Issues

• ConvInteger produces incorrect results with nonzero zero points and padding

  robertknight/rten#1536 · Filter by author wyh2001 opened on Oct 27, 2026
  1

• Support specifying thread pool used for model optimizing and weight prepacking

  robertknight/rten#1533 · Filter by author robertknight opened on Oct 22, 2026
  1

• Document how to balance compilation speed and performance in debug builds

  robertknight/rten#1521 · Filter by author robertknight opened on Oct 10, 2026

• Support 2D symbolic value tensors in shape inference

  shape-inference
  robertknight/rten#1452 · Filter by author robertknight opened on Aug 26, 2026

• Consider replacing RustFFT to reduce build times

  build-perf
  robertknight/rten#1448 · Filter by author robertknight opened on Aug 21, 2026

• Optimizations for Ocrs text recognition and detection models

  robertknight/rten#1444 · Filter by author robertknight opened on Aug 18, 2026

• Add Arm SVE support to rten-simd

  robertknight/rten#1396 · Filter by author robertknight opened on Jul 29, 2026

• Prevent overflow in element count and offset computations

  safety
  robertknight/rten#1373 · Filter by author robertknight opened on Jul 22, 2026
  1

• Loading ONNX external data from Hugging Face hub cache doesn't work

  robertknight/rten#1334 · Filter by author robertknight opened on Jul 5, 2026

• Support strictly positive assumption for input dimension sizes

  shape-inference
  robertknight/rten#1307 · Filter by author robertknight opened on Jun 26, 2026

• Replace inline assembly with stabilized NEON dotprod intrinsics

  refactoring / cleanup
  robertknight/rten#1248 · Filter by author robertknight opened on Jun 6, 2026

• Improve shape inference for PP-OCRv5 text detection model

  shape-inference
  robertknight/rten#1225 · Filter by author robertknight opened on May 24, 2026
  1

• ThreadPool scheduling issues and stack overflows

  robertknight/rten#1188 · Filter by author jmeggitt opened on Jan 10, 2026
  3

• Optimize MatMulNBits with int8 compute when LHS input is a matrix

  performance
  robertknight/rten#1164 · Filter by author robertknight opened on Dec 19, 2025
  1

• Support shape inference for sequence ops

  shape-inference
  robertknight/rten#1163 · Filter by author robertknight opened on Dec 19, 2025

• Improve shape inference for models with subgraphs

  performanceshape-inference
  robertknight/rten#1131 · Filter by author robertknight opened on Dec 10, 2025

• Fuse grouped-query attention

  robertknight/rten#1091 · Filter by author robertknight opened on Nov 18, 2025
  2

• Expose mask to integer conversions in rten-simd

  robertknight/rten#1079 · Filter by author RReverser opened on Nov 12, 2025
  11

• Model::total_params undercounts parameters when model uses 4-bit weights

  robertknight/rten#1036 · Filter by author robertknight opened on Oct 27, 2025
  1

• Warn about potential accuracy/performance-impacting fallbacks when loading ONNX models

  robertknight/rten#973 · Filter by author robertknight opened on Oct 12, 2025
  3

• Allow "scalarish" inputs (scalars or single-element vectors) for some operators

  models
  robertknight/rten#942 · Filter by author robertknight opened on Sep 17, 2025
  1

• Remove internal operator implementation functions from public API

  robertknight/rten#911 · Filter by author robertknight opened on Sep 1, 2025
  1

• Optimize int8 kernels for symmetric quantization of weights

  performance
  robertknight/rten#835 · Filter by author robertknight opened on Jul 31, 2025
  1

• Improve GEMM multi-threading performance

  performance
  robertknight/rten#707 · Filter by author robertknight opened on May 23, 2025
  1

• Support ONNX functions

  models
  robertknight/rten#655 · Filter by author robertknight opened on Apr 9, 2025
  2