Skip to content

Pull requests: Luce-Org/lucebox

New pull request New
68 Open 587 Closed
68 Open 587 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(qwen35): decouple pooled-prefill ubatch from the pool chunk size
#741 opened Sep 17, 2026 by dusterbloom Collaborator Loading…
7
fix(dflash): let the drafter see its trained context window (deep-context accept)
#740 opened Sep 17, 2026 by dusterbloom Collaborator Loading…
13
feat(server): opt-in logprobs on chat completions (AR path, qwen35 + ds4)
#737 opened Sep 16, 2026 by Graffioh Contributor • Draft
5 of 6 tasks
test(server): strengthen regression coverage and harden query-window search
#734 opened Sep 14, 2026 by Graffioh Contributor • Draft
2
feat(server): carry drafter feature ring in concurrent prefix checkpoints
#730 opened Sep 14, 2026 by Graffioh Contributor Loading…
5 tasks done
12
feat(ds4v): dflash vision serving runtime with bounded IQ85 conversion
#722 opened Sep 10, 2026 by marcelormendes Contributor Loading…
84
refactor(test): organize component suites and replace simulated coverage
#709 opened Sep 5, 2026 by Graffioh Contributor Loading…
34
perf(dspark): keep the Markov chain graph instead of rebuilding it per call
#705 opened Sep 4, 2026 by maikzz32 Contributor Loading…
3
feat(moe): register mixed-qtype decode tables for hot-only storage
#703 opened Sep 4, 2026 by maikzz32 Contributor Loading…
4
feat(moe): cold-owner None, and expert-major prefill for a hot-only stack
#701 opened Sep 4, 2026 by maikzz32 Contributor Loading…
4
feat(ggml): in-graph collective node via a caller-registered callback
#700 opened Sep 4, 2026 by maikzz32 Contributor Loading…
perf(qwen35): retain speculation during paged prefill
#686 opened Aug 31, 2026 by Graffioh Contributor • Draft
feat(pflash): improve AMD defaults and enable rocWMMA
#669 opened Aug 27, 2026 by Graffioh Contributor • Draft
experimental: integrate luce_odistill bridge
#663 opened Aug 26, 2026 by Graffioh Contributor • Draft
feat(test): add manual Qwen 3.6 production qualification
#636 opened Aug 20, 2026 by Graffioh Contributor • Draft
135
perf(ds4): elide intermediate prefill logits
#633 opened Aug 18, 2026 by cheese-cakee Contributor • Draft
10
fix(qwen35): size mixed concurrent target-step graphs
#622 opened Aug 18, 2026 by Graffioh Contributor • Draft
feat(harness): add interactive modes and GPU selection
#611 opened Aug 15, 2026 by Graffioh Contributor Loading…
2 tasks
5
feat(mix): batched (MMQ) path for qtypes 105/106, routed on batch width
#607 opened Aug 13, 2026 by geometric Bot Loading…
1
feat(harness): add OMP client support
#603 opened Aug 12, 2026 by Graffioh Contributor Loading…
2 tasks
16
build(deps): bump the github-actions group across 1 directory with 9 updates
#601 opened Aug 11, 2026 by dependabot Bot Loading…
1
test(dflash): add exact prefill differential tracing
#592 opened Aug 9, 2026 by cheese-cakee Contributor • Draft
perf(dflash): add opt-in exact prefill bands
#590 opened Aug 6, 2026 by cheese-cakee Contributor Loading…
8
feat(deepseek4): add experimental PFlash prefill compression
#589 opened Aug 6, 2026 by davide221 Contributor • Draft
feat(kimi-k3): add DSpark Q8 artifact and shared speculative runtime
#573 opened Aug 1, 2026 by davide221 Contributor • Draft
3
ProTip! Type g p on any issue or pull request to go back to the pull request listing page.