[pull] main from NVIDIA:main#604
Merged
Merged
Conversation
* Disable the RHT fusion for non-SM100 family devices 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 the compilation error 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 --------- Signed-off-by: Przemek Tredak <ptredak@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
… buffer (#2900) * [PyTorch] Add bulk_allocate utility and use it in quantized tensor allocators Introduces transformer_engine/pytorch/csrc/extensions/allocate.cpp with a general-purpose bulk_allocate function: given parallel lists of shapes, dtypes, and per-tensor byte alignments, it computes a packed layout, does a single CUDA allocation, and returns at::from_blob views whose deleters keep the backing buffer alive. The three internal bulk_allocate_*_tensors helpers in cast.cpp are refactored to call bulk_allocate instead of each owning a copy of the make_torch_view lambda and the offset-computation loops (~120 lines removed). The new function is also exposed via pybind11 so Python can allocate packed CUDA buffers directly without going through a quantizer. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Tim Moon <tmoon@nvidia.com> * Bulk-allocate wgrads in grouped linear impls Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Apply review suggestions Make optional args for device and alignment. Handle case where base data_ptr is unaligned. Align grouped linear wgrad buffers to 256B. Signed-off-by: Tim Moon <tmoon@nvidia.com> * Nits from Claude Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix incorrect call to `bulk_allocate` Signed-off-by: Tim Moon <tmoon@nvidia.com> * Fix ambiguous return type Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Use c10::Device consistently Signed-off-by: Tim Moon <tmoon@nvidia.com> --------- Signed-off-by: Tim Moon <tmoon@nvidia.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
…rability when processing the output of Flash Attn forward pass (#2825) fix(CP, FA): when processing the output of Flash Attn forward pass, the conditional logic in the FA version contains a vulnerability, fix it Signed-off-by: zhujian <zhujian.whu.cs@gmail.com>
* Use fast unfused cast mxfp8 kernels by default Signed-off-by: Oleg Goncharov <ogoncharov@nvidia.com> * Removed dead code Signed-off-by: Oleg Goncharov <ogoncharov@nvidia.com> * Use fast kernel for full 32-element chunks only 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 * Fix Signed-off-by: Oleg Goncharov <ogoncharov@nvidia.com> * Fixed grid size overflow 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: Oleg Goncharov <ogoncharov@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Przemyslaw Tredak <ptredak@nvidia.com>
* Build Docs fix Signed-off-by: Varun Thumbe <vthumbe@nvidia.com> * fix doxygen warnings Signed-off-by: Varun Thumbe <vthumbe@nvidia.com> * class doc fix Signed-off-by: Varun Thumbe <vthumbe@nvidia.com> --------- Signed-off-by: Varun Thumbe <vthumbe@nvidia.com>
#2979) add wait per multi-proc test cleanup Signed-off-by: Phuong Nguyen <phuonguyen@nvidia.com>
* Avoid CPU offload wait_event for validation Signed-off-by: Vasudevan Rengasamy <vrengasamy@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Vasudevan Rengasamy <vrengasamy@nvidia.com> --------- Signed-off-by: Vasudevan Rengasamy <vrengasamy@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
7 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )