Update dependency helion to v1.2.0

[konflux-internal-p02]

ℹ️ Note

This PR body was truncated due to platform limits.

This PR contains the following updates:

Package	Change	Age	Confidence
helion	==1.0.0 → ==1.2.0

---

Release Notes

pytorch/helion (helion)

v1.2.0

Compare Source

What's Changed

• [Pallas] Add Pallas interpret CI job (revival of #​1938) by @​norx1991 in #​2491
• Disable rope-bwd from nightly perf CI for now by @​choijon5 in #​2494
• torch_tpu: update pin to f863c2b by @​cota in #​2528
• Improve CuTe RMS norm reduction path by @​oulgen in #​2540
• [pallas] support dynamic shapes in indirect gather lowering by @​cota in #​2493
• Enable >32-thread persistent reduction on CuTe for RMS norm perf by @​oulgen in #​2541
• Cache CuTe launcher imports and dtype mapping in hot path by @​oulgen in #​2542
• Add CuTe reduction-kernel autotuner seed by @​oulgen in #​2543
• Fuse two-pass loads of x in CuTe reduction kernels by @​oulgen in #​2544
• Add wide-chunk reduction seed and tidy fuse-two-pass-loads helper by @​oulgen in #​2545
• [Pallas] correct RNG seed buffer argument ordering in launcher by @​AmesingFlank in #​2539
• [cutedsl] move CuTe state out of DeviceFunction by @​jansel in #​2497
• [cutedsl] split CuTe tcgen05 config logic by @​jansel in #​2498
• [cutedsl] replace CuTe golden codegen tests by @​jansel in #​2499
• [cutedsl] handle CuTe edge epilogue autotune by @​jansel in #​2500
• [cutedsl] support flat edge TMA stores with scalar fallback by @​jansel in #​2501
• [cutedsl] enable clustered edge TMA and seed configs by @​jansel in #​2502
• [cutedsl] add hybrid TMA store pipeline for edge tiles by @​jansel in #​2503
• [cutedsl] tune edge K-tail epilogue placements by @​jansel in #​2504
• [cutedsl] split scheduler edge epilogues by @​jansel in #​2505
• [cutedsl] add clc scheduler diagnostics and lifecycle fixes by @​jansel in #​2506
• [cutedsl] add aux TMA load mode by @​jansel in #​2507
• [cutedsl] seed aux TMA and clc persistence candidates by @​jansel in #​2508
• [cutedsl] retile and pack aux epilogue by @​jansel in #​2509
• [cutedsl] vectorize and stage aux epilogue stores by @​jansel in #​2510
• [cutedsl] add staged scheduler mailbox diagnostic by @​jansel in #​2511
• [cutedsl] skip static full tma fallback by @​jansel in #​2512
• [cutedsl] add tcgen05 lifecycle context and kloop ownership by @​jansel in #​2513
• [cutedsl] introduce pure matmul object model for tcgen05 by @​jansel in #​2514
• [cutedsl] add explicit tcgen05 epilogue tile and two-cta kloop by @​jansel in #​2515
• [cutedsl] cache cute launch arguments by @​jansel in #​2516
• [cutedsl] add guarded flat tcgen05 role coordinates by @​jansel in #​2517
• [cutedsl] add tcgen05 tvm ffi launch with seed configs by @​jansel in #​2518
• [cutedsl] introduce target1 clc scheduler object by @​jansel in #​2519
• [cutedsl] add target1 direct entry path by @​jansel in #​2520
• [cutedsl] Update benchmarking code by @​jansel in #​2521
• [Pallas] int32 gather lowering by @​thcmbs in #​2550
• [Autotuner] Cap inner-tile block size by outer tile by @​thcmbs in #​2478
• Use fixed seed for nightly benchmarks by @​choijon5 in #​2561
• torch_tpu: update pin to 5281ef1 by @​cota in #​2562
• Emit vectorized 128-bit loads for CuTe inner reductions by @​oulgen in #​2546
• [Pallas] Fix mask broadcast on size-1 tensor dimensions by @​AmesingFlank in #​2556
• Enable vec loads for bf16/fp16 CuTe reductions via Uint16 bitcast by @​oulgen in #​2547
• [cutedsl] extend tcgen05 direct-entry path to additional targets by @​jansel in #​2551
• [cutedsl] add tcgen05_aux_stages and tcgen05_consumer_regs autotune knobs by @​jansel in #​2552
• [cutedsl] stage dynamic_persistent scheduler object infrastructure by @​jansel in #​2553
• [cutedsl] Expand hillclimber benchmarking to 32 targets by @​jansel in #​2554
• Fix XPU CI SYCL runtime setup by @​choijon5 in #​2566
• (Onboarding task) examples: add XSA (exclusive self-attention) kernel by @​calebmkim in #​2564
• Vec-aware two-pass load fusion for CuTe reductions by @​oulgen in #​2548
• [autoreview] Add --codex-no-sandbox flag for hosts without bubblewrap by @​choijon5 in #​2568
• Support tensor_descriptor + epilogue subtile by @​ethche in #​2557
• [CI] Bump Pyrefly, ruff, cute by @​oulgen in #​2569
• [cute] skip test_xsa on cute by @​oulgen in #​2581
• [cute] Softmax perf: vec hoist, fuser alias, hoist warp reduce, heuristics by @​oulgen in #​2570
• [cute] V=8 fp16/bf16 via 4+4 split; multi-row investigation tests by @​oulgen in #​2571
• [cute] Merge sibling constexpr V-loops + elide warp-reduce dtype wraps by @​oulgen in #​2572
• [cute] Warp-per-row layout for softmax-shape reductions by @​oulgen in #​2573
• Reject tensor_descriptor indexing when block size exceeds tensor dim (#​2555) by @​mengluy0125 in #​2555
• [cute] LICM for reciprocals: hoist 1/divisor out of inner loops by @​oulgen in #​2574
• [Pallas] Fix synchronize_device skipping sync for tuple returns on TPU by @​thcmbs in #​2580
• [Pallas] remove jax_export_ignore_forward_compatibility setting by @​cota in #​2452
• [cute] Cleaner LICM: alias DCE + FMA-friendly scale hoist by @​oulgen in #​2575
• [Pallas] Add support for non zero dim in gather by @​thcmbs in #​2558
• [cute] Software-pipeline inner vec loads to hide HBM latency by @​oulgen in #​2576
• [cute] Optional carried-only gate for the load pipeline pass by @​oulgen in #​2577
• [cute] Rewrite online softmax_two_pass → equivalent 3-pass form by @​oulgen in #​2578
• [test] Split cute compiler-pass tests by pass, drop kernel name from filenames by @​oulgen in #​2579
• [CI] bump cutlass by @​oulgen in #​2610
• A jagged_hstu_attention example that works on Pallas TPU by @​AmesingFlank in #​2218
• torch_tpu: update pin to adbe52d by @​cota in #​2602
• test_examples: skip test_jagged_dense_add on Pallas by @​cota in #​2607
• [metal] Auto-cap threadgroup size via CuteND/CuteFlattenedTileStrategy by @​aditvenk in #​1855
• [cosmetic/UI] Quieter console output and nicer progress formatting by @​choijon5 in #​2598
• [Pallas] Propagate inner tile alignment min_size to bounding outer tiles by @​thcmbs in #​2559
• [Pallas] Suppress Pallas RNG x64 warning spam in random tests by @​thcmbs in #​2617
• [metal] Add DeviceIR MPPGraph lowering for aten matmul by @​aditvenk in #​2472
• [Pallas] Avoid bool reshape lowering by @​thcmbs in #​2618
• ci: bump TPU jax/jaxlib pin to 0.10.1 by @​thcmbs in #​2621
• [Pallas] Fix indirect gather by @​thcmbs in #​2619
• [Pallas] Add generic fallback scatter by @​thcmbs in #​2620
• Migrate H100 CI to OSDC runners (pytorch/pytorch cu128 base image) by @​huydhn in #​2625
• torch_tpu: update pin to dd5b8dd by @​cota in #​2633
• [cute] Fix benchmark errors on CI for cute by @​oulgen in #​2622
• [cute] Require cuda13+ by @​oulgen in #​2623
• [compile] add ast backend customization hook by @​hinriksnaer in #​2624
• torch_tpu: update pin to e2250c4 by @​cota in #​2638
• [pallas] fix emit_pipeline's output mapping and indexing by @​cota in #​2599
• [Pallas] Adjust layernorm example tolerances and split bwd test by @​thcmbs in #​2560
• [pallas] fix fori_loop's output mapping by @​cota in #​2639
• [Pallas] Run each TPU benchmark shape in its own subprocess by @​norx1991 in #​2495
• examples/fp8_attention: enable for Pallas backend by @​cota in #​2640
• [CI] Start testing 3.14t by @​oulgen in #​2642
• [cute] Un-xfail/un-skip 29 CuTe example tests across reductions, attention, matmul by @​oulgen in #​2647
• [pallas] fast-path host-side launcher with pre-computed cache entries by @​choijon5 in #​2593
• [pallas] bypass JaxCallable for static-shape kernels via direct call_custom_kernel by @​choijon5 in #​2594
• [pallas] cache static-shape output meta-tensor allocation in pallas launcher by @​choijon5 in #​2595
• [pallas] direct-call hot-path squeezes: sig-lock + pre-baked invoke closure by @​choijon5 in #​2597
• [Pallas] Fix inplace output preload for aliased outputs (shared by multiple tiles) by @​thcmbs in #​2637
• [Pallas] bf16xint16 example tolerance by @​thcmbs in #​2663
• [Autotuning] Use matmul heuristics as default seeds by @​choijon5 in #​2428
• Fused all_gather_fp8_gemm kernel implementation by @​LironKesem in #​2436
• Don't roll reductions over hl.arange dims into loops by @​oulgen in #​2660
• Cap CuTe synthetic-lane reductions to a single warp by @​oulgen in #​2662
• [cute] Support scalar associative_scan over non-last dimension by @​oulgen in #​2665
• [cute] Fix fp16 reduction identity dtype and reshape-into-reduction by @​oulgen in #​2666
• [cute] Fix transpose shape-chains and add affine hl.arange reshape store by @​oulgen in #​2667
• [cute] Reduction-aware layout propagation for mixed scalar/block stores by @​oulgen in #​2668
• [cute] Synthetic thread axes for free hl.arange, dependent stores, and matmul fallbacks by @​oulgen in #​2669
• [compile] Migrate Types to a dedicated module by @​hinriksnaer in #​2603
• [cute] Un-mark stale CuTe xfails/skips and guard Triton-only assertions by @​oulgen in #​2670
• [cute] Lower baddbmm with static M=N collapsed onto the reduced N dim by @​oulgen in #​2673
• [cute] Lower chunked-recurrence GEMM (gdn_fwd_h) by @​oulgen in #​2674
• [Autotuner] Clear JIT fast-path caches during rebenchmarking by @​ethche in #​2659
• Add backward support to attention examples by @​ethche in #​2657
• [autotuner] Add Triton row-reduction and rope split/join heuristics by @​choijon5 in #​2648
• [autotuner] Surface workload hints and compiler analysis to the LLM prompt by @​choijon5 in #​2649
• [flaky test] Skip bf16xint16 GEMM test under Pallas CPU interpret by @​choijon5 in #​2683
• [tests] Launcher-contract tests in test_misc by @​yushangdi in #​2634
• [Pallas] Fix wrapper tensor output classification by @​thcmbs in #​2664
• [cute] Add on-disk compile caching for the CuTe backend by @​oulgen in #​2685
• [Pallas] Add fp8_attention to TPU benchmark sweep by @​norx1991 in #​2677
• [cutedsl] remove dead code and consolidate backend helpers by @​jansel in #​2650
• [benchmark] Cap per-shape autotune wall-clock in nightly CI by @​choijon5 in #​2686
• [cutedsl] expand epilogue store support and autotuner seeds by @​jansel in #​2651
• [cutedsl] add best-of-K autotuning and stabilize benchmarking by @​jansel in #​2652
• [cutedsl] migrate to CuTe 4.5.1 APIs by @​jansel in #​2653
• [cutedsl] productionize aux-TMA selection via search projection by @​jansel in #​2654
• [cutedsl] overlap epilogue store via a dedicated store warp by @​jansel in #​2655
• [cutedsl] make tcgen05 ab_stages budget-aware searchable by @​jansel in #​2656
• [cutedsl] require CuTe 4.5.1 / tvm-ffi / CUDA 13 and drop build-compat shims by @​jansel in #​2678
• [autotuner] add per-fragment config priors to bias the initial population by @​jansel in #​2679
• [cutedsl] add cute config-value priors (distribution-aware random seeds) by @​jansel in #​2680
• [cutedsl] generalize hardcoded tcgen05 FFI seeds into a shape-derived heuristic by @​jansel in #​2681
• [cutedsl] generalize tcgen05 TargetN specialization into structural handling by @​jansel in #​2682
• [benchmarks] Fix flash_attention accuracy after attention returns (out, lse) by @​choijon5 in #​2691
