Skip to content

Pull requests: sgl-project/SpecForge

New pull request New
83 Open 479 Closed
83 Open 479 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

misc(train): rely on tokenizer-owned chat templates; drop the packaged Inkling template plan
#743 opened Aug 1, 2026 by Boreas618 Contributor Loading…
2 of 3 tasks
2
fix(export): preserve both RoPE scaling schemas
#742 opened Aug 1, 2026 by Boreas618 Contributor Loading…
2
docs: document CUDA allocator fragmentation mitigation
#739 opened Aug 1, 2026 by Boreas618 Contributor Loading…
2
fix(data): parse tools JSON only for strings, not native lists
#738 opened Aug 1, 2026 by Boreas618 Contributor Loading…
1 of 6 tasks
1
fix: degrade empty anchor sampling to a masked batch to avoid multi-r…
#737 opened Aug 1, 2026 by Boreas618 Contributor Loading…
2 of 6 tasks
1
feat: add Kimi K3 DSpark disaggregated reproduction
#735 opened Aug 1, 2026 by maocheng23 Collaborator • Draft
5 of 7 tasks
1
Support Inkling managed capture
#734 opened Aug 1, 2026 by maocheng23 Collaborator • Draft
1
Correctness fixes for DFlash training path
#731 opened Jul 31, 2026 by dcw02 Loading…
2
[Feature] Multimodal (image+text) DFlash training via server-side capture (Qwen3.5)
#730 opened Jul 29, 2026 by curnane-lab Contributor Loading…
3 of 6 tasks
3
feat(dspark): add Ulysses sequence parallel training
#728 opened Jul 28, 2026 by maocheng23 Collaborator • Draft
4 of 5 tasks
2
feat(dspark): add Qwen3.5-4B draft config and managed-local NPU disaggregated recipe
#727 opened Jul 28, 2026 by curnane-lab Contributor • Draft
1 of 6 tasks
1
feat(domino): add Qwen3.5-4B managed-local NPU disaggregated recipe
#726 opened Jul 28, 2026 by curnane-lab Contributor • Draft
1 of 6 tasks
1
feat(dspark): add offline USP training support
#725 opened Jul 28, 2026 by calyoung80 Loading…
2 of 6 tasks
2
feat(npu): managed-local disaggregated DFlash online training on Ascend NPU
#722 opened Jul 27, 2026 by curnane-lab Contributor Loading…
1 of 6 tasks
3
Add FSDP-aware hybrid Muon optimizer
#717 opened Jul 22, 2026 by maocheng23 Collaborator • Draft
3 of 6 tasks
1
Fuse Domino cross entropy with Triton
#716 opened Jul 22, 2026 by zou3519 Loading…
4
Domino/Dflash: avoid blocking syncs
#715 opened Jul 21, 2026 by zou3519 Loading…
8
DFlash/Domino: Add support for the FlexAttention FLASH backend
#714 opened Jul 21, 2026 by zou3519 Loading…
@zou3519
11
feat(runtime): add heterogeneous one-producer multi-consumer DFlash fanout
#707 opened Jul 18, 2026 by heiheiha798 • Draft
13
feat: --trim-loss-positions (compute teacher/logits/loss only at supe…
#705 opened Jul 17, 2026 by julyanghar Contributor Loading…
4 tasks done
13
rocm: Domino support (validated + example)
#701 opened Jul 17, 2026 by zhentaocc • Draft
1 of 2 tasks
1
rocm: DFlash support (validated + example)
#700 opened Jul 17, 2026 by zhentaocc • Draft
2 of 3 tasks
1
rocm: EAGLE3 support (lazy yunchang + graceful fa/usp errors + example)
#699 opened Jul 17, 2026 by zhentaocc • Draft
2 of 3 tasks
1
[ROCm] AMD ROCm enablement — tutorial + managed_local Mooncake teardown/pin fixes
#698 opened Jul 17, 2026 by zhentaocc Loading…
5 tasks done
6
[WIP] Add EDSD: Entropy-Driven Design for Faster Speculative Decoding
#696 opened Jul 16, 2026 by KerwinKai Contributor • Draft
6 tasks
6
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.