- 21 Oct, 2025 1 commit
-
-
Zhongbo Zhu authored
* pipeclean, fix nvfp4 padding of 32 alignment Signed-off-by:
Zhongbo Zhu <zhongboz@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * numerical test passed Signed-off-by:
Tim Moon <tmoon@nvidia.com> * Fix invalid conversion from tensor to int Signed-off-by:
pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by:
Tim Moon <4406448+timmoon10@users.noreply.github.com>
-
- 20 Oct, 2025 2 commits
-
-
Kirthi Shankar Sivamani authored
* Fix CI failures due to deterministic attention Signed-off-by:
Kirthi Shankar Sivamani <ksivamani@nvidia.com> * some more cleanup Signed-off-by:
-
fzyzcjy authored
* Update permutation.py Signed-off-by:
fzyzcjy <5236035+fzyzcjy@users.noreply.github.com> * Update permutation.py Signed-off-by:
-
- 18 Oct, 2025 1 commit
-
-
Kirthi Shankar Sivamani authored
* Support wheel build for cuda 13 Signed-off-by:
-
- 17 Oct, 2025 4 commits
-
-
Alp Dener authored
Make `CanonicalizeGemmInput()` support non-TN layout FP8 GEMM on Blackwell with column-wise/transposed data (#2233) Modified CanonicalizeGemmInput() logic to pull from column-wise data for FP8 GEMM on Blackwell when row-wise is not available. Signed-off-by:Alp Dener <adener@nvidia.com>
-
Haowen Zheng authored
Signed-off-by:
将来 <jianglai.zhw@alibaba-inc.com> Co-authored-by:
-
Tim Geypens authored
Signed-off-by:Tim Geypens <tim.geypens@gmail.com>
-
Kevin Tong authored
* CUDA RHT Signed-off-by:
Kevin Tong <kevin@augmentcode.com> * Fix cuda graphs Signed-off-by:
-
- 16 Oct, 2025 2 commits
-
-
xiaoxi-wangfj authored
* [PyTorch] Add record_stream and untyped_storage func op in QuantizedTensor Signed-off-by:
xiaoxi-wangfj <690912414@qq.com> * Update transformer_engine/pytorch/tensor/float8_blockwise_tensor.py Co-authored-by:
-
Selvaraj Anandaraj authored
Signed-off-by:
Selvaraj Anandaraj <selvaraja@login-ptyche02.ptyche.clusters.nvidia.com> Co-authored-by:
-
- 15 Oct, 2025 1 commit
-
-
Paul Gibbons authored
* fixes for start_end_list usage in TE debug Signed-off-by:
Paul Gibbons <pgibbons@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by:
Paweł Gadziński <62263673+pggPL@users.noreply.github.com>
-
- 14 Oct, 2025 4 commits
-
-
Tim Moon authored
* Require cuDNN 9.14.0+ for fused attention with FP8 current scaling Signed-off-by:
-
Kirthi Shankar Sivamani authored
* Initial API change Signed-off-by:
-
Kshitij Lakhani authored
* Add BRCM support when creating a test mask for fused attn Signed-off-by:
Kshitij Lakhani <klakhani@nvidia.com> * Add support for BRCM to correctly generate the mask needed for calculating the seqlens and offsets for THD Signed-off-by:
-
Evgeny Tsykunov authored
* Fix update_quantized in ref nvfp4 quantizer Signed-off-by:
Evgeny <etsykunov@nvidia.com> * Subclass quantization API Signed-off-by:
-
- 13 Oct, 2025 5 commits
-
-
Selvaraj Anandaraj authored
* FSDP grad fusion support Signed-off-by:
Selvaraj Anandaraj <selvaraja@login-ptyche01.ptyche.clusters.nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Re-factored grad overwriting usage Signed-off-by:
Selvaraj Anandaraj <selvaraja@nvidia.com> * Update transformer_engine/pytorch/ops/fused/backward_linear_add.py Co-authored-by:
