A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit floating point (FP8) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.
active 2025-02-04 → 2026-08-08 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 551 days from 2025-02-04 to 2026-08-08. Pushes: 565 total, peak 7 in a day. Pull requests: 941 total, peak 12 in a day. Issues: 358 total, peak 33 in a day. Comments: 4,414 total, peak 66 in a day. Stars: 469 total, peak 6 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 |
|---|---|---|---|---|
| greptile-apps[bot] | 1.4K | 0 | 0 | 744 |
| timmoon10 | 1.3K | 97 | 157 | 603 |
| ptrendx | 997 | 91 | 45 | 452 |
| ksivaman | 583 | 103 | 116 | 203 |
| jberchtold-nvidia | 565 | 30 | 85 | 262 |
| phu0ngng | 425 | 40 | 64 | 209 |
| cyanguwa | 394 | 18 | 28 | 186 |
| zhongbozhu | 352 | 0 | 14 | 197 |
| pggPL | 278 | 23 | 44 | 136 |
| vthumbe1503 | 248 | 20 | 14 | 127 |
| KshitijLakhani | 245 | 54 | 40 | 84 |
| yaox12 | 231 | 19 | 25 | 136 |
| tdophung | 169 | 15 | 26 | 69 |
| sudhakarsingh27 | 164 | 11 | 16 | 78 |
| kwyss-nvidia | 141 | 0 | 2 | 81 |
| sbhavani | 96 | 0 | 4 | 17 |
| denera | 78 | 3 | 18 | 32 |
| Oleg-Goncharov | 57 | 2 | 13 | 26 |
| negvet | 57 | 2 | 5 | 31 |
| janekb04 | 49 | 0 | 11 | 22 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#3318adityasingh24002026-08-08 02:48[PyTorch] Advance FusedAdam step counter for empty param groups
- Issue comment#3328greptile-apps[bot]2026-08-07 19:07[Bug] use_single_group_runtime_offsets arg is not supported for ScaledSrelu
- Issue comment#3327denera2026-08-07 18:03[PyTorch] Fix deferred initialization in fusible ops
- Pull request#3301github-actions[bot]2026-08-03 15:29
- Issue comment#2668CodersAcademy0062026-07-09 05:15[Inquiry] Support status and roadmap for MXFP8 on SM120 (Blackwell)
- Issue comment#1910sbhavani2026-07-02 22:31How to Perform 3D Tensor Multiplication with FP8 Data Type (Beyond te.Linear)?
- Pull request#3167dependabot[bot]2026-07-02 08:29
- Pull request#3163vthumbe15032026-07-01 22:43
- Issue comment#439sbhavani2026-06-29 23:20Parallel residual Transformer layer
- Issue#439sbhavani2026-06-29 23:20Parallel residual Transformer layer
- Issue#556sbhavani2026-06-29 23:19[PYTORCH::FP8] FP8 significantly slow down when scaling up to 1000+ GPUs
- Issue comment#556sbhavani2026-06-29 23:19[PYTORCH::FP8] FP8 significantly slow down when scaling up to 1000+ GPUs
- Issue#417sbhavani2026-06-29 21:41View size incompatible with tensor's size and stride
- Pull request#3104pggPL2026-06-29 09:12
- Issue comment#3143vthumbe15032026-06-26 23:49Graph Safe Current Scaling Support for GroupedLinear Module/Ops
- Issue comment#3042zianglih2026-06-15 01:12Use cuDNN for row-scaled NVFP4 grouped GEMM
- Issue comment#3127timmoon102026-06-14 06:32Expert Parallelism: common C API + NCCL EP backend
- Pull request#3126phu0ngng2026-06-13 01:46
- Issue comment#3036phu0ngng2026-06-11 00:02[JAX] Expert Parallelism: JAX primitives + VJPs
- Issue comment#3034phu0ngng2026-06-10 23:50Expert Parallelism: common C API + NCCL EP backend
- Issue#2500KshitijLakhani2026-06-09 17:48[JAX] Sharding constraint for rng_state in fused_attn_fwd makes it is impossible to use in shard_map
- Pull request#3102int-smart2026-06-07 05:05
- Pull request#3099zhongbozhu2026-06-06 07:38
- Pull request#3099github-actions[bot]2026-06-06 07:38
- Pull request#3094francesco-bertolotti2026-06-06 02:23
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 469 stars here means stars gained during the window, not the repo's star count.