Automated nightly build for version 0.6.18 (dev20260814)

## What's Changed * fix: use explicit .ptr on scalar struct fields in monolithic MLA decode to get cleaner logs by @jdebache in https://github.com/flashinfer-ai/flashinfer/pull/3458 * fix: One sided MOE A2A warp token policy hangs in some cases, disable… by @djns99 in https://github.com/flashinfer-ai/flashinfer/pull/3371 * test: correct XQA NVFP4 skip reason to SM120/SM121 by @deng451e in https://github.com/flashinfer-ai/flashinfer/pull/3510 * fix XQA NVFP4 head dim by @Njuapp in https://github.com/flashinfer-ai/flashinfer/pull/3534 * tests: split test_trtllm_gen_attention.py into prefill / decode / decode-xqa shards by @bkryu in https://github.com/flashinfer-ai/flashinfer/pull/3162 * docs(misc): expose DiT/RoPE norms; new RSTs for GDN decode/prefill/Mamba by @kangbintNV in https://github.com/flashinfer-ai/flashinfer/pull/3446 * perf: cache cudaGetDeviceProperties in CudaDevice (fmha_v2) by @aws-jiadingg in https://github.com/flashinfer-ai/flashinfer/pull/3522 * FMHAv2 on SM120 for head_dim 256/512 + sliding-window masks by @dbari in https://github.com/flashinfer-ai/flashinfer/pull/3518 * feat: add FlashInfer Trace Apply by @yongwww in https://github.com/flashinfer-ai/flashinfer/pull/3240 * Fix smem race in `FilteredTopK` overflow refinement by @awgu in https://github.com/flashinfer-ai/flashinfer/pull/3529 * Extend Require Workspace Size and Disable Overlapping Criritcal Workspace Sections w/ 'auto' backend + autotuning for MLA decode by @Vinnie6167 in https://github.com/flashinfer-ai/flashinfer/pull/3465 * perf(sampling): Optimize top_k_top_p_sampling_from_logits/from_probs for large-vocab small-k sampling by @bkryu in https://github.com/flashinfer-ai/flashinfer/pull/3461 * Use an environment variable to control the number of reserved SMs for overlapping in TRT-LLM fused MoE by @jinyangyuan-nvidia in https://github.com/flashinfer-ai/flashinfer/pull/3483 * Remove excessive CuteDslMoEWrapper memory allocation by @nvjullin in https://github.com/flashinfer-ai/flashinfer/pull/3404 * docs: close v0.6.13 doc-check gaps + fix(moe) misleading topk_indices ICHECK message by @kangbintNV in https://github.com/flashinfer-ai/flashinfer/pull/3546 * Unified MoE API: MoELayer with cross-backend NVFP4 autotune by @aleozlx in https://github.com/flashinfer-ai/flashinfer/pull/3093 * Support smaller DSv4 sparse MLA head counts by @PerkzZheng in https://github.com/flashinfer-ai/flashinfer/pull/3545 * feat(bench): add GDN routines to flashinfer_benchmark.py by @bkryu in https://github.com/flashinfer-ai/flashinfer/pull/3572 * Add BF16 MoE SwiGLU OA params by @ruoqianguo in https://github.com/flashinfer-ai/flashinfer/pull/3532 * perf(gemm): update mm_fp4 b12x SM120 NVFP4 dense GEMM kernel by @yichengj0 in https://github.com/flashinfer-ai/flashinfer/pull/3560 * feat: add sm90 delta rule dsl prefill by @guangyunh-nv in https://github.com/flashinfer-ai/flashinfer/pull/3477 * fix: headDim=512 GQA decode (#3343) by @djmmoss in https://github.com/flashinfer-ai/flashinfer/pull/3393 * Add Qidi Sang to CODEOWNERS by @saltyminty in https://github.com/flashinfer-ai/flashinfer/pull/3593 * [feat] Add gated tanh-GELU (GeluTanh) activation to CUTLASS fused MoE (GEMMA 4) by @jhaotingc in https://github.com/flashinfer-ai/flashinfer/pull/3501 * [chore] Add jiahanc to GDN commiter by @jiahanc in https://github.com/flashinfer-ai/flashinfer/pull/3600 * fix(ci): skip TRT-LLM Gen BF16 MoE SwiGLU test outside SM10.x by @ruoqianguo in https://github.com/flashinfer-ai/flashinfer/pull/3580 * fix(mla): warn when backend='auto' falls back to non-Blackwell kernel on SM>=100 by @elwhyjay in https://github.com/flashinfer-ai/flashinfer/pull/3405 * [fix] Fix SM100 GDN prefill hang by @jiahanc in https://github.com/flashinfer-ai/flashinfer/pull/3581 * feat: improve gated delta rule benchmark script by @guangyunh-nv in https://github.com/flashinfer-ai/flashinfer/pull/3616 * feat: delta rule work with zero length sequence by @guangyunh-nv in https://github.com/flashinfer-ai/flashinfer/pull/3536 * feat(cutile): introduce cuTile backend (mm_bf16 + bmm_bf16 + gemm_fp8_nt_groupwise) by @yifeis-nv in