Skip to content

Pull requests: AMD-AGI/Primus

New pull request New
27 Open 1,119 Closed
27 Open 1,119 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

fix(megatron): stop the tp_group deprecation warning from breaking compiled graphs
#1169 opened Sep 17, 2026 by jianhan-amd • Draft
21
feat(gpt-oss): enable RCCL-SDMA parameter gather by default
#1168 opened Sep 16, 2026 by suachong Collaborator Loading…
3
refactor(hybrid): rename Hylo stack back to Zebra-Llama
#1167 opened Sep 16, 2026 by vanshbhatia-amd Member Loading…
12
Add qk_clip (MuonClip) patch for hybrid models and the distributed optimizer
#1164 opened Sep 16, 2026 by vanshbhatia-amd Member Loading…
35
Move the MaxDiffusion backend to transformers 5
#1159 opened Sep 15, 2026 by olehtika Contributor • Draft
9
feat(turbo): Llama-3.1-8B FP4 dense MLP, residual RMSNorm, and skip-y
#1156 opened Sep 15, 2026 by kylasa Collaborator Loading…
12
fix(te): recover gfx950 hipBLASLt wgrad failures via workspace retry and torch fallback
#1141 opened Sep 11, 2026 by kailashg26 Contributor • Draft
6 tasks done
3
Feat/diffusion/wan21 megatron training
#1132 opened Sep 10, 2026 by jianhan-amd • Draft
14
refactor(examples): group per-backend material under its backend dire…
#1130 opened Sep 10, 2026 by Xiaoming-AMD Collaborator • Draft
1
test(mxfp6): cover BF16 sensitive-layer routing at the real 57-block shape
#1115 opened Sep 8, 2026 by gphuang Contributor • Draft
2 of 4 tasks
2
perf(mxfp6): fuse Flux QK norm with RoPE and the LN-modulate backward
#1107 opened Sep 6, 2026 by jasainio Contributor • Draft
fix: exclude FP8 quantized weights from MXFP4 de-osc
#1081 opened Sep 3, 2026 by hann-wang Loading…
8
feat(mxfp6): MXFP6 (E2M3) Flux linears, a fused MLP, and the MLPerf recipe
#1064 opened Aug 31, 2026 by jasainio Contributor • Draft
5 of 7 tasks
fix(flux): make MLPerf validation measure what it claims to
#1055 opened Aug 28, 2026 by jasainio Contributor • Draft
5 of 6 tasks
16
[megatron] Implement gemma4
#1053 opened Aug 27, 2026 by yeandy Contributor • Draft
3
feat(megatron): add Qwen3-1.7B model preset and BF16 pretrain configs
#1042 opened Aug 26, 2026 by clairesonglee Collaborator • Draft
fix(deepseek_v4): contract mHC comb on the column-normalised axis
#1040 opened Aug 25, 2026 by yuankaichen-amd Contributor Loading…
[WIP]feat(turbo): route attention through Primus-Turbo's flex_attention when enabled
#1033 opened Aug 25, 2026 by botaohu001 Contributor Loading…
1
feat(mxfp4): switch MXFP4 linears to FP8 at a preset iteration
#1010 opened Aug 21, 2026 by jasainio Contributor • Draft
6 of 7 tasks
1
feat(megatron): support graduated precision boundaries
#980 opened Aug 12, 2026 by gphuang Contributor • Draft
3 of 4 tasks
3
fix(megatron): preserve diffusion resume state
#970 opened Aug 11, 2026 by gphuang Contributor • Draft
4 of 5 tasks
@gphuang
6
ci(config): fail the schema check on real drift, not just report it
#963 opened Aug 10, 2026 by WangLingxun Collaborator • Draft
Fix/ideogram4 ragged cu seqlens compile
#960 opened Aug 10, 2026 by amital-amd Contributor • Draft
2
[Torchtitan] Enable mock dataset to avoid torchtitan data download error
#945 opened Aug 3, 2026 by clairesonglee Collaborator • Draft
[WIP] integrate MORI for FSDP all-gather
#940 opened Jul 31, 2026 by lorri-rao Contributor • Draft
ProTip! Type g i on any issue or pull request to go back to the issue listing page.