Skip to content

feifeibear/LLMSpeculativeSampling

View on GitHub ↗Related repositories →

Fast inference from large lauguage models via speculative decoding(original version)

active 2023-08-162026-03-19 (UTC)

Complete coverage26,602 / 26,602 hourly files (100%) · 2 absent upstream2023-08-152026-08-27 (UTC)
Events
1.1K
Pushes
37
Pull requests
39
Issues
32
Stars
799
Forks
87

Activity over time

Daily event counts in the loaded window

Line chart, 947 days from 2023-08-16 to 2026-03-19. Pushes: 37 total, peak 12 in a day. Pull requests: 39 total, peak 15 in a day. Issues: 32 total, peak 4 in a day. Comments: 57 total, peak 7 in a day. Stars: 799 total, peak 17 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
feifeibear113373429
sf-Liu13018
haiduo6006
cliangyu6022
pppyb4001
pendulum4453001
hynnn3002
doquangg2010
tszdanger2001
wenxin-zhu2000
taoxunqiang2002
briancpark2000
huangzl191010
YEOMJINSEOP1000
Sun2sky1261001
java639401000
jayeshthk1001
wiluen1000
zhaoyang-star1000
KONGM1MG1001

Recent activity

Latest issues, pull requests and releases

  • Issue#42YZH09052025-06-19 01:51
    the speed of speculative decoding
  • Issue#41java639402025-05-29 11:08
    why autoregressive_sampling and speculative_sampling_v2 output different tokens
  • Issue comment#36Sun2sky1262025-05-21 07:59
    Deepmind version's probability distribution
  • Issue#40doquangg2025-04-01 17:50
    Incompatibility with HuggingFace Llama 2
  • Pull request#39doquangg2025-04-01 17:33
  • Issue comment#38jayeshthk2025-03-06 17:44
    add timing measurements for autoregressive and speculative sampling m…
  • Pull request#37huangzl192025-01-21 12:13
  • Issue#35feifeibear2024-12-10 09:22
    这个项目和transformers中实现的speculative方法区别是什么
  • Issue comment#34feifeibear2024-11-28 03:05
    Bloom's kvcache are both (bs, head, seq, head_dim) in new version of transformers
  • Issue comment#35feifeibear2024-11-28 03:04
    这个项目和transformers中实现的speculative方法区别是什么
  • Issue comment#34KONGM1MG2024-11-13 09:35
    Bloom's kvcache are both (bs, head, seq, head_dim) in new version of transformers
  • Issue#36YEOMJINSEOP2024-11-02 11:26
    Deepmind version's probability distribution
  • Issue#35wiluen2024-10-11 11:33
    这个项目和transformers中实现的speculative方法区别是什么
  • Issue#34shadow1505192024-10-06 05:58
    Bloom's kvcache are both (bs, head, seq, head_dim) in new version of transformers
  • Issue#33feifeibear2024-08-22 03:35
    Please add a license to this repo
  • Issue comment#33feifeibear2024-08-22 03:34
    Please add a license to this repo
  • Issue#33mbrukman2024-08-21 19:00
    Please add a license to this repo
  • Issue#29feifeibear2024-07-29 01:31
    Algorithm error about target model sampling in Deepmind's version
  • Issue#24feifeibear2024-07-29 01:30
    does it support batchsize > 1
  • Issue#30feifeibear2024-07-29 01:30
    Unable to Reproduce Results from README File
  • Pull request#32feifeibear2024-07-29 01:30
  • Pull request#31feifeibear2024-07-29 01:30
  • Issue comment#29sf-Liu2024-07-26 18:04
    Algorithm error about target model sampling in Deepmind's version
  • Issue comment#29haiduo2024-07-26 03:05
    Algorithm error about target model sampling in Deepmind's version
  • Issue comment#29sf-Liu2024-07-25 22:32
    Algorithm error about target model sampling in Deepmind's version

Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 799 stars here means stars gained during the window, not the repo's star count.