• [autotune] final-pick verification: re-rank top configs to beat timing noise by @​choijon5 in #​2626
• [pallas] lower bf16/fp16 matmuls through pl.dot on TPU by @​choijon5 in #​2629
• [pallas] route whole-matrix matmuls through XLA's lax.dot_general by @​choijon5 in #​2631
• [autotune] rank TPU matmul candidates by true on-chip time (the big win) by @​choijon5 in #​2632
• [cute] Reject missing-keepdim reduction broadcast in type propagation by @​oulgen in #​2695
• [cute] Enable cute-gated tests: structural atomics + config reuse by @​oulgen in #​2688
• [cute] Support float32 atomic_max/atomic_min via integer-bitcast dispatch by @​oulgen in #​2689
• [cute] Lower layernorm-style reductions over a matmul-distributed output dim by @​oulgen in #​2690
• [pallas] revert bf16/fp16 matmul pl.dot lowering, keep lax.dot_general by @​choijon5 in #​2701
• [fix CI] Make TestLauncher sync device-agnostic to fix XPU CI by @​choijon5 in #​2698
• [Pallas] Factor shared compile/output helpers from launchers by @​AmesingFlank in #​2645
• [Pallas] Add Kernel.jax_fn for pure-JAX export of Pallas kernels by @​AmesingFlank in #​2646
• [Pallas] Fix BlockSpec for emit_pipeline nested in outer non-grid loop by @​norx1991 in #​2676
• [cute] Emit cross-thread reduction for a reduced inner tiled dim by @​oulgen in #​2706
• Fix fbcode CI torch.compile fusion with newer PyTorch by @​choijon5 in #​2567
• [triton][AMD] Enable load cache modifier autotuning by @​ethche in #​2712
• check_example: add optional cosine similarity by @​cota in #​2671
• dot_precision: introduce default, high and highest by @​cota in #​2697
• torch_tpu: update pin to e2b5601 by @​cota in #​2672
• [compiler] Add MemoryOpFact linking loads/stores to metadata by @​ethche in #​2710
• [CI] Bump nightly CUDA lanes from cu128 to cu130 by @​choijon5 in #​2713
• [pallas] enable emit_pipeline BlockSpecs for jagged tile ranges by @​ethche in #​2716
• Fix formatting by @​oulgen in #​2723
• Kernel metadata hooks by @​IshanAryendu in #​2699
• [Autotuner] Run accuracy check in benchmark worker subprocess by @​tarinduj in #​2700
• Fix ref_mode _handle_setitem mask expansion for non-leading int-tensor dims by @​hinriksnaer in #​2693
• [cute] Remove dead code and de-duplicate AST-pass helpers by @​oulgen in #​2724
• test_squeeze_and_excitation_net_bwd_db: enable on Pallas by @​cota in #​2732
• torch_tpu: update pin to 013936a by @​cota in #​2733
• Add multi tile loop support to autodiff by @​karthickai in #​2338
• Fix dynamic shapes tracing support for non-padding backends by @​jbschlosser in #​2692
• Add AWS Bedrock provider to LLM autotuner transport by @​yushangdi in #​2734
• [pallas] Tile geglu/swiglu over the natural N-D shape on TPU by @​choijon5 in #​2725
• [pallas] Skip the inner-loop pad when its begin is block-aligned by @​choijon5 in #​2726
• [pallas] Tile batch_softmax over the native 3D shape by @​choijon5 in #​2727
• [pallas] torch.compile(tpu) capture for Helion Pallas kernels by @​choijon5 in #​2730
• [cute] Fix autotune IndexError on reduction kernels with nested tiling by @​fulvius31 in #​2735
• Fix Pallas/TPU alignment constraints to enable example/aot_examples.py by @​yarongmu-google in #​2753
• [triton] Simplify floordiv/remainder for non-negative tile indicies by @​hinriksnaer in #​2757
• [cute] Enable FP8 (e4m3) scaled_mm on the tcgen05 tensor-core path by @​yushangdi in #​2736
• [triton] contiguity hint for swizzle gathers; improved triton nvpf4_gemv by @​ethche in #​2738
• [cute] Whitelist fp8 in persistent role-local pure-call set by @​yushangdi in #​2739
• Cache CUDA device capability lookups on the kernel dispatch hot path by @​yushangdi in #​2746
• Switch H100 CI image back to nvidia/cuda from pytorch/pytorch by @​huydhn in #​2767