https://github.com/flashinfer-ai/flashinfer/pull/3426 * chore: remove leftover cpp srcs from #3477 by @guangyunh-nv in https://github.com/flashinfer-ai/flashinfer/pull/3613 * feat: add mxfp8 quant to moe a2a combine by @IwakuraRein in https://github.com/flashinfer-ai/flashinfer/pull/3376 * fix: scope trtllm-gen block_size guard to its own backend by @elwhyjay in https://github.com/flashinfer-ai/flashinfer/pull/3428 * feat(attention): add SM120 sparse MLA kernels by @lucifer1004 in https://github.com/flashinfer-ai/flashinfer/pull/3395 * test: match test output scaling with cutedsl kernel (scale before bf16 cast) by @saltyminty in https://github.com/flashinfer-ai/flashinfer/pull/3596 * fix(sampling): fix shared-memory race and out-of-range token id in SamplingFromLogitsKernel by @kahyunnam in https://github.com/flashinfer-ai/flashinfer/pull/3624 * feat(attention): head_dim=512 support for attention prefill & decode for Gemma 4 on SM120/121 by @bkryu in https://github.com/flashinfer-ai/flashinfer/pull/3576 * [fix] explicitly validate grouped_gemm_nt_masked layout contract to prevent NaN contamination by @kahyunnam in https://github.com/flashinfer-ai/flashinfer/pull/3574 * fix(gemm): relax b12x FP4 K constraint from 128 to 32 (TMA alignment) by @yichengj0 in https://github.com/flashinfer-ai/flashinfer/pull/3646 * Add MXFP8 MoE SwiGLU OA parameters by @ruoqianguo in https://github.com/flashinfer-ai/flashinfer/pull/3504 * Add support for non-multiple VEC_COLS in fused_rmsnorm_silu for bf16/fp8 by @xueweilnvidia in https://github.com/flashinfer-ai/flashinfer/pull/3417 * fix(topk): eliminate multi-CTA radix top-k stream hangs on SM120/SM121 by @waynehacking8 in https://github.com/flashinfer-ai/flashinfer/pull/3615 * feat: add sm120 delta rule dsl prefill by @guangyunh-nv in https://github.com/flashinfer-ai/flashinfer/pull/3479 * Add SM120 NVFP4 attention JIT path by @tiffany940107 in https://github.com/flashinfer-ai/flashinfer/pull/3640 * Add MXFP8 MoE GEMM entry (cute SM120 backend) by @CarstyYou in https://github.com/flashinfer-ai/flashinfer/pull/3562 * feat: Add BF16_FP4 GEMM with cuDNN and CuTe-DSL backends for SM120/121 for W4A16 workloads by @bkryu in https://github.com/flashinfer-ai/flashinfer/pull/3597 * perf(gdn): make GDN kernels compilation batch-size agnostic (support dynamic batch shapes for vLLM integration) by @kahyunnam in https://github.com/flashinfer-ai/flashinfer/pull/3649 * feat: add sm90 cp delta rule dsl by @guangyunh-nv in https://github.com/flashinfer-ai/flashinfer/pull/3481 * fix(sampling): terminate top-p search at adjacent float bounds by @djmmoss in https://github.com/flashinfer-ai/flashinfer/pull/3623 * fix: fix tinygemm barrier bug by @yweng0828 in https://github.com/flashinfer-ai/flashinfer/pull/3630 * feat(gdn): BF16 state recovery/decode kernel with per-request K and f… by @ameynaik-hub in https://github.com/flashinfer-ai/flashinfer/pull/3502 * feat(benchmark): add MLA --mla_is_var_seq / --mla_cute_dsl_impl knobs by @lunarz-dev in https://github.com/flashinfer-ai/flashinfer/pull/3695 * Enable SwapAB in `mm_fp4` `cute-dsl` backend when M is not a multiple of 8. by @b8zhong in https://github.com/flashinfer-ai/flashinfer/pull/3667 * fix(moe): Fix unbounded weight-cache growth in b12x MoE by @bkryu in https://github.com/flashinfer-ai/flashinfer/pull/3709 * Add Relu2 + ungated MoE to CuteDSL MoE by @b8zhong in https://github.com/flashinfer-ai/flashinfer/pull/3642 * trtllm_batch_decode_with_kv_cache_mla trtllm-gen backend cum_seq_lens_q support by @saltyminty in https://github.com/flashinfer-ai/flashinfer/pull/3238 * fix: pass XQA NVFP4 scale-factor strides by @Njuapp in https://github.com/flashinfer-ai/flashinfer/pull/3608 * feat(attention): FP8 KV cache support for Hopper SM90 MLA by @qsang-nv in https://github.com/flashinfer-ai/flashinfer/pull/3694 * [GDN] sm100: support more state dtype by @Observer007 in https://github.com/flashinfer-ai/flashinfer/pull/3715 * feat: cuTile Grouped MXFP8 Quantization by @philipphack in https://github.com/flashinfer-ai/flashinfer/pull/3657 * Accept uint8 workspaces in CUTE DSL MLA decode by @leejnau in https://github.com/flashinfer-ai/flashinfer/pull/3599 * tests: split