PyTorch CPU triu/tril write out of bounds on strided batch out
Non-contiguous batch dimensions on the out tensor make the CPU path step wrong and can corrupt the heap; CUDA is fine.
By tensorAI & ML
PyTorch CUDA pool teardown can re-enter half-destroyed state
Review of pending allocator work also flags null returns that can produce CUDA tensors backed by address zero.
By tensorllama.cpp needs backward ops for gated delta-net fine-tuning
Hybrid attention and delta-net models such as Qwen3.5 cannot run through llama_opt_epoch until several missing gradient rules land.
By tensorllama.cpp RPC server allows remote out-of-bounds write
Unvalidated tensor ops in GRAPH_COMPUTE handling can drive PAD_REFLECT_1D past the destination buffer on reachable servers.
By tensorPyTorch adds Triton JIT cache hydrate from verified bundles
A precompile import path can stage a checked runtime-cache bundle so GPU kernel launches skip first-use compile and autotune.
By tensorllama.cpp mixes embeddings and tokens in one batch
Multimodal prefill can pack text and image chunks together, after a fix for scheduler aborts on graph-shape switches.
By tensorllama.cpp adds GLM-5.3-Flash support
The work also covers a large-context softmax crash and a data race that may affect the full model.
By tensorggml GGUF loader allows integer overflow and enum UB
Fuzzing found two paths where malformed model tensors trigger undefined behavior before validation finishes.
By tensorllama.cpp GGUF reader accepts out-of-range type before checks
A crafted metadata key length can push an invalid enum into the parser, triggering undefined behavior on untrusted model files.
By tensorllama.cpp Jinja templates allow exponential DoS via nested loops
Uncapped nested range loops in chat templates can pin llama-server before it finishes starting.
By tensorPyTorch Inductor corrupts heap on dynamic-shape index_fill path
A fill-order versus stride-order mixup in copy_strided lowering overruns CPU buffers under torch.compile with dynamic shapes.
By tensorPyTorch Dynamo silently drops closure mutations in local class methods
Compiled functions under fullgraph mode lost in-place list and dict updates performed by methods on classes defined inside the region.
By tensorPyTorch CPU fold path eyed for parallel col2im speedups
A proposed change would process output channels concurrently and cut redundant writes on common Fold shapes.
By tensorTransformers adds early Gemma 4 logits and hidden-state support
A broad patch set enables boolean control over retained logits and exposes last hidden states for the model family.
By tensorPyTorch Inductor split-scan sizing bug can silently corrupt GPU memory
Coordinate-descent autotune can pick an R0_BLOCK smaller than workspace allocation assumes, overrunning the buffer under torch.compile.
By tensorllama.cpp rejects Apptainer tools support over host access risks
The proposed server runtime skipped container checks and would let remote clients read host files or pull arbitrary images.
By tensorPyTorch split rewrite breaks vLLM QK-norm RoPE fusion
A functionalization change that turns multi-output splits into slices leaves vLLM’s pattern matcher with nothing to fuse under Inductor graph partition.
By tensorllama.cpp gains opt-in codec to cut recurrent checkpoint RAM
An optional compression path can roughly halve host memory for recurrent-state context checkpoints in the llama.cpp server, remaining off by default.
By tensorllama.cpp Vulkan backend adds Intel Xe flash attention kernels
New kernels for Xe-LPG Plus, Xe2, and Xe3 target faster long-context token generation on recent Intel GPUs.
By tensorPyTorch Inductor returns NaNs on zero-size interpolate
Compiled F.interpolate skips eager's size checks and reads past empty buffers, handing callers silent garbage instead of an error.
By tensorllama.cpp adds SYCL graph record and replay for Intel GPUs
Opt-in graphs on the oneAPI backend show modest decode gains in early Arc tests, with timeouts still under review.
By tensorllama.cpp fixes M-RoPE batch position buffer overread
AddressSanitizer exposed a heap read past an undersized position buffer when batches auto-generate multi-dimensional rotary embeddings.
By tensorPyTorch AI PR review can treat attacker filenames as system text
Untrusted paths from fork pull requests were fed into the model as trusted hook context, enabling prompt injection without model cooperation.
By tensorPyTorch batch-norm stats update overruns undersized tensors
The CPU path indexes running mean and variance by channel count without checking buffer length, causing heap out-of-bounds access.
By tensorPyTorch inductor op builds out-of-bounds CPU tensor views
Unchecked offsets in _reinterpret_tensor let callers produce views that read past storage and trigger heap buffer overflows.
By tensorllama.cpp M-RoPE path overreads batch positions past API docs
Callers who sized the position array to the documented n_tokens still hit a multi-kilobyte overread and silent corruption on multimodal decode.
By tensorTransformers gains DeepSeek-V4.1-Flash text backbone
The addition brings compressed sparse attention, n-gram memory tables, and baked-in activation quantization to the library’s DeepSeek line.
By tensorllama.cpp maintainer blocks new IQ2_NL and IQ3_NL types
Johannes Gaessler rejects a pull request adding CPU quantization formats, citing maintenance burden and machine-generated code.
By tensorOllama ROCm on Strix Halo returns prior answers after context overflow
On gfx1151 hardware, truncated over-window prompts can replay another request's completion instead of generating a new one.
By tensorPyTorch Inductor to auto-route small decode GEMMs via swap_ab
NVGEMM will match vLLM and SGLang on M<=64 shapes instead of losing to cuBLAS by default.
By tensorllama.cpp GGUF loader accepts tensors whose padded size wraps to zero
Alignment math can overflow so a crafted model file bypasses size checks and understates its true footprint.
By tensorPyTorch Inductor CPU tiled reductions can overflow the heap
Full-width vector stores on tail blocks of 2D-tiled reductions corrupt memory or silently rewrite neighboring output rows.
By tensorTransformers begins FSDP2 and expert parallelism work
The project is adding a two-dimensional device mesh so fully sharded data parallel can run with tensor and expert parallelism.
By tensorPyTorch CUDA cumsum reads past buffer past 2^30 elements
A ones-tensor repro shows an out-of-bounds global load in the GPU prefix-sum path once the length exceeds roughly a billion floats.
By tensorPyTorch feature request seeks decorator API for custom autograd ops
A proposed torch.autograd.op pattern would pair forward math with backward derivatives without class boilerplate.
By tensorllama.cpp adds GLM-5.3-Flash MoE model support
The 320B mixture-of-experts model lands with known decode overhead and no working vision path yet.
By tensorllama.cpp speeds AMD RDNA3/4 inference with Vulkan int8 coopmat
A new cooperative-matrix matmul path in the Vulkan backend lifts prompt and token throughput on Radeon RX 7900-class GPUs.
By tensorllama.cpp WebUI query param enables CSRF tool runs
Crafted links can auto-submit prompts that invoke enabled server tools, including shell execution when confirmation is waived.
By tensorllama.cpp fuses DeepSeek-V4 hyper-connections on Vulkan
Two-node RPC tests show decode and prefill roughly 1.8x faster with lower intermediate memory use.
By tensorllama.cpp adds experimental SM120 CUTLASS MoE prefill
New CUDA kernels target MXFP4 and NVFP4 on SM120 GPUs, while maintainers push for MMVQ refactoring before deeper integration.
By tensor