• [Pallas] gate strict 128-alignment on tensor lane dim by @​yarongmu-google in #​2765
• [pallas] ci: temporarily freeze PyTorch wheel version by @​cota in #​2771
• [cute] Support column-major (K-major) B on the tcgen05 fp8 TMA path by @​yushangdi in #​2740
• [cute] Deep AB staging for fp8 to close the compute-bound gap by @​yushangdi in #​2741
• [cute] Fused-scale epilogue: scalar colvec read for per-row scale by @​yushangdi in #​2742
• Fix UnicodeEncodeError in codegen under ASCII locale by @​karthickai in #​2775
• [test] Restore mutated kernel configs/settings in test_examples to stop leaks by @​yushangdi in #​2777
• Fix tritonbench auto-install in benchmarks/run.py by @​hinriksnaer in #​2776
• Add a SymInt-free tensor specialization key for exact torch.Tensor args by @​yushangdi in #​2759
• [cute] Vectorize fp8 loads in the SIMT matmul fallback by @​yushangdi in #​2772
• [cute] Fix warp-reduce hoist double-count for matmul split-K reductions by @​yushangdi in #​2773
• [cute] Paired fp8 decode (cvt.f16x2.e4m3x2) to close the GEMV perf gap by @​yushangdi in #​2774
• [pallas] fix BlockSpec indexing for 0D tensors by @​cota in #​2783
• [pallas] fix boundary mask with squeezed leading dimensions by @​cota in #​2784
• [Pallas] Jagged row-tile building blocks by @​thcmbs in #​2717
• [pallas] test_examples: enable on Pallas by @​cota in #​2785
• [cute] 2-CTA MMA at bm=128 (CTA tile 64xbn) for fp8 tcgen05 by @​yushangdi in #​2778
• Add regression test for scalar-arg Triton specialization by @​yushangdi in #​2793
• [cute] Pre-wait rowvec aux register hoist for the bm=128 2-CTA family by @​yushangdi in #​2779
• Fix pyrefly type errors in distributed examples by @​yushangdi in #​2801
• [cute] Persist autotune winner from memory instead of recompiling by @​fulvius31 in #​2768
• [cute] Reuse reduction thread axis for free hl.arange in sibling branches by @​yushangdi in #​2786
• [Pallas] Add pallas_loop_type = compact_worklist by @​ethche in #​2782
• [examples] Add a simpler concat implementation by @​hinriksnaer in #​2766
• [test] Add partial slice indexing test for [:n] and [n:] patterns by @​hinriksnaer in #​2799
• [examples] Correctness fixes to reduction example kernels for the seed-heuristic curriculum by @​calebmkim in #​2760
• [autotuner] Reduction fact layer: ReductionFact + AccumulatorFact + enriched MemoryOpFact by @​calebmkim in #​2761
• [autotuner] Triton reduction seed heuristic (generalizable core) by @​calebmkim in #​2762
• [Pallas] Implement is_row_map_axis legality gate for jagged carry by @​thcmbs in #​2718
• [cute] Add CuteFp8GemmSkinnyMHeuristic autotuner seed for skinny-M FP8 GEMM by @​yushangdi in #​2808
• Fix pyrefly errors from torch SymInt.sympy() -> sympy.Basic by @​ethche in #​2813
• [cute] Clear L2 in the generic autotune benchmark loops, which is used by cute backend by @​yushangdi in #​2815
• testing: reject @​skipIfFn calls without parentheses by @​cota in #​2810
• chore: Bump actions/checkout from 6 to 7 by @​dependabot[bot] in #​2811
• [autotune] Fix pyrefly missing-attribute on L2 cache-clear calls by @​yushangdi in #​2820
• Pin pytorch version in benchmark-tpu to existing ci version instead of nightly by @​ethche in #​2818
• [cute] Sample CUDA stream per launch to fix empty-graph capture by @​yushangdi in #​2817
• [Pallas] Defer tile load masks past transposes onto the sublane axis by @​ethche in #​2812
• [chore] Upgrade pyrefly to 1.1.1 by @​oulgen in #​2825
• Lean kernel-artifact telemetry for cost-model dataset by @​IshanAryendu in #​2770
• Fix Docker CuTe install by @​fallintoplace in #​2831
• [Pallas] Enable row-slab streaming for compact jagged loads by @​ethche in #​2827
• Fix invalid Python in hl.tile matmul docstring example by @​yurekami in #​2832

