refactor: rebase to deepgemm 2.0.0+#1
Open
KyeeHuang wants to merge 32 commits into
Open
Conversation
Co-authored-by: Shixian Cui <shixian@amazon.com>
…ze setting. (deepseek-ai#115)" This reverts commit ac428e2. This PR causes wgrad to hang during testing. Revert it until we resolve the issue
* Add more GPU architectures support * Update layout.py * Optimize performance, Add SM90 support, Add 1D2D SM100 support * Add fmtlib submodule at commit 553ec11 --------- Co-authored-by: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
Anonymous cloning is more difficult with git@ (it requires per user SSH keys), which makes CI automation more difficult. This commit reverts to using https:// as it was before deepseek-ai#112
…ecursive_directory_iterator to avoid an ABI breakage we met (deepseek-ai#131)
* Add torch as dependency. Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com> * Add pyproject.toml Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com> * Fix setup.py. Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com> * Add build dependency. Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com> --------- Signed-off-by: Yuxian Qiu <142763828+yuxianq@users.noreply.github.com>
Fix import-time error: "undefined symbol: _ZN3c104cuda20getCurrentCUDAStreamEa"
* Fix assertion error in smxx_layout.hpp for mn % 4 != 0 cases * Fix assertion error in smxx_layout.hpp for mn % 4 != 0 cases * Align submodule files * Fix assertion error in smxx_layout.hpp for mn % 4 != 0 cases * fix(smxx_layout): support mn%4!=0 and num_groups>1 via torch * fix(smxx_layout): support mn%4!=0 and num_groups>1 via torch * fix: correct logic for entering get_mn_major_tma_aligned_packed_ue8m0_tensor_torch
* Fix sm90 GEMM * Fix typo --------- Co-authored-by: Kuai Yu <yukuai@deepseek.com>
- Add support for legacy CUDA versions; now compatible with CUDA 12.3 and newer - Add support for NVRTC compilation - Other fixes and code refactoring
Make various updates and fixes
- Add BF16 support for SM90 and SM100 - Refactor Python APIs - Other fixes and code refactoring
Signed-off-by: Xiwen Yu <13230610+VALLIS-NERIA@users.noreply.github.com>
- Replace non-existent test_core.py with test_bf16.py - Add test_fp8.py and test_lazy_init.py to testing section
KyeeHuang
changed the title
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
16 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.
I recently rebased the original per-tensor solution to version 2.0.0+, and it can pass the tests in test_fp8.py.
I only rebased the implementation for the Hopper architecture. I suspect that the per-tensor approach may not yield additional performance benefits on the Blackwell architecture.