Skip to content

vllm-project/compressed-tensors

View on GitHub ↗Related repositories →

A safetensors extension to efficiently store sparse quantized tensors on disk

active 2025-10-152026-08-05 (UTC)

Partial coverage11,239 / 11,968 hourly files (94%) · 2 absent upstream · 726 failed, retryable2025-03-302026-08-10 (UTC)— sampled evenly across the window, so rankings and trends hold; absolute counts scale up.
Events
1.3K
Pushes
484
Pull requests
123
Issues
19
Stars
34
Forks
10

Activity over time

Daily event counts in the loaded window

Line chart, 295 days from 2025-10-15 to 2026-08-05. Pushes: 484 total, peak 17 in a day. Pull requests: 123 total, peak 6 in a day. Issues: 19 total, peak 3 in a day. Comments: 251 total, peak 15 in a day. Stars: 34 total, peak 2 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
kylesayrs4382532384
brian-dellabetta15471037
dsikka128701220
mergify[bot]9406034
HDCharles9235618
coderabbitai[bot]344016
shanjiaz321735
dhuangnm292530
yiliu3013017
Etelis10036
fynnsu9101
colldata795012
LudovicoYIN5041
andy-neuma4211
rahul-tuli4001
deepak-kumar-neu3300
mgoin3120
AndreasKaratzas2002
ArkaSanka2002
dbarbuzzi2000

Recent activity

Latest issues, pull requests and releases

  • Issue comment#795xin3he2026-07-27 04:01
    [Regression] The broadcast format is incorrect for fp8 attention (Q scale).
  • Pull request#751ZewenShen-Cohere2026-07-01 19:00
  • Issue comment#748coderabbitai[bot]2026-06-23 17:09
    [Offloading] Add warnings for dtype mismatches
  • Pull request#729kylesayrs2026-06-22 21:19
  • Issue#652kylesayrs2026-06-22 17:44
    Kimi-K2.5: AttributeError: 'CompressedLinear' object has no attribute 'weight'
  • Issue comment#728kylesayrs2026-06-10 01:25
    [Offload] Use explicit class patching for `load_offloaded_model`
  • Issue comment#720coderabbitai[bot]2026-06-01 20:03
    Add Buildkite pipeline for test-check on H100 duo and L4 solo
  • Issue comment#684Yatimai2026-05-23 14:15
    fix: use file reference counting to handle shared tensor deletion
  • Issue#711pasta-paul2026-05-20 21:15
    [Bug] `dispatch_with_map` raises `AttributeError: '0' is not an nn.Module` on sharded models (sharded MoE expert ModuleList)
  • Pull request#704HDCharles2026-05-11 20:21
  • Issue comment#700yiliu302026-05-11 00:31
    Support CI test with XPU
  • Pull request#699mergify[bot]2026-05-04 22:59
  • Pull request#699mergify[bot]2026-05-04 22:56
  • Issue#695brian-dellabetta2026-04-29 22:22
    match_named_modules O(N×M) causes multi-minute silent load tail on fine-grained MoE checkpoints (Qwen3-30B-A3B-AWQ, ~18K Linear modules)
  • Issue comment#696brian-dellabetta2026-04-29 21:33
    [not for merge] validate token
  • Issue comment#696mergify[bot]2026-04-29 21:29
    [not for merge] validate token
  • Issue comment#692coderabbitai[bot]2026-04-28 18:47
    [Bugfix] Do not throw "unexpected non-targeted tensor" validation error for ignored tensors
  • Issue comment#684kylesayrs2026-04-28 05:02
    fix: use file reference counting to handle shared tensor deletion
  • Pull request#690kylesayrs2026-04-27 21:58
  • Pull request#662mergify[bot]2026-04-27 20:24
  • Pull request#666mergify[bot]2026-04-22 15:43
  • Issue comment#677Vishnu-sai-teja2026-04-18 06:07
    [convert] Add Converter for AutoAWQ format
  • Pull request#624kylesayrs2026-04-16 02:57
  • Issue comment#683mergify[bot]2026-04-16 01:37
    Add MXFP4 decompression support
  • Issue comment#662vkduy2026-04-14 04:18
    [MPS] Add MPS device offloading and unit tests with pytest

Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 34 stars here means stars gained during the window, not the repo's star count.