New Contributors

• @​calebmkim made their first contribution in #​2564
• @​LironKesem made their first contribution in #​2436
• @​IshanAryendu made their first contribution in #​2699
• @​tarinduj made their first contribution in #​2700
• @​jbschlosser made their first contribution in #​2692
• @​fallintoplace made their first contribution in #​2831
• @​yurekami made their first contribution in #​2832

Full Changelog: pytorch/helion@v1.1.0...v1.2.0

v1.1.0

Compare Source

What's Changed

• [Pallas] Use TPU-aware synchronize_device across autotuner and testing by @​norx1991 in #​1944
• Refactor if-else branches, unblock test_if_new_variable_in_static_range for Pallas TPU by @​AmesingFlank in #​1935
• [Pallas] Register pallas_loop_type only when inner loops exist by @​norx1991 in #​1915
• [Pallas] Fix emit_pipeline program_id mapping with loop_order reordering by @​norx1991 in #​1916
• [Pallas] Add expected-failure test for non-power-of-2 RDIM size by @​norx1991 in #​1945
• Update docs: add missing API entries by @​choijon5 in #​1948
• [Benchmark] Add compile time measurement to CI benchmarks by @​choijon5 in #​1952
• use is_symm_mem_tensor if the API is available by @​shunting314 in #​1933
• [Pallas] When not doing tiling for entire kernel, use explicit BlockSpecs instead of None by @​AmesingFlank in #​1960
• Fix FROM_BEST_AVAILABLE matching with hl.specialize() after #​1883 by @​fulvius31 in #​1940
• [Pallas] Fix indexing scalars using SMEM memory space by @​AmesingFlank in #​1955
• [Pallas] Add xfail tests for scalar .begin index not collapsing dims by @​norx1991 in #​1971
• Skip test_hl_rand_mixed_argument_order on MTIA due to unaligned address crash by @​karthickai in #​1977
• Fix _supports_maxnreg() to guard against non-CUDA backends by @​karthickai in #​1981
• Skip TestRandomPhiloxParity class on MTIA (#​1979) by @​karthickai in #​1982
• [Pallas] Fix acccesing tensors using index from hl.grid() by @​AmesingFlank in #​1956
• [Pallas] Support using traced size-1 tensor as condition predicate, unblocking test_if_arg_indexed_scalar by @​AmesingFlank in #​1957
• [Pallas] Fix atomic_add dtype cast and VMEM preload for fori/pipeline launchers by @​thcmbs in #​1966
• [Pallas] Use exact RDIM size instead of next-power-of-2 by @​norx1991 in #​1954
• [Pallas] Add support for accessing tensors with the pattern of tile.index + offset /.id/.begin/.end by @​AmesingFlank in #​1968
• [metal] Reuse Inductor's MetalOverrides for MSL expression emission by @​aditvenk in #​1853
• [metal] Add Metal codegen handlers for load, store, and mask_to by @​aditvenk in #​1854
• [Pallas] Enable a subset of test_grid tests for Pallas by @​AmesingFlank in #​1985
• [Pallas] Add a test for accessing tensors with hl.grid() index + offset by @​AmesingFlank in #​1988
• Relax rms_norm example tolerance for Pallas bf16 by @​thcmbs in #​1983
• [autotuner] Introduce BenchmarkProvider abstraction for kernel benchmarking by @​hinriksnaer in #​1928
