ByteDance-Seed/Triton-distributed

Distributed Compiler and Optimized Parallel Kernels

Stars: 1.5k
Forks: 172
Open beginner issues: 0
Indexed issues: 48

Dominant language: Python
License: MIT
Last GitHub push: Aug 6, 2026
Latest indexed: Sep 14, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed
PR merge metrics: No merged PRs in 30d

Open All

48 open issues indexed

Tried to generate ptx automatically with xla？ Open

question

ByteDance-Seed/Triton-distributed#3 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Is there a related Docker image available? Open

enhancement

ByteDance-Seed/Triton-distributed#6 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
fp8 support Open

enhancement

ByteDance-Seed/Triton-distributed#9 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
which one is all reduce test file~~ Open

enhancement

ByteDance-Seed/Triton-distributed#12 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 14, 2026
vllm Op benchmark Open

question

ByteDance-Seed/Triton-distributed#13 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
[Bug] Run into CUDA error: misaligned address on H100 Open

bug

ByteDance-Seed/Triton-distributed#22 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 14, 2026
pip install bugs Open

bug

ByteDance-Seed/Triton-distributed#33 · 5 comments · 1 reaction · 0 assignees · Indexed Sep 14, 2026
test_gemm_rs performance issue Open

ByteDance-Seed/Triton-distributed#40 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
nvshmemx_getmem_nbi_block failed when test nvshmem in H800 Open

bug

ByteDance-Seed/Triton-distributed#43 · 6 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Experiment script requirement Open

enhancement

ByteDance-Seed/Triton-distributed#47 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Why is there no performance improvements on H20 GPUs？ Open

question

ByteDance-Seed/Triton-distributed#48 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Support MLA Decode Open

enhancement

ByteDance-Seed/Triton-distributed#60 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
rocSHMEM (AMD) support status Open

AMD

ByteDance-Seed/Triton-distributed#69 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 14, 2026
When to release MoE+A2A overlapping kernels？ Open

ByteDance-Seed/Triton-distributed#87 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 14, 2026
How to use all_to_all_kernel in Megatron-LM Open

ByteDance-Seed/Triton-distributed#88 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Performance issue about A2A on 8xH800 Open

ByteDance-Seed/Triton-distributed#98 · 8 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
About issues running megatritonkernel Open

ByteDance-Seed/Triton-distributed#99 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 14, 2026
Any comparison between mega-triton-kernel & sglang/vllm? Open

ByteDance-Seed/Triton-distributed#100 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Add a new backend? Open

enhancement

ByteDance-Seed/Triton-distributed#101 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
[Question] For gemm_rs kernel, does FUSE_SCATTER option achieve better performance or the non fused one? Open

ByteDance-Seed/Triton-distributed#105 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Discussion: Enabling NVIDIA-Equivalent Distributed Optimizations on AMD GPUs Open

AMD

ByteDance-Seed/Triton-distributed#108 · 5 comments · 0 reactions · 1 assignee · Indexed Sep 14, 2026
复现mega_triton_kernel报错 Open

enhancement NVIDIA

ByteDance-Seed/Triton-distributed#110 · 2 comments · 0 reactions · 2 assignees · Indexed Sep 14, 2026
Passing num_ctas directly to the Triton kernel is not supported Open

ByteDance-Seed/Triton-distributed#115 · 7 comments · 0 reactions · 1 assignee · Indexed Sep 14, 2026
All2all performance comparison between Triton-distributed and DeepEP Open

bug

ByteDance-Seed/Triton-distributed#117 · 5 comments · 1 reaction · 0 assignees · Indexed Sep 14, 2026
Overlap is slower than non-overlap in MoE+RS on A100*8 Open

enhancement

ByteDance-Seed/Triton-distributed#125 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Bug: assert_allclose fails in test_ag_moe.py for large model shape (K=14336) due to numerical divergence Open

enhancement

ByteDance-Seed/Triton-distributed#127 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 14, 2026
Bug:TP=4 init_nvshmem_by_torch_process_group errors with PAI-Megatron-Patch on 8 X A100 Open

ByteDance-Seed/Triton-distributed#132 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
TileLink release plans Open

ByteDance-Seed/Triton-distributed#135 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 14, 2026
Why some tile task does NOT set InputDependencyDesc property? Open

ByteDance-Seed/Triton-distributed#141 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 14, 2026
[Bug] `symm_at` fails with `constexpr_type` when buffer parameter is None - "'constexpr_type' object has no attribute 'is_block'" Open

ByteDance-Seed/Triton-distributed#152 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 14, 2026
[BUG] Multi-node reduce-scatter crash due to unregistered output buffer in NVSHMEM P2P Open

ByteDance-Seed/Triton-distributed#155 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Possible incorrect barrier offset in `tile_kernel_dispatch_token_intra_node` zero-token expert path Open

ByteDance-Seed/Triton-distributed#169 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
[feature] Implement of UniEP? Open

ByteDance-Seed/Triton-distributed#176 · 0 comments · 1 reaction · 0 assignees · Indexed Sep 14, 2026
Are there any plans to support nvshmem4py-cu13? Open

ByteDance-Seed/Triton-distributed#177 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
swiglu_backward writes every row's grad_fc1_output to row 0 (missing dA/dB output row-pointer offset) → NaN backward in fused EP-MoE Open

ByteDance-Seed/Triton-distributed#180 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
`calc_gather_scatter_index_v2_kernel`: unmasked histogram reduction counts padding lanes into expert 0 when `ntokens * topk % 1024 != 0` Open

ByteDance-Seed/Triton-distributed#181 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
[BUG] `reduce_1d_persistent_kernel`: `tl.atomic_add` into a `torch.empty` Open

ByteDance-Seed/Triton-distributed#182 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
gen_amd_libdevice_extra.py --out cannot accept a file path Open

ByteDance-Seed/Triton-distributed#186 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
LittleKernel Scalar Types table is missing from rendered docs Open

ByteDance-Seed/Triton-distributed#188 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Build guide references nonexistent .codebase test scripts Open

ByteDance-Seed/Triton-distributed#190 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
setup.py fails to recognize repository checkouts Open

ByteDance-Seed/Triton-distributed#192 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Kernel API reference signatures are out of sync with implementations Open

ByteDance-Seed/Triton-distributed#194 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Strict Sphinx build has duplicate objects and unresolved PyTorch references Open

ByteDance-Seed/Triton-distributed#195 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Native license check fails for 39 tracked source files Open

ByteDance-Seed/Triton-distributed#198 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
Pinned Ruff check reports 21 errors in AMD modules Open

ByteDance-Seed/Triton-distributed#199 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
gqa_fwd_batch_decode: every non-power-of-two head dim fails to compile — the rope-split Q load indexes the wrong axis Open

ByteDance-Seed/Triton-distributed#200 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
copy_tensor: the persistent 2-D arm computes row offsets in int32, so a copy whose row extent passes 2 GiB reads and writes out of bounds Open

ByteDance-Seed/Triton-distributed#201 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026
[NVIDIA] low_latency_all_to_all_v2: int32 buffer offsets overflow at large num_experts * max_m * hidden, corrupting dispatch/combine Open

ByteDance-Seed/Triton-distributed#203 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 14, 2026

A short digest of beginner-friendly GitHub issues.