Issues

• TensorOperationInWrapper warning for tensor is not None in host code

  pytorch/helion#4350 · Filter by author qcc4cp opened on Nov 8, 2026

• ReductionLowering asserts on a reduction over a tunable-sized dimension for backends that do not pad factory tensors

  pytorch/helion#4316 · Filter by author qcc4cp opened on Nov 6, 2026

• [random] randint fails to compile with integer kernel arguments as bounds

  pytorch/helion#4268 · Filter by author 0z5a opened on Nov 4, 2026

• [randint] Intervals wider than 2^31 leave values unreachable

  pytorch/helion#4265 · Filter by author 0z5a opened on Nov 4, 2026

• [RFC] Benchmarking and launcher portability for out-of-tree Triton drivers

  pytorch/helion#4038 · Filter by author yzchang-plus opened on Oct 16, 2026

• [RFC] Backend capability hooks for out-of-tree Triton drivers

  pytorch/helion#4037 · Filter by author yzchang-plus opened on Oct 16, 2026

• [RFC] Device-type backend routing for out-of-tree accelerator backends

  pytorch/helion#4036 · Filter by author yzchang-plus opened on Oct 16, 2026

• sm121 heuristic: extend to the full decode grid? (follow-up to #3415)

  pytorch/helion#3772 · Filter by author yashb98 opened on Oct 5, 2026
  7

• Rewrote Mamba SSM kernels using Helion

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

• Nightly Benchmark Tracker Issue

  pytorch/helion#3269 · Filter by author yushangdi opened on Aug 3, 2026
  5

• 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