, whose multilinear structure is unde...
The paper proposes a new federated learning approach called FedALS that reduces communication costs by varying aggregation frequencies across model layers. It derives tighter generalization bounds for one‑round and multi‑round federated learning, linking these bounds to local updates and data heterogeneity. Based on representation‑learning insights, the authors argue that infrequent aggregation of early layers and more frequent aggregation of final layers yields more generalizable models, especially in non‑iid settings, and demonstrate the method’s effectiveness experimentally.
The paper explores using large language models (LLMs) to replace traditional compiler backends, a process termed AI lowering. An LLM agent translates Triton kernels directly into NVIDIA PTX, achieving 0.83x–3.34x the performance of autotuned Triton on a variety of GPUs and ML kernels. The study also extends a PTX verifier to support modern GPU features, highlighting the potential for AI compilers to reduce engineering effort for new hardware.
arXiv:2607. 07565v1 Announce Type: cross Abstract: One-shot federated learning (OSFL) addresses the communication overhead of federated learning by limiting training to a single round, but doing so without sacrificing model quality is non-trivial, particularly when client data distributions diverge.
arXiv:2606. 09312v1 Announce Type: new Abstract: Tensor program optimization is essential for modern machine learning systems, but its search space is enormous.
arXiv:2607. 29071v1 Announce Type: cross Abstract: Federated learning of foundation models faces a fundamental resource-asymmetry challenge: the institutions holding the most valuable domain-specific data cannot host billion-parameter models.
TensorCommitments (TCs) is a lightweight, tensor-native proof‑of‑inference scheme that enables verifiable inference for large language models (LLMs) without requiring the verifier to rerun the model or possess a powerful GPU. By binding each inference to a commitment stored in multivariate Terkle Trees, TCs detect tampering with only a 0.97% overhead for the prover and 0.12% for the verifier on LLaMA2. The approach improves robustness against tailored LLM attacks by up to 48% compared to previous methods that needed a verifier GPU.