Skip to content

Unturtle: the dLLM method layer on top of unsloth — conversion, objectives, inference acceleration, and canonical evaluation for diffusion language models

active 2026-03-302026-08-12 (UTC)

Complete coverage26,389 / 26,389 hourly files (100%) · 2 absent upstream2023-08-152026-08-18 (UTC)
Events
536
Pushes
157
Pull requests
57
Issues
113
Stars
0
Forks
0

Activity over time

Daily event counts in the loaded window

Line chart, 136 days from 2026-03-30 to 2026-08-12. Pushes: 157 total, peak 14 in a day. Pull requests: 57 total, peak 12 in a day. Issues: 113 total, peak 30 in a day. Comments: 33 total, peak 6 in a day. Stars: 0 total, peak 0 in a day.

  • Pushes
  • Pull requests
  • Issues
  • Comments
  • Stars

Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.

Top contributors

Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity

ContributorContributionsPushesPRsComments
nishide-dev3601575733

Recent activity

Latest issues, pull requests and releases

  • Issue#163nishide-dev2026-08-12 17:58
    [P2][docs][testing] mark real-checkpoint tests by actual resource requirement, not checkpoint identity
  • Issue comment#160nishide-dev2026-08-12 13:32
    ✨ feat(elf-pack): #153 ELF reference method pack — verbatim-port parity, plugin isolation, official OWT cell (#153)
  • Pull request#133nishide-dev2026-08-10 13:31
  • Pull request#118nishide-dev2026-08-09 14:39
  • Issue comment#88nishide-dev2026-08-08 03:00
    ✨ feat(benchmarks): end-to-end sparse LM-head benchmark + backbone support decision (#61)
  • Issue comment#67nishide-dev2026-08-07 16:49
    [P0][roadmap] evolve Unturtle into the fast systems/method layer for diffusion + flow language models
  • Issue comment#66nishide-dev2026-08-07 16:49
    [P2][research] RFC + minimal backend for continuous/latent diffusion and flow-matching LMs
  • Pull request#86nishide-dev2026-08-07 16:14
  • Pull request#26nishide-dev2026-06-12 04:31
  • Issue#24nishide-dev2026-06-12 02:32
    [P3] align GSM8KEvaluator mask_token_id resolution with harness adapter
  • Issue#303nishide-dev2026-05-31 05:51
    [survey] dLM literature survey + adoption queue
  • Issue#297nishide-dev2026-05-29 15:06
    [Phase A2] Physically migrate models into taxonomy layout
  • Issue#288nishide-dev2026-04-29 07:18
    [Phase N] rename eval/formal → eval/lighteval and GSM8K full-run
  • Issue#286nishide-dev2026-04-25 13:50
    [Phase 2] add Unturtle custom model adapter for lighteval
  • Issue comment#284nishide-dev2026-04-25 13:14
    [Phase 1] add lighteval formal evaluation scaffold
  • Pull request#285nishide-dev2026-04-25 11:19
  • Issue#280nishide-dev2026-04-24 04:52
    chore: resolve ruff/ty violations across repo and add PR CI workflow
  • Issue#266nishide-dev2026-04-23 05:06
    [Studio] Chat generation: apply chat template and stream intermediate denoising steps
  • Pull request#275nishide-dev2026-04-23 04:53
  • Pull request#273nishide-dev2026-04-23 02:42
  • Issue#272nishide-dev2026-04-22 06:57
    feat(studio): BD3LM block-diffusion streaming via step/stream callbacks
  • Issue#272nishide-dev2026-04-22 06:57
    feat(studio): BD3LM block-diffusion streaming via step/stream callbacks
  • Issue comment#263nishide-dev2026-04-21 03:47
    fix(a2d): Qwen2/3 forward use layer_types[i] (transformers>=4.57)
  • Issue comment#260nishide-dev2026-04-21 01:36
    ✨ feat(diffusion): OpenAI-compatible diffusion streaming
  • Issue#255nishide-dev2026-04-20 04:55
    install.sh: observability & test-coverage follow-ups from PR #254 review

Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 0 stars here means stars gained during the window, not the repo's star count.