Skip to content

Pull requests: deepspeedai/DeepSpeed

New pull request New
157 Open 3,826 Closed
157 Open 3,826 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

Isolate DeepCompile list-schedule test ops
#8319 opened Aug 25, 2026 by tohtana Collaborator Loading…
2
Add configurable dtype for ZeRO checkpoint export
#8318 opened Aug 25, 2026 by gaoxiaomo Loading…
1
4
Fix the seq-first Ulysses all2all output layout
#8317 opened Aug 25, 2026 by vineethsaivs Contributor Loading…
[Ulysses] Carry the KV head count per DistributedAttention
#8316 opened Aug 25, 2026 by alanhuangyoo Contributor Loading…
1
1
Raise the per-tensor norms to norm_type when combining them
#8313 opened Aug 24, 2026 by vineethsaivs Contributor Loading…
6
Forward barrier device_ids to communication backends
#8312 opened Aug 24, 2026 by tohtana Collaborator Loading…
2
[AutoTP] Add support for vocab-parallel LM head and related configurations
#8309 opened Aug 24, 2026 by jinyouzhi Contributor • Draft
Honor adam_w_mode in the CPU multi_tensor_adam binding
#8307 opened Aug 24, 2026 by PKUWZP Collaborator Loading…
6
Recognize Qwen3.5's RMSNorm variants in AutoTP module loading
#8306 opened Aug 24, 2026 by promptsmith1990 Loading…
1
5
Forward free_data through partition() instead of hardcoding True
#8305 opened Aug 24, 2026 by vineethsaivs Contributor Loading…
3
Honor --include/--exclude in the SLURM launcher
#8304 opened Aug 24, 2026 by vineethsaivs Contributor Loading…
6
[AutoTP] Fix training lm_head routing
#8302 opened Aug 24, 2026 by gaoxiaomo Loading…
9
[Apple Silicon Support Phase 1] Add Metal FusedAdam kernel and CPU Adam build for Apple Silicon
#8300 opened Aug 23, 2026 by PKUWZP Collaborator Loading…
15
feat(opsd): share prompt prefill across rollout samples
#8296 opened Aug 23, 2026 by nathon-lee Contributor Loading…
5
Enable optimized Adam backend for MuonWithAuxAdam optimizer
#8278 opened Aug 19, 2026 by jinyouzhi Contributor Loading…
22
Fix ZeRO parameter alignment for grouped_mm
#8277 opened Aug 19, 2026 by fwerkor Contributor Loading…
1
10
fix: Only bind device id when needed, Fixes #8248
#8269 opened Aug 17, 2026 by pengdurice Contributor Loading…
1
4
Make the WarmupCosineLR ratio flags reach the config
#8268 opened Aug 17, 2026 by vineethsaivs Contributor Loading…
1
Fix comms logger KeyError when log_name is omitted
#8267 opened Aug 17, 2026 by jinyouzhi Contributor Loading…
4
Fallback for unsupported Hybrid Engine policies
#8265 opened Aug 17, 2026 by LiRunGuo Loading…
1
2
Fix ZeRO-3 synchronization during OPSD rollout
#8264 opened Aug 17, 2026 by LiRunGuo Loading…
1
4
Fix Triton NFS detection crash when df wraps long device names
#8259 opened Aug 16, 2026 by atirna Loading…
1
2
Route ZeRO/SuperOffload pin sites through accelerator pin_memory
#8256 opened Aug 15, 2026 by sfc-gh-truwase Collaborator Loading…
4 tasks done
27
Fix shared loss gradient accumulation
#8245 opened Aug 11, 2026 by tohtana Collaborator Loading…
1
3
[AutoTP] Replace tp_shard process-wide globals with per-model AutoTPMeta
#8241 opened Aug 10, 2026 by delock Collaborator Loading…
1
17
ProTip! Follow long discussions with comments:>50.