All pull requests

Pull requests

• feat(deepseek4): serve image requests in the concurrent batch

  #760·davide221 opened Sep 23, 2026Contributor·

  ·

• ci(e2e): run Qwen and DS4 end to end on any free lucebox

  #756·Graffioh opened Sep 23, 2026Collaborator·

  ·
  3

• feat(launch): UX improvements regarding auto device placement, DeepSeek V4 flags and profiles, pass-through container entrypoint

  #753·Graffioh opened Sep 23, 2026Collaborator·

  ·
  1

  13

• fix(cmake): adopt DFLASH27B_* options left from before the rename

  #751·Graffioh opened Sep 22, 2026Collaborator·

  ·
  4

• feat(pflash): auto skip-park residency with VRAM estimate and OOM fallback

  #747·Graffioh opened Sep 20, 2026Collaborator·

  ·

• feat(pflash): Qwen3.5-0.8B drafter scorer, segment probe and strict budget selection

  #746·Graffioh opened Sep 19, 2026Collaborator·

  ·

• DO NOT MERGE

  #745·jagtesh opened Sep 19, 2026·

  ·
  41

• fix(dflash): let the drafter see its trained context window (deep-context accept)

  #740·dusterbloom opened Sep 17, 2026Collaborator·

  ·
  14

• feat(server): opt-in logprobs on chat completions (AR path, qwen35 + ds4)

  #737·Graffioh opened Sep 16, 2026Collaborator·

  ·

• test(server): strengthen regression coverage and harden query-window search

  #734·Graffioh opened Sep 14, 2026Collaborator·

  ·
  2

• feat(server): carry drafter feature ring in concurrent prefix checkpoints

  #730·Graffioh opened Sep 14, 2026Collaborator·

  ·
  12

• refactor(test): organize component suites and replace simulated coverage

  #709·Graffioh opened Sep 5, 2026Collaborator·

  ·
  34

• perf(dspark): keep the Markov chain graph instead of rebuilding it per call

  #705·maikzz32 opened Sep 4, 2026Contributor·

  ·
  4

• Qwen35: preserve speculation during mixed prefill and stabilize decode arithmetic

  #686·Graffioh opened Aug 31, 2026Collaborator·

  ·

• feat(pflash): improve AMD defaults and enable rocWMMA

  #669·Graffioh opened Aug 27, 2026Collaborator·

  ·

• experimental: integrate luce_odistill bridge

  #663·Graffioh opened Aug 26, 2026Collaborator·

  ·

• feat(test): add manual Qwen 3.6 production qualification

  #636·Graffioh opened Aug 20, 2026Collaborator·

  ·
  135

• perf(ds4): elide intermediate prefill logits

  #633·cheese-cakee opened Aug 18, 2026Contributor·

  ·
  10

• fix(qwen35): size mixed concurrent target-step graphs

  #622·Graffioh opened Aug 18, 2026Collaborator·

  ·

• feat(harness): add interactive modes and GPU selection

  #611·Graffioh opened Aug 15, 2026Collaborator·

  ·
  5

• feat(mix): batched (MMQ) path for qtypes 105/106, routed on batch width

  #607·geometric[bot] opened Aug 13, 2026Bot·

  ·
  1

• build(deps): bump the github-actions group across 1 directory with 9 updates

  #601·dependabot[bot] opened Aug 11, 2026Bot·

  ·
  1

• test(dflash): add exact prefill differential tracing

  #592·cheese-cakee opened Aug 9, 2026Contributor·

  ·

• perf(dflash): add opt-in exact prefill bands

  #590·cheese-cakee opened Aug 6, 2026Contributor·

  ·
  8

• feat(deepseek4): add experimental PFlash prefill compression

  #589·davide221 opened Aug 6, 2026Contributor·

  ·