- 13 Aug, 2025 2 commits
-
-
Paweł Gadziński authored
* code drop Signed-off-by:
Pawel Gadzinski <pgadzinski@nvidia.com> * fix Signed-off-by:
Kirthi Shankar Sivamani <ksivamani@nvidia.com> Co-authored-by:
pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by:
-
jberchtold-nvidia authored
* Add L2_jax_distributed_unittest Signed-off-by:
Jeremy Berchtold <jberchtold@nvidia.com> * Add L1 entry for NORM_INPUT_SHAPES that was missing Signed-off-by:
-
- 09 Aug, 2025 1 commit
-
-
Daniel Stokes authored
* fix: Add stream synchronization before destroying MPI communicator (#1979) Signed-off-by:
djns99 <40156487+djns99@users.noreply.github.com> * feat: Implement column-wise userbuffer overlap for comm+GEMM operations Add support for overlapping column-wise allgather communication with GEMM operations to improve training performance: * **Core infrastructure changes:** - Update bulk_overlap_columnwise_ag() to accept explicit stream parameter - Modify userbuffers send/recv loops to use rank-ordered iteration - Add userbuffers_send_all/recv_all function declarations * **Python integration:** - Add bulk_overlap_ag_with_external_gemm() C++ extension function - Expose new overlap function via pybind11 bindings - Update overlap method configurations to include more ring_exchange ops * **LayerNorm MLP optimization:** - Enable column-wise quantization for FC2 gradient output - Implement overlap of allgather communication with FC2 DGRAD GEMM - Use fill_userbuffers_buffer_for_all_gather for efficient buffering This optimization allows overlapping communication and computation phases more effectively, reducing training wall-clock time by hiding allgather latency behind GEMM execution. Signed-off-by:
Tim Moon <4406448+timmoon10@users.noreply.github.com>
-
- 08 Aug, 2025 2 commits
-
-
Paweł Gadziński authored
* turn on userbuffers for layers without debug Signed-off-by:
-
Phuong Nguyen authored
* rm cudaGraph compatible trait from GroupedGEMM and groupedQuantize Signed-off-by:
Phuong Nguyen <phuonguyen@nvidia.com> * add grouped_gemm jitting in the unit test Signed-off-by:
-
- 07 Aug, 2025 1 commit
-
-
Phuong Nguyen authored
* rm batch_dim, sequence_dim, sequence_parallel_output Signed-off-by:
-
- 06 Aug, 2025 2 commits
-
-
hx authored
[PyTorch] fix input_quantizer usage for save_original_input; fix blockwise FP8 convert_and_update_tensor (#1978) * fix input_quantizer in save_original_input bwd Signed-off-by:
Hongxiao Bai <hongxiaob@nvidia.com> * fix get shape of blockwise tensor with only compact colwise data Signed-off-by:
-
jberchtold-nvidia authored
* Pytest timings Signed-off-by:
-
- 04 Aug, 2025 1 commit
-
-
Tim Moon authored
* Add basic kernel for swapping first two tensor dims Signed-off-by:
Tim Moon <tmoon@nvidia.com> * Add NVRTC kernel for swapping first dims Signed-off-by:
-
- 01 Aug, 2025 1 commit
-
-
Autumn1998 authored
* fix bug if all values<0 Signed-off-by:
tongliu <tongliu@nvidia.com> * minor fix Signed-off-by:
Xin Yao <xiny@nvidia.com>
-
- 29 Jul, 2025 2 commits
-
-
Tim Moon authored
* remove reciprocal op Signed-off-by:
zhongboz <zhongboz@nvidia.com> * Refactor Quantizer::create_tensor function Signed-off-by:
-
Kirthi Shankar Sivamani authored
* Add verbosity only for failing tests Signed-off-by:
-
- 25 Jul, 2025 3 commits
-
-
Tim Moon authored
* Add ops for dropout and constant scale Signed-off-by:
-
Kirthi Shankar Sivamani authored
* Remove deprecated device arg Signed-off-by:
-
Evgeny Tsykunov authored
* Support RMSNorm for QK Signed-off-by:
Evgeny <etsykunov@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * rms -> RMSNorm, l2 -> L2Normalization (align with current pattern) Signed-off-by:
-
- 24 Jul, 2025 4 commits
-
-
Oleg Goncharov authored
* Fixed integer overflow when computing offsets Signed-off-by:
Oleg Goncharov <ogoncharov@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by:
-
Jan Bielak authored
* Mark output tensors as not deletable in backward Signed-off-by:
Jan Bielak <jbielak@nvidia.com> * Add `in_place` kwarg to `MakeExtraOutput` Signed-off-by:
-
Evgeny Tsykunov authored
* Increase intermediate precision and reuse tensors from fwd Signed-off-by:
-
Phuong Nguyen authored
* add manage_primitives() helper * disable GEMM primitives for non-MXFP8 recipes * implement the NVTE_JAX_CUSTOM_CALLS + deprecate NVTE_JAX_CUSTOM_CALLS_RE * replace NVTE_JAX_CUSTOM_CALLS_RE with NVTE_JAX_CUSTOM_CALLS in TE tests and examples * fix use_jax_gemm contextmanager Signed-off-by:
-
- 23 Jul, 2025 2 commits
-
-
jberchtold-nvidia authored
Fix current scaling test_helper.py and enable test_helper.py in L0 Signed-off-by: -
Charlene Yang authored
* fix current device for cuDNN/cuBLAS handles Signed-off-by:
Charlene Yang <8636796+cyanguwa@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add unit test Signed-off-by:
-
- 22 Jul, 2025 2 commits
-
-
Jan Bielak authored
* Refactor _OperationFuserAutogradFunction.forward to use less parameters Signed-off-by:
-
Oleg Goncharov authored
* Fixed conflicts Signed-off-by:
-
- 21 Jul, 2025 1 commit
-
-
Charlene Yang authored
* exclude 9.10.0/.1 for certain configs Signed-off-by:
-
- 19 Jul, 2025 1 commit
-
-
jberchtold-nvidia authored
Update tolerance of distributed layernorm MLP for FP8 Signed-off-by:
-
- 18 Jul, 2025 2 commits
-
-
Phuong Nguyen authored
* enable cudnn norm tests Signed-off-by:
-
Phuong Nguyen authored
* set precision=HIGHEST for the ref_grouped_gemm impl in the unit test Signed-off-by:
-
- 17 Jul, 2025 2 commits
-
-
Phuong Nguyen authored
* remove unnecessary padding Signed-off-by:
-
hx authored
* save original input Signed-off-by:
-
- 16 Jul, 2025 1 commit
-
-
Paweł Gadziński authored
* some initial code Signed-off-by:
Paweł Gadziński <62263673+pggPL@users.noreply.github.com> * fix Signed-off-by:
root <root@prenyx0221.a51.clusters.nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * lint fix Signed-off-by:
Pawel Gadzinski <pgadzinski@gmail.com> * Update build.yml Signed-off-by:
-
- 15 Jul, 2025 1 commit
-
-
Emmanuel Ferdman authored
* [JAX] Resolve test conflict in JAX helper tests Signed-off-by:
Emmanuel Ferdman <emmanuelferdman@gmail.com> * [JAX] Resolve test conflict in JAX helper tests Signed-off-by:
jberchtold-nvidia <158520091+jberchtold-nvidia@users.noreply.github.com>
-
- 14 Jul, 2025 3 commits
-
-
Alp Dener authored
* added XLA FFI custom op for TE/common nvte_cublas_gemm Signed-off-by:
Alp Dener <adener@nvidia.com> started GemmPrimitive, abstract done Signed-off-by:
-
Autumn1998 authored
* fix underterminsic problem in CI Signed-off-by:
-
hx authored
* optimize permute Signed-off-by:
-
- 12 Jul, 2025 1 commit
-
-
Jan Bielak authored
* Fix clearing tensor data in backward removing is_first_op Signed-off-by:
-
- 11 Jul, 2025 1 commit
-
-
jberchtold-nvidia authored
Update test tolerance for L40 Signed-off-by:
-
- 10 Jul, 2025 2 commits
-
-
Autumn1998 authored
* add router fusion Signed-off-by:
-
Paweł Gadziński authored
* push Signed-off-by:
-
- 09 Jul, 2025 2 commits
-
-
Hua Huang authored
* Fix JAX grouped gemm error on CUDA 12.9.1 & later by using 16B alignment for scale ptr Signed-off-by:
Hua Huang <huah@nvidia.com> * Pad MXFP8 scales with 2*-127 instead of NaNs Signed-off-by:
-
Tim Moon authored
* Add tests for loading previously-generated checkpoints Signed-off-by:
-
