This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a change to branch nightly
in repository https://gitbox.apache.org/repos/asf/tvm.git
from cca84d3082 [Bug][CodeGen,Cuda]fix cast fp16 to int8/uint8 in cuda
(#13641)
add e5a7f5fb5f [TOPI] Expose mem_scope from generic conv2d variants to be
more reusable (#13680)
add b6851f344e [CLML] Version compatibility and various test cases
(#13670)
add 6dbb7e1a5e [BENCHMARKS][ADRENO] Documentation for Adreno (Texture)
benchmarks (#13679)
add e24d4fb78b [ONNX] Add converter for QAttention from Microsoft
onnxruntime contrib opset (#13654)
add 45a8a44b86 [BugFix][Runtime] Add missing check for `PackedFunc`
(#13687)
add 17f88eaf78 [COMMUNITY] @blackkker -> Reviewer (#13686)
add d8357a08c0 [Git] Ignore python/requirements directory (#13684)
add 49ed54478b [Schedule][Bugfix] Fix decompose padding wrt the single
child subtree (#13646)
No new revisions were added by this update.
Summary of changes:
.gitignore | 3 +
CONTRIBUTORS.md | 1 +
apps/benchmark/README.md | 16 +-
cmake/modules/contrib/CLML.cmake | 16 +-
include/tvm/runtime/packed_func.h | 5 +
python/tvm/relay/frontend/onnx.py | 297 ++++++++++++++++
python/tvm/relay/op/contrib/clml.py | 58 ++--
python/tvm/topi/generic/conv2d.py | 6 +-
src/relay/backend/contrib/clml/codegen.cc | 2 +-
src/runtime/contrib/clml/clml_runtime.cc | 38 ++-
src/tir/schedule/primitive/decompose_padding.cc | 17 +-
tests/cpp/packed_func_test.cc | 12 +
tests/python/contrib/test_clml/infrastructure.py | 58 +++-
tests/python/contrib/test_clml/test_network.py | 15 +-
tests/python/contrib/test_clml/test_ops.py | 377 +++++++++++++++++++--
tests/python/frontend/onnx/test_forward.py | 184 ++++++++++
.../test_tir_schedule_decompose_padding.py | 63 ++++
tests/scripts/task_build_adreno_bins.sh | 6 +-
tests/scripts/task_config_build_adreno.sh | 4 +-
19 files changed, 1077 insertions(+), 101 deletions(-)
