Skip to content

Pull requests: ROCm/ATOM

New pull request New
255 Open 1,800 Closed
255 Open 1,800 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(deepseek-v41): native DeepSeek-V4.1-Flash — text, vision, paged runtime and DSpark
#2314 opened Sep 20, 2026 by valarLip Collaborator • Draft
5 of 6 tasks
43
ci(mesh): benchmark GLM agentic 2P2D with kv_cache_aware routing
#2313 opened Sep 20, 2026 by Jasen2201 Contributor • 5/5 • Draft
1
Prefer PITT /models cache for benchmark containers
#2310 opened Sep 20, 2026 by gyohuangxin Member Loading…
3
feat(mesh): route calibrated HBM and CPU aware prefill/decode pairs
#2308 opened Sep 20, 2026 by Jasen2201 Contributor • 4/5 • Draft
1
feat(cache): publish native HBM and CPU catalog with exact rendering
#2307 opened Sep 20, 2026 by Jasen2201 Contributor • 3/5 • Draft
feat(cache): isolate native layouts and expose execution topology
#2306 opened Sep 20, 2026 by Jasen2201 Contributor • 2/5 • Draft
1
fix(offload): memoize the external-tier lookup per HBM frontier (native + plugin)
#2305 opened Sep 20, 2026 by PerryZhang01 Contributor Loading…
8
ci(release): serve the smoke model from the runner cache
#2304 opened Sep 20, 2026 by zufayu Contributor Loading…
2
[Spec Decode] Add opt-in synthetic forward for forced acceptance
#2301 opened Sep 20, 2026 by gbyu-amd Contributor Loading…
1
Enable GLM-5.3-Flash under vllm serve on the ATOM plugin.
#2299 opened Sep 20, 2026 by sajandhy • Draft
1 task
1
[Lumen-RL] Receive trained weights over RCCL, transactionally
#2298 opened Sep 19, 2026 by i-chaochen Loading…
1 task done
11
[Lumen-RL] Generic collective_rpc across all DP and TP ranks, with capability discovery
#2297 opened Sep 19, 2026 by i-chaochen Loading…
1 task done
5
Add GPT-OSS IQ2R 2-bit MoE integration
#2295 opened Sep 18, 2026 by ssharma4-amd Loading…
2
Add hybrid mega/standard MoE dispatch for MegaMxfp4MoEMethod
#2293 opened Sep 18, 2026 by amd-weisun • Draft
1
[feat] Supports loading Quark NVFP4 and online quantization of NVFP4 to MXFP4 (minimax-m3) ci:atom
#2292 opened Sep 18, 2026 by haoyangli0109 Contributor Loading…
2
GLM-5.3 on the LMCache byte-offload connector: record it, measure it, and fix the slot that would break it on main
#2291 opened Sep 18, 2026 by PerryZhang01 Contributor Loading…
1
[DCP] Allow query replication (QREP) with MTP ci:atom
#2290 opened Sep 18, 2026 by yitingw1 Contributor • Draft
1 task
1
feat(ci): collect per-GPU hardware metrics in agentic reports
#2287 opened Sep 18, 2026 by Jasen2201 Contributor Loading…
1 task
3
[FlyDSL] [Gfx1250] mega moe stage1 gfx1250 no overlap ci:atom
#2286 opened Sep 18, 2026 by XingerZhu Loading…
1 task
3
[sgl+atom] Upgrade SGLang to v0.5.19 ci:sglang
#2280 opened Sep 18, 2026 by zhangxinyuanliuhengyu Contributor Loading…
3 tasks
2
fix(plugin/vllm): keep the platform hook from dropping ATOMPlatform
#2275 opened Sep 17, 2026 by linsun12 Contributor Loading…
3
[Lumen-RL] Enable Rollout Router Replay (R3)
#2274 opened Sep 17, 2026 by sudhu2k Loading…
1 task
@sudhu2k
9
[CI][ROCm][ATOMesh] Add MiniMax-M3-FP8 2P1D TP8 catalog cases
#2272 opened Sep 17, 2026 by avininjamay8 Loading…
4
[minimax-m3]: integrate flydsl index_score kernel
#2269 opened Sep 17, 2026 by ganyi1996ppo Contributor Loading…
13
Add Kimi-K3 (hybrid) LMCache KV offload to the vLLM plugin path
#2268 opened Sep 17, 2026 by PerryZhang01 Contributor Loading…
1
ProTip! Follow long discussions with comments:>50.