• [Pallas] Use HBM BlockSpecs for output-only tensors to save VMEM by @​norx1991 in #​1984
• Emit autotune failure summary warnings by @​allgather in #​1994
• [Pallas] Remove unused is_device_loop variable in _pallas_index_str by @​AmesingFlank in #​1990
• [Pallas] Fix BlockSpecs for 2D tl.grid([m, n]), unblocking test_scalar_access_hl_grid_2d by @​AmesingFlank in #​1986
• Reduce some autotuner overhead without changing kernel behavior by @​svdrecbd in #​1885
• [compile] Add pre_codegen hook to Backend ABC by @​hinriksnaer in #​1976
• Adding AMD Mi350x machines to CI with new labels. by @​umechand-amd in #​1835
• [Pallas] Fix accessing tensor via hl.grid() index within a device loop, unblocking test_scalar_access_hl_grid_2d_nested by @​AmesingFlank in #​1989
• Support mtia in LocalAutotuneCache by @​Hamlin-Li in #​1996
• removed stale comments by @​hinriksnaer in #​1997
• Fix invalid default config for kernels with large tensor numel by @​fulvius31 in #​1839
• chore: Bump actions/github-script from 8 to 9 by @​dependabot[bot] in #​2000
• Fix _benchmark dropping configs that fail compilation by @​fulvius31 in #​1942
• [metal] Add MSL AST walker for Python-to-C++ translation by @​aditvenk in #​1794
• Enable tensor_descriptor based atomic ops by @​ethche in #​1953
• [metal] Add @​metal_jit decorator for AST-to-MSL compilation by @​aditvenk in #​1991
• Epilogue subtiling: store indexing fix, example, and tuple output support in run_example by @​choijon5 in #​1907
• [Compiler] Added backend registry by @​hinriksnaer in #​1967
• [metal] Wire @​metal_jit into MetalBackend and simplify launcher by @​aditvenk in #​1992
• [Pallas] Skip trivial reduction mask when RDIM size equals actual dim by @​norx1991 in #​1993
• [Pallas] Fix TPU min_dot_size for matmul autotuning by @​norx1991 in #​1999
• chore: Bump actions/upload-pages-artifact from 4 to 5 by @​dependabot[bot] in #​2015
• Make _get_tile_with_offset_info accept a torch.fx.Node as arg, instead of the entire CodegenState by @​AmesingFlank in #​2005
• [metal] Skip tests better to pass in internal CI by @​aditvenk in #​2016
• Add to_code() as backend-agnostic alias for to_triton_code() by @​norx1991 in #​2012
• [Pallas] Fix fori_loop multi-dim index decomposition with nested loops by @​norx1991 in #​1917
• custom config filter by @​shunting314 in #​1847
• [Pallas] Reject 64-bit input tensors and fix tiling ZeroDivisionError by @​norx1991 in #​1950
• Bump CI libtpu to 0.0.40 by @​AmesingFlank in #​2017
• [Pallas] Make test_tensor_access_tile_index_offset more meaningful and reveal codegen issue in Pall

✂ Note

PR body was truncated to here.

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

To execute skipped test pipelines write comment /ok-to-test.

---

Documentation

Find out how to configure dependency updates in MintMaker documentation or see all available configuration options in Renovate documentation.