27 Pull requests merged by 13 people

• [ROCm] Cherry-pick heterogeneous vectorized elementwise kernel optimizations from upstream

  #2163 merged May 18, 2025

• [release/2.6] Make triton build ROCm version agnostic

  #2130 merged May 17, 2025

• [ROCm] Cherry-pick heterogeneous vectorized elementwise kernel optimizations from upstream

  #2161 merged May 16, 2025

• [release/2.6] Change gfx110x BLAS preferred backend (Take 2)

  #2158 merged May 16, 2025

• [release/2.5][ROCm] update state check for test_trace_while_active*

  #2110 merged May 16, 2025

• [AUTOGENERATED] [release/2.3] Remove --no-index and --no-deps flags

  #2142 merged May 15, 2025

• [AUTOGENERATED] [release/2.4] Remove --no-index and --no-deps flags

  #2143 merged May 15, 2025

• Updated default workspace size for gfx95

  #2098 merged May 14, 2025

• [release/2.7] Prefer hipblaslt for gfx1200, gfx1201

  #2125 merged May 14, 2025

• [release/2.6] Use -ge for numeric comparison

  #2151 merged May 14, 2025

• [ROCm] Improve in-place vectorized elemwise kernel performance on MI300

  #2111 merged May 14, 2025

• Fix fp8 inductor tests

  #2107 merged May 14, 2025

• [release/2.6] Update ROCm version check for triton wheel build

  #2149 merged May 14, 2025

• [rocm6.5_internal_testing] Update related_commits

  #2131 merged May 14, 2025

• [release/2.7] Update related_commits

  #2132 merged May 14, 2025

• [release/2.5] Update related_commits

  #2134 merged May 14, 2025

• [AUTOGENERATED] [rocm6.4_internal_testing] [release/2.5] test_decompose_mem_bound_mm.py tolerance increase for navi3x

  #2146 merged May 14, 2025

• [AUTOGENERATED] [release/2.6] [release/2.5] test_decompose_mem_bound_mm.py tolerance increase for navi3x

  #2144 merged May 14, 2025

• [AUTOGENERATED] [release/2.5] Remove --no-index and --no-deps flags

  #2124 merged May 13, 2025

• [AUTOGENERATED] [release/2.7] Remove --no-index and --no-deps flags

  #2123 merged May 13, 2025

• [AUTOGENERATED] [rocm6.4_internal_testing] Remove --no-index and --no-deps flags

  #2122 merged May 13, 2025

• [AUTOGENERATED] [rocm6.5_internal_testing] Remove --no-index and --no-deps flags

  #2121 merged May 13, 2025

• [release/2.5] test_decompose_mem_bound_mm.py tolerance increase for navi3x

  #2129 merged May 13, 2025

• [release/2.6] Change gfx110x BLAS preferred backend

  #2053 merged May 13, 2025

• Use BUILD_BYPRODUCTS instead of INSTALL_BYPRODUCTS in ExternalProject_Add

  #2109 merged May 12, 2025

• [rocm6.4_internal_testing] Replaced ROCm specific skips to generalized conditions

  #2100 merged May 12, 2025

• Test PR to test creation of comment by Pytorch CI

  #1630 merged May 12, 2025

14 Pull requests opened by 5 people

• [AUTOGENERATED] [release/2.7] [rocm6.4_internal_testing] Replaced ROCm specific skips to generalized conditions

  #2126 opened May 13, 2025

• [AUTOGENERATED] [rocm6.5_internal_testing] [rocm6.4_internal_testing] Replaced ROCm specific skips to generalized conditions

  #2127 opened May 13, 2025

• [AUTOGENERATED] [release/2.7] [release/2.6] Change gfx110x BLAS preferred backend

  #2128 opened May 13, 2025

• [release/2.6] Update related_commits

  #2133 opened May 13, 2025

• [AUTOGENERATED] [release/2.7] [release/2.5] test_decompose_mem_bound_mm.py tolerance increase for navi3x

  #2145 opened May 14, 2025

• [AUTOGENERATED] [rocm6.5_internal_testing] [release/2.5] test_decompose_mem_bound_mm.py tolerance increase for navi3x

  #2147 opened May 14, 2025

• Enable rocSolver for batch eigen decomposition on ROCm

  #2148 opened May 14, 2025

• [AUTOGENERATED] [rocm6.5_internal_testing] [release/2.7] Prefer hipblaslt for gfx1200, gfx1201

  #2152 opened May 14, 2025

• [rocm6.5_internal_testing] Fix tf32 tests

  #2155 opened May 15, 2025

• [release/2.6] Fix tf32 tests

  #2156 opened May 15, 2025

• [AUTOGENERATED] [release/2.7] [release/2.6] Change gfx110x BLAS preferred backend (Take 2)

  #2159 opened May 16, 2025

• [rocm6.4_internal_testing] kineto: add missing <string> include to RocLogger.h

  #2160 opened May 16, 2025

• [AUTOGENERATED] [release/2.5] Make triton build ROCm version agnostic

  #2164 opened May 17, 2025

• [AUTOGENERATED] [release/2.7] Make triton build ROCm version agnostic

  #2165 opened May 17, 2025

12 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [In Progress] Enable mx f8 fp4 support on ROCm

  #2046 commented on May 16, 2025 • 1 new comment

• dockerhub images appear to be build images instead of runtime

  #1965 commented on May 14, 2025 • 0 new comments

• [release/2.5] Enabled force_shape_pad for test_pad_mm and test_slice_mm_bandwidth_computation

  #1755 commented on May 15, 2025 • 0 new comments

• [release/2.5] Fixed string comparison in test_cpp_wrapper_hipify

  #1760 commented on May 15, 2025 • 0 new comments

• [release/2.5] AMDSMI/layernorm cherry picks

  #1764 commented on May 15, 2025 • 0 new comments

• [AUTOGENERATED] [release/2.5] Cherry-pick PR-1767

  #1794 commented on May 15, 2025 • 0 new comments

• [Do NOT MERGE] [release/2.5] Enable tf32 testing on test_nn

  #1859 commented on May 15, 2025 • 0 new comments

• Diable L2_cahce_size display on ROCm platforms

  #1876 commented on May 15, 2025 • 0 new comments

• [release/2.5] enable NHWC batchnorm with MIOpen

  #1982 commented on May 15, 2025 • 0 new comments

• Schedule remainder loop chunk in threadblock 0.

  #1985 commented on May 15, 2025 • 0 new comments

• [Release 2.6] Triton/inductor related optimisations

  #2008 commented on May 15, 2025 • 0 new comments

• [rocm6.5_internal_testing] fix sparse tests, enable fp16/bf16 for testing

  #2108 commented on May 15, 2025 • 0 new comments