Skip to content

chrisjmccormick/shared-subspaces

View on GitHub ↗Related repositories →

Research code exploring shared subspaces in attention and feedforward networks. Includes SVD analysis of pre-trained models, experiments with shared output projections, and latent-space-efficient Transformer variants.

active 2025-07-292025-11-17 (UTC)

Complete coverage26,687 / 26,687 hourly files (100%) · 2 absent upstream2023-08-152026-08-30 (UTC)
Events
314
Pushes
204
Pull requests
53
Issues
10
Stars
4
Forks
1

Activity over time

Daily event counts in the loaded window

Line chart, 112 days from 2025-07-29 to 2025-11-17. Pushes: 204 total, peak 56 in a day. Pull requests: 53 total, peak 22 in a day. Issues: 10 total, peak 5 in a day. Comments: 7 total, peak 2 in a day. Stars: 4 total, peak 1 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
chrisjmccormick268204504
Adefioye7033
cursor[bot]1000

Recent activity

Latest issues, pull requests and releases

  • Pull request#21Adefioye2025-10-25 16:10
  • Issue comment#20Adefioye2025-10-24 17:01
    Conducted experiment to investigate the effect of intra-block subspaces on MOE transformer model
  • Issue comment#18chrisjmccormick2025-10-23 23:16
    Add config files to experiment on the performance between dense and MOE blocks in attention layers
  • Issue comment#20chrisjmccormick2025-10-23 23:08
    Conducted experiment to investigate the effect of intra-block subspaces on MOE transformer model
  • Pull request#20Adefioye2025-10-21 07:46
  • Pull request#19chrisjmccormick2025-10-03 17:46
  • Issue comment#19chrisjmccormick2025-10-03 17:41
    Merge all changes for GPT-2 scale pre-training experiments
  • Pull request#19chrisjmccormick2025-10-03 17:41
  • Pull request#17chrisjmccormick2025-09-12 16:53
  • Pull request#16chrisjmccormick2025-09-11 01:16
  • Issue comment#5chrisjmccormick2025-09-10 02:36
    Enable torch.compile and bf16 optimizations in SubspaceEncoder
  • Issue#5chrisjmccormick2025-09-10 02:36
    Enable torch.compile and bf16 optimizations in SubspaceEncoder
  • Pull request#15Adefioye2025-09-07 14:18
  • Pull request#14chrisjmccormick2025-09-07 01:25
  • Pull request#14chrisjmccormick2025-09-07 01:24
  • Pull request#13chrisjmccormick2025-09-07 01:21
  • Pull request#13chrisjmccormick2025-09-07 01:21
  • Issue#4Adefioye2025-09-04 09:54
    Add a better MLM head to the SharedSubspaceEncoder
  • Issue comment#4Adefioye2025-09-04 01:57
    Add a better MLM head to the SharedSubspaceEncoder
  • Issue#12chrisjmccormick2025-09-03 23:46
    Comparative analysis of MLA vs MLA-o model behavior
  • Issue#11chrisjmccormick2025-09-03 23:46
    MoE router health analysis
  • Issue#10chrisjmccormick2025-09-03 23:42
    Query vs Output subspace impact comparison
  • Issue#7chrisjmccormick2025-09-03 22:59
    Fix RoPE dimension split issue for balanced head configurations
  • Issue#6chrisjmccormick2025-09-03 22:58
    Implement example packing for efficient sequence utilization
  • Issue comment#4Adefioye2025-08-20 13:19
    Add a better MLM head to the SharedSubspaceEncoder

Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 4 stars here means stars gained during the window, not the repo's star count.