test_trtllm_gen_fused_moe.py into shards by @feih-nv in https://github.com/flashinfer-ai/flashinfer/pull/3635 * perf(moe): Enhance CuteDSL NVF4 MOE Perf by @liyuhannnnn in https://github.com/flashinfer-ai/flashinfer/pull/3564 * fix(mamba): reject SM120/SM121 in SSDCombined with a clear error by @waynehacking8 in https://github.com/flashinfer-ai/flashinfer/pull/3668 * docs(gemm): add missing .rst entries for mm_bf16_fp4 and prepare_bf16_fp4_weights by @kangbintNV in https://github.com/flashinfer-ai/flashinfer/pull/3710 * docs: add missing parameter entries to docstrings and env vars by @kangbintNV in https://github.com/flashinfer-ai/flashinfer/pull/3627 * rename back by @aleozlx in https://github.com/flashinfer-ai/flashinfer/pull/3730 * docs: document scale_major_mode param and FLASHINFER_AUTOTUNE_DIR env var by @kangbintNV in https://github.com/flashinfer-ai/flashinfer/pull/3696 * feat: add mxfp4/nvfp4 quant to moe a2a combine by @IwakuraRein in https://github.com/flashinfer-ai/flashinfer/pull/3643 * chore: fix enable_pdl for trtllm-gen routing and finalize kernel by @IwakuraRein in https://github.com/flashinfer-ai/flashinfer/pull/3588 * Bug fix (gdn): Layout contract fix from #3649 by @kahyunnam in https://github.com/flashinfer-ai/flashinfer/pull/3693 * Ameyn/fix fp32 mtp pool out indices by @ameynaik-hub in https://github.com/flashinfer-ai/flashinfer/pull/3490 * Prune moe tests by @aleozlx in https://github.com/flashinfer-ai/flashinfer/pull/3733 * Fix the output allocation consistency of trtllm-gen MoE APIs by @b8zhong in https://github.com/flashinfer-ai/flashinfer/pull/3678 * feat(comm): Support per-token LoRA Info in MoE a2a comm payloads by @JyChang012 in https://github.com/flashinfer-ai/flashinfer/pull/3375 * docs(gdn): document missing ssm_state_indices param in gated_delta_rule_mtp by @kangbintNV in https://github.com/flashinfer-ai/flashinfer/pull/3725 * feat(moe): enable DSFp8 + LoRA delta path by @zetacat in https://github.com/flashinfer-ai/flashinfer/pull/3708 * Improve GDN prefill perf by ~20-25% (mainloop efficiency) by @jhjpark in https://github.com/flashinfer-ai/flashinfer/pull/3742 * test: fix CUDA OOM in batch-prefill custom-mask test on 24GB CI GPUs by @waynehacking8 in https://github.com/flashinfer-ai/flashinfer/pull/3609 * Remove pagesize 16/32 assertion from xqa nvfp4 sm120 by @Njuapp in https://github.com/flashinfer-ai/flashinfer/pull/3724 ## New Contributors * @deng451e made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3510 * @Njuapp made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3534 * @aws-jiadingg made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3522 * @awgu made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3529 * @ruoqianguo made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3532 * @yichengj0 made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3560 * @jhaotingc made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3501 * @elwhyjay made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3405 * @yifeis-nv made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3426 * @lucifer1004 made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3395 * @waynehacking8 made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3615 * @tiffany940107 made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3640 * @CarstyYou made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3562 * @feih-nv made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3635 * @liyuhannnnn made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3564 * @JyChang012 made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3375 * @zetacat made their first contribution in https://github.com/flashinfer-ai/flashinfer/pull/3708 **Full Changelog**: https://github.com/flashinfer-ai/flashinfer/compare/v0.6.13rc2...v0.6.14

Automated nightly build for version 0.6.13 (dev20260613)

Automated nightly build for version 0.6.13 (dev20260610)

**Full Changelog**: https://github.com/flashinfer-ai/flashinfer/compare/v0.6.12rc2...v0.6.12rc3

Automated nightly build for version 0.6.12 (dev20260528)

Automated nightly build for version 0.6.12 (dev20260523)

Automated nightly build for version 0.6.11 (dev20260513)