Skip to content

Official code release for "SuperBPE: Space Travel for Language Models"

active 2025-04-102026-05-20 (UTC)

Partial coverage17,162 / 21,924 hourly files (78%) · 2 absent upstream · 4,761 failed, retryable2024-02-102026-08-12 (UTC)— sampled evenly across the window, so rankings and trends hold; absolute counts scale up.
Events
97
Pushes
14
Pull requests
2
Issues
14
Stars
35
Forks
7

Activity over time

Daily event counts in the loaded window

Line chart, 406 days from 2025-04-10 to 2026-05-20. Pushes: 14 total, peak 6 in a day. Pull requests: 2 total, peak 1 in a day. Issues: 14 total, peak 2 in a day. Comments: 24 total, peak 5 in a day. Stars: 35 total, peak 3 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
PythonNut17915
alisawuffles11505
MattYoon4002
Iambestfeed4004
xiulinyang3003
yi-yvonne-yu2000
Yi-Yu-Yvonne2001
ShreyGanatra2001
narutatsuri2001
huyiwen1001
rrenaud1000
lsy6411000
klasocki1010
wang997111231000
sjcheon861001

Recent activity

Latest issues, pull requests and releases

  • Issue comment#6PythonNut2026-05-03 07:28
    Require a huge amount of memory
  • Issue comment#24ShreyGanatra2026-04-13 05:46
    pre-training olmo code missing
  • Issue comment#23PythonNut2026-04-13 05:46
    SuperBPE produces garbage characters for non-English languages (Sinhala)
  • Issue#24ShreyGanatra2026-04-01 04:32
    pre-training olmo code missing
  • Issue comment#21PythonNut2026-01-17 21:48
    Is there a way to package this?
  • Pull request#22klasocki2025-12-16 17:33
  • Issue#20rrenaud2025-11-11 22:47
    Upstreaming minimal set of changes to HF tokenizers to get rid of the need for the fork
  • Issue comment#19alisawuffles2025-11-04 23:22
    Stage 2 Ignores Outputs from Stage 1?
  • Issue#19narutatsuri2025-10-31 02:24
    Stage 2 Ignores Outputs from Stage 1?
  • Issue comment#14narutatsuri2025-10-31 02:07
    num_inherit_merges has no effect in Stage 2 tokenizer training
  • Issue#17PythonNut2025-10-14 21:16
    Unusually slow training speed
  • Issue#16PythonNut2025-10-14 21:10
    Please add a license
  • Issue comment#15xiulinyang2025-10-06 03:29
    Incorporating superbpe into HF training pipeline
  • Issue comment#15Iambestfeed2025-10-06 03:14
    Incorporating superbpe into HF training pipeline
  • Issue comment#15xiulinyang2025-10-06 03:02
    Incorporating superbpe into HF training pipeline
  • Issue comment#15Iambestfeed2025-10-06 02:25
    Incorporating superbpe into HF training pipeline
  • Issue comment#15xiulinyang2025-10-06 02:02
    Incorporating superbpe into HF training pipeline
  • Issue comment#17Iambestfeed2025-10-05 10:18
    Unusually slow training speed
  • Issue comment#15Iambestfeed2025-10-05 06:31
    Incorporating superbpe into HF training pipeline
  • Issue#14yi-yvonne-yu2025-09-08 03:05
    num_inherit_merges has no effect in Stage 2 tokenizer training
  • Issue#13yi-yvonne-yu2025-09-06 13:40
    Request for tokenizer data-prep scripts (olmo-mix-p99-subset)
  • Issue#12Yi-Yu-Yvonne2025-07-23 17:40
    extend_tokenizer result merges differ from original despite copying merges.txt
  • Issue comment#7alisawuffles2025-07-07 18:27
    Citation request for a similar work at EMNLP 2023 paper
  • Issue comment#4alisawuffles2025-07-01 00:06
    More comparison
  • Issue comment#11alisawuffles2025-06-30 23:35
    Pre-Training data

Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 35 stars here means stars gained during the window, not the repo's star count.