Skip to content

Pull requests: ROCm/ATOM

New pull request New
146 Open 1,155 Closed
146 Open 1,155 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

feat(prezero): wire split-K GEMM prezero into Kimi MLA/MoE decode via a torch.compile pass
#1490 opened Jul 6, 2026 by ColorsWind Loading…
[fix][sgl-atom] avoid Qwen3-32B GSM8K truncation
#1489 opened Jul 6, 2026 by zhangxinyuanliuhengyu Contributor Loading…
fix(moe): auto-degrade FP8 block_n/k from 128 to 64 on alignment mismatch (#442)
#1488 opened Jul 6, 2026 by haowu1234 Loading…
[atom-sglang-accuracy] Modify checkbox to dropdown for sglang accuracy Mi355 inputs
#1485 opened Jul 6, 2026 by junyyang-amd Contributor Loading…
Optimize GLM5.2 MTP
#1481 opened Jul 6, 2026 by whx-sjtu Contributor Loading…
Fix sglang minimax m3 cuda graph capture problem
#1480 opened Jul 6, 2026 by Phi-C Contributor • Draft
1 task
feat(quant): implement fp4_act_quant Triton kernel for DeepSeek-V4 (#807)
#1479 opened Jul 6, 2026 by haowu1234 Loading…
1
1
[Benchmark] Update Kimi2.7 recipe and workflow
#1478 opened Jul 6, 2026 by qichu-yun Contributor Loading…
1 task
perf(quant_v4): Triton Walsh-Hadamard rotate_activation kernel
#1477 opened Jul 6, 2026 by haowu1234 Loading…
1
ci(atomesh): add Spur cluster benchmark support
#1475 opened Jul 6, 2026 by wanzhenchn Contributor • Draft
1 task
enable minimax m3 fp8 index cache and fuse index_score_kernel with partial_topk kernel
#1473 opened Jul 6, 2026 by ganyi1996ppo Contributor • Draft
1 task
3
Enable triton MoE for DSv4 decode
#1469 opened Jul 6, 2026 by azaidy Contributor Loading…
1
debug: run15 cu_seqlens_q stale probe (3-state dump + guard)
#1468 opened Jul 5, 2026 by JiaoliangYu Contributor • Draft
perf: optimize model loading speed
#1465 opened Jul 4, 2026 by ftyghome Loading…
7
[ci][mesh] publish benchmark data on dashboard
#1461 opened Jul 3, 2026 by wanzhenchn Contributor Loading…
[Feature] Supports dequantization of common Quark formats
#1459 opened Jul 3, 2026 by haoyangli0109 Contributor • Draft
[NOT READY][GLM5.2 FP8/MXFP4] optimize atom native GLM5.2
#1458 opened Jul 3, 2026 by zejunchen-zejun Collaborator • Draft
add SemiAnalysis aiperf install to ATOM image
#1457 opened Jul 3, 2026 by Yuechguo Contributor Loading…
2
add trace breakdown skill
#1456 opened Jul 3, 2026 by zhuyuhua-v Collaborator • Draft
1 task
2
fix: all greedy && set seed
#1455 opened Jul 3, 2026 by JiaoliangYu Contributor • Draft
1 task
[atom-vllm] enable prefix cache for deepseek v4
#1454 opened Jul 3, 2026 by whx-sjtu Contributor Loading…
[1/3]refactor mesh worker registry into layered pools
#1448 opened Jul 3, 2026 by Yuechguo Contributor • Draft
Remove legacy proxy, update docs, and enhance scripts
#1447 opened Jul 3, 2026 by Jasen2201 Contributor Loading…
1 task
8
[Frontend] openai: multi-model tool-call parsing + reasoning (GLM / MiniMax-M3 / DSML)
#1443 opened Jul 2, 2026 by yhl-amd Contributor Loading…
[Bugfix] Cancel inference on client disconnect + fix non-stream request leak
#1441 opened Jul 2, 2026 by yhl-amd Contributor Loading…
1
ProTip! Follow long discussions with comments:>50.