-
Notifications
You must be signed in to change notification settings - Fork 528
Pull requests: NVIDIA/TransformerEngine
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[Common] Fix checks in quantize_transpose_vector_blockwise_fp4
#2299
opened Oct 23, 2025 by
jberchtold-nvidia
Loading…
8 of 13 tasks
Submodule checkout during setup
build
Build system
enhancement
New feature or request
#2293
opened Oct 22, 2025 by
ksivaman
Loading…
8 of 13 tasks
[common] Remove kvpacked and qkvpacked attention functions for every kernel type.
#2287
opened Oct 20, 2025 by
pggPL
Loading…
8 of 13 tasks
[common] Misc improvements for attention
2.9.0
#2272
opened Oct 15, 2025 by
cyanguwa
Loading…
8 of 13 tasks
[Draft][JAX] E2E encoder sanity test with synthetic data
#2269
opened Oct 13, 2025 by
jberchtold-nvidia
Loading…
8 of 13 tasks
[PyTorch debug] Fixes to debug tests failures
#2268
opened Oct 13, 2025 by
pggPL
Loading…
7 tasks done
[Draft][JAX] Add "initialize" XLA stage to remaining TE/JAX primitives
#2260
opened Oct 10, 2025 by
jberchtold-nvidia
Loading…
8 of 13 tasks
[Common] Split cast/gated kernels by scaling mode
#2248
opened Oct 8, 2025 by
Oleg-Goncharov
Loading…
4 of 13 tasks
[JAX] xla_home logging during JAX build
#2232
opened Oct 3, 2025 by
jberchtold-nvidia
Loading…
13 tasks
[JAX] Add support for sink attention in JAX
#2225
opened Oct 1, 2025 by
pggPL
Loading…
8 of 13 tasks
Enable SWA with CP for THD input format
#2220
opened Sep 30, 2025 by
sudhakarsingh27
Loading…
1 of 6 tasks
Test to see if SWA and Causal compute can be removed from seqlens and…
#2201
opened Sep 25, 2025 by
KshitijLakhani
•
Draft
13 tasks
Honor COMPACT data_format for FP8 blockwise scales in MoE up-projection path to remove 5× redundant rowwise_scale_inv.T.contiguous() passes
#2199
opened Sep 24, 2025 by
xiaoxi-wangfj
•
Draft
2 of 13 tasks
Previous Next
ProTip!
Updated in the last three days: updated:>2025-10-20.