-
jberchtold-nvidia authored
assertion check Signed-off-by:Jeremy Berchtold <jberchtold@nvidia.com>
-
jberchtold-nvidia authored
* Improve error message for cublas fp8 gemm with incorrect shape Signed-off-by:
-
Peter St. John authored
Signed-off-by:
Peter St. John <pstjohn@nvidia.com> Co-authored-by:
-
Selvaraj Anandaraj authored
* Added multi-layout support for attention Signed-off-by:
Pawel Gadzinski <pgadzinski@nvidia.com> --------- Signed-off-by:
-
- 09 Oct, 2025 5 commits
-
-
Peter St. John authored
Don't pickle an empty dict in LayerNorm and BasicOperation layers Signed-off-by:
-
jberchtold-nvidia authored
Signed-off-by:
Phuong Nguyen <phuonguyen@nvidia.com>
-
Kirthi Shankar Sivamani authored
* Update minimum python version to 3.10 and update CI Signed-off-by:
-
Kirthi Shankar Sivamani authored
Deprecate old float8_tensor.py Signed-off-by: -
Kirthi Shankar Sivamani authored
Catch unsupported GEMM during recipe init Signed-off-by:
-
- 08 Oct, 2025 2 commits
-
-
Hua Huang authored
* Try async copy of grouped GEMM group_sizes data Signed-off-by:
Hua Huang <huah@nvidia.com> --------- Signed-off-by:
-
Kirthi Shankar Sivamani authored
Unblock path for fusing NVFP4 quantize and bgrad Signed-off-by:
-
- 07 Oct, 2025 2 commits
-
-
Kirthi Shankar Sivamani authored
* Improve docstring for NVFP4 recipe Signed-off-by:
Przemyslaw Tredak <ptrendx@gmail.com> Signed-off-by:
Copilot <175728472+Copilot@users.noreply.github.com> Signed-off-by:
-
Phuong Nguyen authored
* reuse amax for current scaling Signed-off-by:
-
- 06 Oct, 2025 1 commit
-
-
Phuong Nguyen authored
* not fuse bias for output all reduction case + unit tests Signed-off-by:
-
- 04 Oct, 2025 2 commits
-
-
Kirthi Shankar Sivamani authored
* Fix in FP8 attention selection logic Signed-off-by:
-
Tim Moon authored
Signed-off-by:
-
- 03 Oct, 2025 2 commits
-
-
Jan Bielak authored
* Update to_string(NVTEScalingMode) to include block scaling Signed-off-by:
Jan Bielak <jbielak@nvidia.com> * Add `nvte_swizzle_block_scaling_to_mxfp8_scaling_factors` Signed-off-by:
-
vthumbe1503 authored
Signed-off-by:Varun Thumbe <vthumbe@nvidia.com> *Jax integration for clamped swiglu. This is the continuation of PR which added Clamped Swiglu(used in GPT OSS) support in TE along with Pytorch integration. This PR hooks up the clamped swiglu and dswiglu's nvte APIs to TE Jax.
-
- 02 Oct, 2025 4 commits
-
-
hx authored
* fix overflow of int32 in permute kernels Signed-off-by:
Hongxiao Bai <hongxiaob@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by:
Xin Yao <xiny@nvidia.com> Co-authored-by:
-
jberchtold-nvidia authored
Fix shard map issue Signed-off-by:
-
jberchtold-nvidia authored
Fix code block in fp8_autocast docstring Signed-off-by:
-
Tim Moon authored
* Make sure to set usages for linear op quantizers before forward Signed-off-by:
-
- 01 Oct, 2025 2 commits
-
-
Przemyslaw Tredak authored
* Fix the cublas workspace alignment Signed-off-by:
Przemek Tredak <ptredak@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Co-authored-by:
-
Kirthi Shankar Sivamani authored
Fix passing args to nvfp4 recipe Signed-off-by:
-
