| 1 | #183633 | [ROCm] Remove redundant ROCm scale-mode branch in CUDABlas | Author |
| 2 | #182195 | [ROCm] Add CK BLAS backend torch.mm test | Author |
| 3 | #180731 | [ROCm] Run test_scaled_mm_deepseek_error_messages on mi350 arch | Author |
| 4 | #180518 | [ROCm] Fix evaluate_platform_supports_fp8 false-positive | Author |
| 5 | #180384 | [ROCm] Update scaled_mm DeepSeek error message | Author |
| 6 | #179782 | [ROCm] Enable ROCm swizzle check and update scaled_mm swizzle tests | Author |
| 7 | #178570 | test_scaled_matmul_cuda: Parameterize float8 basics cases | Author |
| 8 | #176924 | Skip 4-GPU distributed tests on 2-GPU runners | Author |
| 9 | #173737 | [ROCm] Enable scaled group mm on gfx950 | Author |
| 10 | #173502 | [ROCm] group_gemm perf: Add new kernel for all K equal cases | Author |
| 11 | #172988 | CUDAScaledBlas - replace FBGEMM_GENAI with MSLK | Author |
| 12 | #172791 | [ROCm] Unifying hipBLASLt architecture lists into common hook methods | Author |
| 13 | #172311 | [ROCm] Use HIPCachingAllocator for CK argument and workspace buffers | Author |
| 14 | #171901 | [ROCm] Add unit test to verify grouped GEMM CK opt-in flag | Author |
| 15 | #171140 | [ROCm] Make grouped GEMM CK opt-in via env and default to fallback path | Author |
| 16 | #171121 | [ROCm] Refactor ROCm CK config generation into shared helper | Author |
| 17 | #170254 | [ROCm] inductor/fp8 test: Check for "cuda" in device type | Author |
| 18 | #169356 | [ROCm] Enable group gemm on gfx90a | Author |
| 19 | #168220 | [ROCm] Fix HIP document url | Author |
| 20 | #167403 | [ROCm] Disable group gemm CK path when CK is not enabled | Author |
| 21 | #167390 | [ROCm][CI] trigger magma build with gfx950 for ROCm7.1 | Author |
| 22 | #167299 | [ROCm][CI] unconditionally add gfx950, gfx115x to PYTORCH_ROCM_ARCH | Author |
| 23 | #167066 | [ROCm] Skip nvfp4 tests on ROCm | Author |
| 24 | #166478 | [ROCm][CI] remove relaxed tolerance for tf32 tests | Co-author |
| 25 | #166334 | [ROCm] Enable group gemm through CK | Author |
| 26 | #165968 | [ROCm] Skip test_blockwise_nvfp4_with_global_scale | Author |
| 27 | #165788 | test_scaled_matmul_cuda: fix infer_scale_swizzle | Author |
| 28 | #165528 | [ROCm] Add scaled_mm v2 support | Author |
| 29 | #164320 | MX: Remove redundant PLATFORM_SUPPORTS_MX_GEMM constant | Author |
| 30 | #163127 | [ROCm] Fix mx fp8 and fp4 code after scaling refactor changes | Author |
| 31 | #163050 | [ROCm] test_aot_inductor: Enable fp8 tests | Author |
| 32 | #162787 | [ROCm] Enable test_fixed_striding | Author |
| 33 | #162600 | [ROCm] rocblas Aten GEMM overload for FP32 output from FP16/BF16 inputs | Author |
| 34 | #161540 | ROCm: Enable overload tests from test_matmul_cuda | Author |
| 35 | #161166 | forward fix of #152198 | Author |
| 36 | #161009 | test_matmul_cuda: Refine MX test skipping | Author |
| 37 | #151360 | AMD/ROCm OCP Micro-scaling Format (mx-fp8/mx-fp4) Support | Co-author |
| 38 | #150250 | ROCm: Add trailing comma for consistency in gfx architecture list | Author |
| 39 | #148945 | ROCm: Enable tf32 testing on test_nn | Author |
| 40 | #148228 | ROCm: Disable torch check for Multiplication of two Float8_e5m2 matrices | Author |
| 41 | #147186 | ROCm: Remove static specifier for allow_tf32 variable | Author |
| 42 | #144692 | ROCm: Skip tests in elastic/utils/distributed_test | Author |
| 43 | #140319 | ROCm: Enable 4 gpu tests for distributed config | Author |
| 44 | #137724 | [ROCm] Enable dist sharded_tensor test suites | Author |
| 45 | #92932 | [ROCm] Fix distributed tests failure and enable ROCm distributed CI | Author |
| 46 | #85760 | [ROCm] [FakeTensorTest] Enable test_fallback_memory_prop | Author |
| 47 | #84100 | [ROCm] Enable test_filtering_env_var | Author |
| 48 | #84077 | [ROCm] add case for FP32MatMulPattern skip property | Author |
| 49 | #82749 | [ROCm] Enable test_ddp_profiling_torch_profiler | Author |
| 50 | #82356 | [ROCm] Enable test_multiprocessing tests | Author |
| 51 | #82005 | [ROCm] Enable test_grad_layout_1devicemodule_1replicaperprocess | Author |
| 52 | #80369 | [test_shape_ops] Increase system memory requirement | Author |
| 53 | #78633 | [ROCm] Enable test_instantiator, test_type_hints | Author |
| 54 | #78401 | [ROCm] TestGradients: Enable grad and gradgrad | Author |
| 55 | #78385 | [ROCm] enable test_lobpcg_ortho_cuda_float64 | Author |
| 56 | #72923 | [LTS] CherryPick: Add multi gpu checker for TestZeroRedundancyOptimizer | Co-author |
| 57 | #68487 | [ROCm] Use hipCUB/rocPRIM scan algorithms for large index support | Author |
| 58 | #65203 | [ROCm] update Magma for 4.3 release | Author |
| 59 | #62502 | [ROCm] update magma | Author |
| 60 | #61313 | [ROCm] Skip test_masked_scatter_large_tensor_cuda | Author |
| 61 | #59339 | [Release-1.9] Disable failing ROCM-4.2 tests | Co-author |
| 62 | #59158 | [ROCm] disable test_Conv2d_groups_nobias for ROCm | Author |
| 63 | #57159 | [Distributed test] Enable ddp_control_flow tests for ROCm | Author |
| 64 | #54714 | [test_c10d] Add wait in nccl high priority stream test | Author |
| 65 | #53850 | [ROCm] Fix the flaky test test_stream_event_nogil | Author |
| 66 | #53564 | [TestZeroRedundancyOptimizer] Add multi gpu checker | Author |
| 67 | #53192 | [ROCm] Disable flaky tests on ROCm | Author |
| 68 | #51510 | [ROCm] disable tests for ROCm 4.0.1 | Author |
| 69 | #50629 | [distributed_test_c10d] Enable disabled ROCm tests | Author |
| 70 | #50421 | [distributed_test] Enable disabled ROCm tests | Author |
| 71 | #49069 | [distributed] Provide parameter to pass GPU ID in barrier function | Author |
| 72 | #48023 | [ROCm] Enable skipped distributed global tests | Author |
| 73 | #47898 | distributed_test: Map rank to GPU accordingly | Author |