ggml-org/llama.cpp: b10731
qwen4exp: support recurrent state rollback (#28123) MTP speculative decoding needs the target state to move back by the number of rejected draft tokens. Without rollback support the context is classified as…
Topic - Edition
AI
ggml-org/llama.cpp: b10730
qwen4exp: sum the indexer heads by slices (#28023) qwen4exp: sum the indexer heads by slices The head reduction went through a transpose and a sum_rows over ne[1], which left sum_rows with ne0 = 4, one block per row for…
pytorch/pytorch: trunk/77849518d146b9938817c4c4455c11798693eefb: [BE] Remove obsolete `isLowPower` check (#195365)
Which only made sense on Intel Macs, which have not been supported for quite some time. Discovered while compiling on MacOS-27, where this function is deprecated. Pull Request resolved: #195365 Approved by:…
pytorch/pytorch: trunk/96043a96360267ded157dfe37c3d8c2a197b0e47: Update Metamates merge rule (#195520)
This PR is auto-generated nightly by this action. Updates the Metamates merge rule with active team members. Pull Request resolved: #195520 Approved by: https://github.com/pytorchbot
pytorch/pytorch: trunk/42f04f29182a6cd7a72a493e9cca1136756cc312: [torch] Make singletons immortal (#194371) (#194371)
Summary: Every value in the torch namespace is a process-lifetime singleton, and on a free-threaded build every one of them is an ordinary mortal PyObject. Loading a mortal object owned by another thread takes the…
pytorch/pytorch: trunk/62146ea0c777ac1a6fcc637a647335b06fba3019: [xpu][fix]Relax fp16 grad tolerance for combinations on XPU (#195403)
Motivation #189102 bumped the gradient tolerances for test_comprehensive_combinations_cuda_float16 after #186595 / #189305 reimplemented torch.combinations on top of index_put, whose fp16 backward Inductor accumulates…
pytorch/pytorch: trunk/3c512baeddc8fdc688d905a33baf89d97ebb050e: Dispatch `index_put_` to `masked_fill_.Scalar` when possible (#188330)
This is a followup on #188267 and #61612 and tries to dispatch index_put_ to masked_fill_.Scalar where possible - e.g. gpu_tensor[mask] = 1.0 - instead of converting the scalar to a CPU tensor first. While this prevents…
pytorch/pytorch: trunk/747833341c75bcb9ed1cf447ad5103ec5b87e6dd
[ROCm] Fix invalid max_seqlen in test_dummy_mha_with_nt, unskip NT te…
pytorch/pytorch: trunk/64417d6eb27cc17a97e8aa2172ffe60f7843fdc4
[CUDA] Fix out-of-bounds/invalid-gradient in layer_norm's huge-M gamm…
pytorch/pytorch: trunk/378040143b63049e5025611bb606ffa18c6582ed: [cuDNN] Upgrade to cuDNN 9.25.1 (#195492)
includes fix for #193893 authored with codex Pull Request resolved: #195492 Approved by: https://github.com/Skylion007
pytorch/pytorch: viable/strict/1788240474: Fix stack buffer overflow in libshm Unix socket path handling (#195183)
Summary Fixes #195182 Replace unbounded strcpy() in Socket::prepare_address() with a length check and bounded copy into sockaddr_un.sun_path. Changes Zero-initialize sockaddr_un with {} Reject paths >= sizeof(sun_path)…
pytorch/pytorch: viable/strict/1788237049
[distributed] Warn pipeline users about eager NCCL communicators (#19…
openclaw/openclaw: release-publish/ce9ef22169f5-33473514087
fix(release): restore evidence reuse and reconcile advisory ranges (#…
openai/codex: rust-v0.153.0-alpha.2
Release 0.153.0-alpha.2
openclaw/openclaw: release-publish/54183c0c9688-33471448769: fix(release): preserve original attempts in reused validation (#134820)
fix(release): preserve original attempts in reused validation chore(ui): refresh measured startup baseline
openai/codex: 0.153.0-alpha.1
Release 0.153.0-alpha.1
Estimating Population-Risk Curves Along Nonconvex Gradient Flows from the Training Sample
arXiv:2608.30261v1 Announce Type: cross Abstract: We estimate the conditional population-risk curve of a realized smooth nonconvex gradient flow from the training sample. Flow approximate leave-one-out (Flow-ALO)…
A-MADiff: Attention-Guided Multi-Agent DRL with Diffusion Policies for Memory-Aware Task Orchestration in Mobile AIGC Networks
arXiv:2608.29255v1 Announce Type: cross Abstract: Artificial Intelligence-Generated Content (AIGC) services employ Generative AI (GenAI) models to automatically generate diverse content. Mobile AIGC networks host GenAI…
Spectral Analysis for Sparse Matrix Computation: Insights and Potential
arXiv:2608.29362v1 Announce Type: cross Abstract: Sparse computations are fundamental to scientific computing, graph analytics, and machine learning, yet their performance is highly sensitive to the diverse sparsity and…
Multi-Marginal Flow Matching with Adversarially Learnt Interpolants
arXiv:2510.01159v3 Announce Type: replace Abstract: Learning the dynamics of a process given sampled observations at several time points is an important but difficult task in many scientific applications. When no…
Loading more...
500 items available