Skip to content

Pull requests: intel/torch-xpu-ops

New pull request New
163 Open 1,586 Closed
163 Open 1,586 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Enable dpclang sycl compiler
#3287 opened Apr 8, 2026 by dvrogozh Loading…
Remove stale ConvTranspose3d complex half parity xfail
#3285 opened Apr 8, 2026 by gplutop7 Loading…
3
Override test case test_big_num_tensors to change torch.cuda to torch.xpu. disable_accelerate Disable accelerate test job in PR CI testing disable_distributed Disable distributed UT test jobs for the PR disable_e2e Disable all e2e test jobs for the PR disable_transformers Disable transformers UT test in PR CI disable_win Disable Windows CI test jobs for the PR
#3283 opened Apr 8, 2026 by BBBela Loading…
2
Add contiguous gather fast path bypassing OffsetCalculator
#3282 opened Apr 8, 2026 by yucai-intel Loading…
8
[DO NOT MERGE][TEST ONLY] Create AGENTS.md disable_all Disable all ci test jobs for the PR, just keep basic lint check
#3281 opened Apr 8, 2026 by Stonepia • Draft
91
Optimize scatter kernel launch with compile-time constants
#3279 opened Apr 8, 2026 by yucai-intel Loading…
6
Use C++20 for SYCL device compilation
#3278 opened Apr 7, 2026 by r-barnes Loading…
2
Add early exit checks in torch.binomial op
#3276 opened Apr 7, 2026 by Silv3S Loading…
1
4
Use std::fma in addcmul and foreach pointwise ops for FMA parity with CUDA
#3275 opened Apr 7, 2026 by AKloniecki Loading…
3 tasks done
1
9
[CI] Seperate Windows CI disable_distributed Disable distributed UT test jobs for the PR disable_e2e Disable all e2e test jobs for the PR disable_ut Disable all non-distributed UT test jobs for the PR
#3274 opened Apr 7, 2026 by RUIJIEZHONG66166 Loading…
7
Fix bf16/fp16 accuracy issue in sparsecsr addmm
#3273 opened Apr 6, 2026 by jenniew Loading…
3
[CI] Fix permission issue for workflows disable_win Disable Windows CI test jobs for the PR
#3271 opened Apr 6, 2026 by chuanqi129 Loading…
6
Refactor sycl::exp call sites for readability in SoftMax and LossCTC kernels
#3269 opened Apr 6, 2026 by jianyizh Loading…
11
Migrate trigonometric and related math functions from std:: to sycl:: namespace
#3268 opened Apr 6, 2026 by jianyizh Loading…
2
Decompose dot_xpu_mkl into mul and sum for long dtype
#3265 opened Apr 3, 2026 by Silv3S Loading…
1
5
Add native XPU SVD implementation using oneMKL gesvd
#3264 opened Apr 3, 2026 by PatrykWilczewski Loading…
1
21
Improve norm kernel performance for 16-bit types on XPU
#3263 opened Apr 3, 2026 by yucai-intel Loading…
5
Add triangular solve function for sparse CSR tensors in XPU
#3261 opened Apr 2, 2026 by tszulist-hbn Loading…
1
8
Skip TypedStorage tests on XPU
#3260 opened Apr 2, 2026 by BartoszKokoszko Loading…
12
Fix generic address space warnings in GroupRadixSort and SegmentedGro…
#3256 opened Apr 2, 2026 by chunhuanMeng Loading…
7
Fix generic address space warnings in NMSKernelFunctor
#3254 opened Apr 2, 2026 by chunhuanMeng Loading…
5
Fix illegal tensor in dbeta of RMSNorm Backward
#3253 opened Apr 1, 2026 by Wetitpig Loading…
Introduce op_ut_slow job for very slow tests. disable_accelerate Disable accelerate test job in PR CI testing disable_distributed Disable distributed UT test jobs for the PR disable_e2e Disable all e2e test jobs for the PR disable_transformers Disable transformers UT test in PR CI
#3252 opened Apr 1, 2026 by BBBela Loading…
19
Removed the BUILD_SEPARATE_OPS build option
#3245 opened Apr 1, 2026 by chunhuanMeng Loading…
2
[windowsCI] Remove hardcoded MKL pin from CI requirements to avoid conflicts with Torch XPU wheel
#3241 opened Apr 1, 2026 by jafraustro Loading…
@jafraustro
11
ProTip! Mix and match filters to narrow down what you’re looking for.