Skip to content

Pull requests: sgl-project/SpecForge

New pull request New
85 Open 439 Closed
85 Open 439 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(draft): read RoPE parameters from Transformers v5 configs
#702 opened Jul 17, 2026 by wiketool Contributor Loading…
2 tasks done
2
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
1 of 2 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: lazy-import yunchang so SpecForge imports on ROCm (foundation)
#698 opened Jul 17, 2026 by zhentaocc • Draft
2 of 3 tasks
1
fix(config): drop redundant qwen3-8b-dta config and example
#697 opened Jul 17, 2026 by curnane-lab Contributor Loading…
2 of 6 tasks
1
1
[WIP] Add EDSD: Entropy-Driven Design for Faster Speculative Decoding
#696 opened Jul 16, 2026 by KerwinKai Contributor • Draft
6 tasks
6
[PR678 stack 13/14] docs: finalize the server-only migration
#694 opened Jul 15, 2026 by maocheng23 Collaborator • Draft
1
[PR678 stack 12/14] refactor(runtime)!: enforce server-only training
#693 opened Jul 15, 2026 by maocheng23 Collaborator • Draft
1
[PR678 stack 11/14] refactor(cli)!: make specforge train the single public entry point
#692 opened Jul 15, 2026 by maocheng23 Collaborator • Draft
1
[PR678 stack 10/14] feat(app): add composition and export boundaries
#691 opened Jul 15, 2026 by maocheng23 Collaborator • Draft
1
[PR678 stack 09/14] feat(capture): add prompt media and offline capture leaves
#690 opened Jul 15, 2026 by maocheng23 Collaborator • Draft
1
[PR678 stack 08/14] feat(training): add compatibility assembly spine
#689 opened Jul 15, 2026 by maocheng23 Collaborator • Draft
1
[PR678 stack 07/14] feat(training): add orchestration and lifecycle services
#688 opened Jul 15, 2026 by maocheng23 Collaborator Loading…
2
refactor: consolidate training on the unified runtime
#678 opened Jul 14, 2026 by maocheng23 Collaborator Loading…
3
feat(training): single-GPU online-mode memory options
#671 opened Jul 12, 2026 by notwitcheer Loading…
6
feat(training): --vocab-mapping-path to pin the vocab mapping across resumes
#670 opened Jul 12, 2026 by notwitcheer Loading…
3
feat(training): gradient checkpointing for the TTT unroll (--draft-gradient-checkpointing)
#669 opened Jul 12, 2026 by notwitcheer Loading…
fix(data): pass per-sample tools list for preformatted batches
#668 opened Jul 10, 2026 by curnane-lab Contributor Loading…
1 of 6 tasks
1
feat: add mtp support
#667 opened Jul 9, 2026 by curnane-lab Contributor Loading…
4 of 6 tasks
1
feat(target_utils): auto-detect embed/lm_head keys for LLM and VLM checkpoints
#666 opened Jul 9, 2026 by curnane-lab Contributor Loading…
6 tasks
3
feat: optional Liger fused RMSNorm/SwiGLU kernels for DFlash training
#665 opened Jul 8, 2026 by jessiewei7 Loading…
3
Fix DFlash sglang target backend for sglang 0.5.13+ scheduler/batch API
#664 opened Jul 8, 2026 by violet73 Loading…
6 tasks
2
fix(optimizer): clip by global grad norm across FSDP ranks in BF16Optimizer
#663 opened Jul 8, 2026 by jessiewei7 Loading…
2
fix(train_dflash): atomic per-file checkpoint writes + _SUCCESS-gated resume
#659 opened Jul 7, 2026 by jessiewei7 Loading…
1
ProTip! Mix and match filters to narrow down what you’re looking for.