MoE training for Me and You and maybe other people
Python · active 2025-12-17 → 2026-05-02 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 137 days from 2025-12-17 to 2026-05-02. Pushes: 32 total, peak 8 in a day. Pull requests: 5 total, peak 3 in a day. Issues: 8 total, peak 4 in a day. Comments: 12 total, peak 3 in a day. Stars: 130 total, peak 39 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
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| xjdr-noumena | 46 | 32 | 5 | 2 |
| github-actions[bot] | 10 | 0 | 0 | 10 |
| zianglih | 1 | 0 | 0 | 0 |
Recent activity
Latest issues, pull requests and releases
- Issue#15zianglih2026-01-18 22:10[RESEARCH] Clarifications about NVFP4 vs MXFP4 format
- Issue#7xjdr-noumena2026-01-16 04:37[03b] Frozen experts for attention ablations
- Issue comment#10xjdr-noumena2026-01-16 04:37[06] Chunked cross-entropy for large vocab
- Issue#14xjdr-noumena2026-01-03 17:36Inference serving system (nmoe.serve)
- Issue comment#13github-actions[bot]2026-01-02 21:48Multi-Token Prediction (MTP) module (Issue 11)
- Pull request#13xjdr-noumena2026-01-02 21:35
- Issue comment#12github-actions[bot]2026-01-02 03:35[07] Eval framework: choice-based + generation-based evals
- Issue comment#11github-actions[bot]2026-01-02 02:47[08] ZeRO-2 chunked reduce-scatter and checkpoint reliability
- Issue comment#10xjdr-noumena2026-01-01 22:48[06] Chunked cross-entropy for large vocab
- Issue#8xjdr-noumena2026-01-01 20:56[09] Blockscaled expert OOM: fused SwiGLU+quant epilogue
- Pull request#9xjdr-noumena2026-01-01 20:56
- Issue comment#9github-actions[bot]2026-01-01 20:32[09] Blockscaled expert OOM: fused SwiGLU+quant epilogue
- Issue comment#9github-actions[bot]2025-12-31 23:29[09] Blockscaled expert OOM: fused SwiGLU+quant epilogue
- Pull request#9xjdr-noumena2025-12-31 23:02
- Issue#7xjdr-noumena2025-12-31 22:08[03b] Frozen experts for attention ablations
- Issue#5xjdr-noumena2025-12-31 22:07[03a] Global/local attention: MLA + SWA mixing
- Pull request#6xjdr-noumena2025-12-31 22:07
- Issue#5xjdr-noumena2025-12-31 21:43[03a] Global/local attention: MLA + SWA mixing
- Pull request#4xjdr-noumena2025-12-31 21:40
- Issue#3xjdr-noumena2025-12-31 20:13[02] Data pipeline: deterministic flow + HYDRA parity
- Issue comment#2github-actions[bot]2025-12-31 00:24[01] RDEP kernels: SonicMoE fusion
- Issue comment#2github-actions[bot]2025-12-30 22:16[01] RDEP kernels: SonicMoE fusion
- Issue comment#2github-actions[bot]2025-12-30 21:34[01] RDEP kernels: SonicMoE fusion
- Issue comment#2github-actions[bot]2025-12-30 20:44[01] RDEP kernels: SonicMoE fusion
- Issue comment#2github-actions[bot]2025-12-29 21:27[01] RDEP kernels: SonicMoE fusion
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 130 stars here means stars gained during the window, not the repo's star count.