Issues

• Rewrote Mamba SSM kernels using Helion

  pytorch/helion#3309 · Filter by author shcho1118 opened on Aug 8, 2026
  1

• [benchmark] Upgrade MI350X jobs to PyTorch 2.14 when available

  pytorch/helion#3271 · Filter by author yushangdi opened on Aug 3, 2026
  2

• Nightly Benchmark Tracker Issue

  pytorch/helion#3269 · Filter by author yushangdi opened on Aug 3, 2026
  3

• Pretuned Benchmark Tracker Issue

  pytorch/helion#3216 · Filter by author yushangdi opened on Jul 31, 2026
  8

• autotuning does't work for nvfp4 cast with inline asm

  pytorch/helion#3177 · Filter by author vkuzo opened on Jul 28, 2026
  1

  

• autotuning does not work on large tensor + swizzling

  pytorch/helion#3176 · Filter by author vkuzo opened on Jul 28, 2026
  1

  

• can helion support in-tile reshape for quantization casting?

  pytorch/helion#3163 · Filter by author vkuzo opened on Jul 27, 2026
  1

  

• Optimize Helion compile overhead

  pytorch/helion#3118 · Filter by author xiaohongchen1991 opened on Jul 21, 2026

• Add failure tracking for benchmark

  pytorch/helion#2972 · Filter by author yushangdi opened on Jun 30, 2026
  

• [Pallas] adjust_block_size_constraints max-propagates per-tensor TPU alignment across all tile-indexed tensors sharing a block_id

  pytorch/helion#2842 · Filter by author norx1991 opened on Jun 22, 2026

• [RFC] Helion Remote Autotuning Service

  pytorch/helion#2745 · Filter by author mrodden opened on Jun 10, 2026
  4

• Same autotuned "static_ranges" config fails across different specialized shapes

  pytorch/helion#2705 · Filter by author xiaohongchen1991 opened on Jun 5, 2026
  3

• Compile-phase subprocess isolation to survive Triton C++ aborts

  precompile
  pytorch/helion#2531 · Filter by author choijon5 opened on May 20, 2026
  1

• Support subprocess benchmarking for non-picklable epilogue arguments

  autotuning
  pytorch/helion#2469 · Filter by author choijon5 opened on May 16, 2026

• Optimize fixed launch overheads

  pytorch/helion#2468 · Filter by author jansel opened on May 16, 2026
  2

  

• [Pallas] test_atomic_add_w_tile_attr broken with recent torch_tpu

  pytorch/helion#2304 · Filter by author cota opened on May 6, 2026
  1

  

• Better support for Larger Grouped Operations in Helion in hl.grid

  user support
  pytorch/helion#2148 · Filter by author melody-lui3 opened on Apr 30, 2026
  1

  

• Kernel perf regression using latest main branch with configs autotuned from v1.0.0

  performanceuser support
  pytorch/helion#2044 · Filter by author xiaohongchen1991 opened on Apr 17, 2026
  1

• [AMD 350x] Helion Layernorm kernel performance is worse compared to inductor for some shapes in backward pass

  rocm
  pytorch/helion#2013 · Filter by author MengjiaoZhou opened on Apr 13, 2026
  8

  

• not use TMA for symm mem tensor

  CUDA error
  pytorch/helion#1846 · Filter by author shunting314 opened on Mar 27, 2026

• Enable Partial Tile Store/Loading

  language supportuser support
  pytorch/helion#1679 · Filter by author mengluy0125 opened on Mar 14, 2026
  5

• Suboptimal config from autotuning

  autotuningenhancementperformanceuser support
  pytorch/helion#1640 · Filter by author gmagogsfm opened on Mar 10, 2026
  2

• [RFC] Helion Autotuning service for instantaneous auto-tuning times

  autotuning
  pytorch/helion#1518 · Filter by author msaroufim opened on Feb 19, 2026
  3

• Overlapping compilation time with benchmarking for the autotuner

  autotuning
  Feature

  pytorch/helion#1400 · Filter by author ethche opened on Feb 9, 2026
  1

  

• [RFC] Helion kernel + Inductor ops prologue / epilogue fusion

  PyTorch interop
  pytorch/helion#1346 · Filter by author yf225 opened on Jan 23, 2026
  12