Skip to content

Pull requests: ROCm/ATOM

New pull request New
32 Open 198 Closed
32 Open 198 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

[Perf][Qwen3-Next] merge qkvz and ba into qkvzba and add triton sharding kernel
#244 opened Feb 27, 2026 by ganyi1996ppo Loading…
1 task
15
[Perf][Qwen3-Next] Fused shared experts for qwen3_next
#242 opened Feb 27, 2026 by ganyi1996ppo Loading…
1 task
8
CI: Make ATOM benchmark can be called from other repos
#240 opened Feb 26, 2026 by gyohuangxin • Draft
[QUARK-403] Add MiniMax-2.1 support
#237 opened Feb 25, 2026 by thpereir Loading…
1 task
Not yet ready for review- Per-layer quantization config with parser registry
#236 opened Feb 25, 2026 by thpereir • Draft
1 task
fix: resolve prefix caching crashes with MTP speculative decoding
#234 opened Feb 24, 2026 by valarLip • Draft
1 task
5
Refactor ATOM for top-k top-p sampling support
#227 opened Feb 20, 2026 by aryaman-gupta Loading…
4
Enable Triton MOE for MXFP4 on gfx950 (MI355X)
#226 opened Feb 20, 2026 by sunway513 • Draft
5 tasks
2
Add FlyDSL MOE backend and Triton fallback for FP8 MoE
#225 opened Feb 20, 2026 by sunway513 • Draft
1 of 3 tasks
2
[QUARK-402] Add Quark GLM4.7-MXFP4 support
#223 opened Feb 18, 2026 by thpereir Loading…
1 task
6
Fix prefix caching crash: recalculate num_new_tokens after block allocation
#222 opened Feb 17, 2026 by ChuanLi1101 Loading…
2 of 3 tasks
1
Enable Triton MXFP4 MoE on gfx950 for GPT-OSS
#220 opened Feb 16, 2026 by ChuanLi1101 Loading…
4 tasks done
5
Enable AllReduce+RMSNorm fusion for GPT-OSS model
#218 opened Feb 15, 2026 by ChuanLi1101 Loading…
4 tasks done
Fix CI container name collision for parallel matrix jobs
#212 opened Feb 13, 2026 by sunway513 Loading…
3 tasks
11
Add support for Kimi-K2 and Kimi-K25 models in configuration and mode…
#208 opened Feb 12, 2026 by thpereir • Draft
1 task
Revert "CI: Use DeepSeek-R1-0528-mtp-mxfp4 models for deepseek fp4 tests"
#206 opened Feb 11, 2026 by gyohuangxin Loading…
2
support mtp>1
#193 opened Feb 7, 2026 by jiayyu Loading…
1 task
28
Add Flux diffusion model support
#170 opened Jan 29, 2026 by ChuanLi1101 Loading…
@carlushuang @valarLip
2
[POC][Deepseek] Engram support, model_runner hash compute overlap
#168 opened Jan 28, 2026 by ZhangLirong-amd • Draft
1 task
Adding prefill decode markers to trace and enable shapes
#156 opened Jan 20, 2026 by msiddaiah Loading…
1 task
10
[recipe] update qwen3 recipe
#154 opened Jan 20, 2026 by gbyu-amd Loading…
1 task
1
qwen3-235b fp4 support
#151 opened Jan 19, 2026 by gbyu-amd • Draft
1 task
feat: Add fused attention output + RMSNorm support for GPT-OSS
#148 opened Jan 17, 2026 by ChuanLi1101 Loading…
kv and output scale loading bug -- FIX
#146 opened Jan 16, 2026 by amirumoAMD Loading…
1 task done
4
[1/N][feat] Make ATOM work with vLLM and SGLang
#126 opened Jan 12, 2026 by zejunchen-zejun Loading…
299
ProTip! Filter pull requests by the default branch with base:main.