- 25 Nov, 2025 1 commit
-
-
Zhongbo Zhu authored
* minor fix of torch view dtype Signed-off-by:
Zhongbo Zhu <zhongboz@nvidia.com> * multi-tensor RHT amax, compiles Signed-off-by:
Tim Moon <tmoon@nvidia.com> * Use size_t for split sections Signed-off-by:
pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
-
- 21 Nov, 2025 2 commits
-
-
Tim Moon authored
Only disable Flash Attention in Userbuffers test on A100 Signed-off-by: -
Sudhakar Singh authored
* Add support for THD+CP+SWA through A2A comms Signed-off-by:
Sudhakar Singh <sudhakars@nvidia.com> * unblock the `padding`+`THD`+`CP(A2A)` with SWA case in A2A forward Signed-off-by:
greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com> Signed-off-by:
-
- 19 Nov, 2025 2 commits
-
-
Tim Moon authored
Disable Flash attention in Userbuffers tests Signed-off-by: -
Charlene Yang authored
* fix test_current_device 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 --------- Signed-off-by:
Kirthi Shankar Sivamani <ksivamani@nvidia.com>
-
- 18 Nov, 2025 4 commits
-
-
Jaime authored
[PyTorch] Implement Selective Activation Checkpointing for LayerNormMLP with checkpoint flag (#2311) * custom tests for selective activation checkpointing for layernorm mlp Signed-off-by:
Jaime Cardenas <jaime@evolutionaryscale.ai> * add selective layernorm mlp to te.pytorch Signed-off-by:
Jaime <102792198+jaimec00@users.noreply.github.com> Signed-off-by:
-
Paweł Gadziński authored
* fix Signed-off-by:
Pawel Gadzinski <pgadzinski@nvidia.com> * fix Signed-off-by:
-
Kirthi Shankar Sivamani authored
* Cache device tensors properly Signed-off-by:
-
Paweł Gadziński authored
* fix Signed-off-by:
-
- 17 Nov, 2025 3 commits
-
-
Charlene Yang authored
* [Common] Deleted unused header (#2324) Deleted unused header Signed-off-by:
Oleg Goncharov <ogoncharov@nvidia.com> Signed-off-by:
Peter Dykas <wdykas@nvidia.com> * [JAX] L1_jax_distributed_test suit with individual executions (#2321) * L1 rework Signed-off-by:
Phuong Nguyen <phuonguyen@nvidia.com> * comment out test_multi_process_grouped_gemm for now Signed-off-by:
Jeremy Berchtold <jberchtold@nvidia.com> * Set HF_HUB_OFFLINE to disable any network calls to HF when the pre-downloaded dataset is available Signed-off-by:
Przemek Tredak <ptredak@nvidia.com> Signed-off-by:
kunlunl <kunlunl@nvidia.com> * Rename keep_columnwise to manual_post_all_gather_processing & Optimize unit test Signed-off-by:
Tim Moon <4406448+timmoon10@users.noreply.github.com> Signed-off-by:
Kshitij Lakhani <klakhani@nvidia.com> * Consolidate documentation for DPA in JAX Co-authored-by:
Kshitij Lakhani <33047503+KshitijLakhani@users.noreply.github.com> * Correctly update the documentation for defaults in JAX DPA Co-authored-by:
tdophung <tdophung@nvidia.com> * Formatting Signed-off-by:
Oleg Goncharov <64355998+Oleg-Goncharov@users.noreply.github.com> Co-authored-by:
root <root@gpu-h100-0496.cm.cluster> Co-authored-by:
Paweł Gadziński <62263673+pggPL@users.noreply.github.com> Co-authored-by:
jberchtold-nvidia <158520091+jberchtold-nvidia@users.noreply.github.com> Co-authored-by:
Kunlun Li <94586211+kunlunl@users.noreply.github.com> Co-authored-by:
Michael Goldfarb <mgoldfarb@nvidia.com> Co-authored-by:
wdykas <73254672+wdykas@users.noreply.github.com>
-
Evgeny Tsykunov authored
* Enable reference current scaling recipe Signed-off-by:
Evgeny <etsykunov@nvidia.com> * minor Signed-off-by:
-
Kirthi Shankar Sivamani authored
Initial changes to remove pytorch overheads Signed-off-by:
-
- 14 Nov, 2025 3 commits
-
-
Paweł Gadziński authored
* init Signed-off-by:
root <root@ptyche0312.ptyche.clusters.nvidia.com> Co-authored-by:
-
jberchtold-nvidia authored
* Refactor to avoid storing a global quantization config so direct recipe passing works as intended Signed-off-by:
-
Robin Zhang authored
* reset cudagraph Signed-off-by:
Robin Zhang <robinz@nvidia.com> * use closure instead of mutable default values Signed-off-by:
-
- 13 Nov, 2025 2 commits
-
-
jberchtold-nvidia authored
* Support for checkpointing quantizations Signed-off-by:
JAX Toolbox <jax@nvidia.com> * Checkpoint quantizations Signed-off-by:
-
Evgeny Tsykunov authored
Fix amax computation using output_t data in normalization Signed-off-by:
-
- 12 Nov, 2025 1 commit
-
-
Sudhakar Singh authored
* enable applying rope offsets in backwared Signed-off-by:
-
- 11 Nov, 2025 1 commit
-
-
vthumbe1503 authored
* fix for float8 tensor fsdp2 training Signed-off-by:
Varun Thumbe <vthumbe@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * zeros_like should return fp32 for fsdp2 to work Signed-off-by:
-
- 10 Nov, 2025 1 commit
-
-
Teddy Do authored
* Changing default activations in MLP, TransformerLayer, dropout rate after FC1 to 0, and return_layernorm_output to False Signed-off-by:
-
- 07 Nov, 2025 1 commit
-
-
jberchtold-nvidia authored
* Add test to check jaxpr that amax is reused for nvfp4 recipe Signed-off-by:
-
- 06 Nov, 2025 1 commit
-
-
Kunlun Li authored
* Make cast_master_weights_to_fp8 compatible with older MCore version Signed-off-by:
-
- 05 Nov, 2025 1 commit
-
-
Paweł Gadziński authored
* code drop Signed-off-by:
-
- 04 Nov, 2025 1 commit
-
-
Paweł Gadziński authored
* code drop Signed-off-by:
-
- 03 Nov, 2025 1 commit
-
-
Phuong Nguyen authored
* L1 rework Signed-off-by:
-
- 30 Oct, 2025 2 commits
-
-
Kirthi Shankar Sivamani authored
* Fix attention backend and tests for sm120 Signed-off-by:
-
Kshitij Lakhani authored
* Fix: Skip determinism tests for bprop for all sm >=100 Signed-off-by:
-
- 29 Oct, 2025 1 commit
-
-
vthumbe1503 authored
* changes working Signed-off-by:
Przemyslaw Tredak <ptrendx@gmail.com> Signed-off-by:
-
- 25 Oct, 2025 1 commit
-
-
Charlene Yang authored
* add max_score for fused/unfused F16 non-CP Signed-off-by:
-
- 23 Oct, 2025 2 commits
-
-
Paweł Gadziński authored
* fix perf issue Signed-off-by:
-
jberchtold-nvidia authored
* Make SR rng state always 2D (num_devices, 4) Signed-off-by:
-
- 22 Oct, 2025 2 commits
-
-
jberchtold-nvidia authored
* [JAX] Support recipe flags for disabling SR, RHT, and 2D quantization Signed-off-by:
-
Evgeny Tsykunov authored
* rename experimental -> custom_recipes Signed-off-by:
Evgeny Tsykunov <e.tsykunov@gmail.com> * Update tests/pytorch/test_custom_recipe.py Co-authored-by:
-
- 21 Oct, 2025 3 commits
-
-
Kunlun Li authored
* Add post-processing API for FP8 primary weights to support CUDA Graph Signed-off-by:
-
Oleg Goncharov authored
* Removed activations from NVFP4 CPP tests. Removed CMake debugging flags Signed-off-by:
-
Zhongbo Zhu authored
* pipeclean, fix nvfp4 padding of 32 alignment Signed-off-by:
-
- 20 Oct, 2025 1 commit
-
-
Kirthi Shankar Sivamani authored
* Fix CI failures due to deterministic attention Signed-off-by:
-
- 17 Oct, 2025 2 commits
-
-
Neil Tenenholtz authored
* Fix test of FSDP2 by correcting init logic and applying autocast This fixes multiple issues in the FSDP2 test, namely 1. Previously fp8 init was performed when `args.fp8_init == False`. I have updated the logic to match what I presume was intended by leveraging the nullcontext context manager. 2. `te.fp8_autocast` was previously not called; the recipe was created but was unused. The autocast context manager now wraps the model's computation. Signed-off-by:
Neil Tenenholtz <ntenenz@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix typo Signed-off-by:
-
Kshitij Lakhani authored
* Fix imports in test for deprecated jax.experimental.pjit Signed-off-by:
Kshitij Janardan Lakhani <klakhani@login-eos01.eos.clusters.nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by:
-
- 14 Oct, 2025 1 commit
-
-
Kirthi Shankar Sivamani authored
* Initial API change Signed-off-by:
-
