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 cc0f27a8b1 [Relay] Remove overwriting of matmul shapes when they are
static (#13615)
add 06be0b3b26 [Frontend] [ONNX] Support sequence_lens of GRU (#13587)
add c5911a6f23 [ETHOSN] Add support for experimental compiler option
(#13410)
add ce97138ebe [TVMScript] Fix print round-tripable multi thread env
binding (#13622)
add cdb4eea138 [TOPI][Hexagon] Implement global_avg_pool2d for hexagon
(#13614)
add 7674ea84fe Add check for non-contiguous memory access when lowering to
async dma… (#13613)
add 37f6aa0c7e [MetaSchedule] Fix tensorcore winograd task extraction
(#13625)
No new revisions were added by this update.
Summary of changes:
python/tvm/relay/frontend/common.py | 57 +++++-
python/tvm/relay/frontend/onnx.py | 18 +-
python/tvm/relay/op/strategy/cuda.py | 2 +
python/tvm/topi/hexagon/qnn/__init__.py | 1 +
python/tvm/topi/hexagon/qnn/global_avg_pool2d.py | 95 ++++++++++
python/tvm/topi/hexagon/slice_ops/__init__.py | 1 +
.../topi/hexagon/slice_ops/global_avg_pool2d.py | 52 ++++++
python/tvm/topi/hexagon/utils.py | 12 ++
src/printer/tvmscript_printer.cc | 1 -
src/relay/backend/contrib/ethosn/codegen.cc | 10 +-
src/relay/backend/contrib/ethosn/codegen_ethosn.h | 4 +
src/tir/transforms/lower_async_dma.cc | 28 +++
tests/python/contrib/test_ethosn/test_codegen.py | 54 ++++++
.../python/contrib/test_hexagon/infrastructure.py | 13 ++
.../test_hexagon/test_async_dma_pipeline.py | 206 +++++++++++++++++++++
.../test_global_avg_pool2d.py} | 104 +++++------
tests/python/frontend/onnx/test_forward.py | 40 +++-
.../test_meta_schedule_relay_integration.py | 18 ++
tests/python/unittest/test_tvmscript_roundtrip.py | 14 ++
19 files changed, 656 insertions(+), 74 deletions(-)
create mode 100755 python/tvm/topi/hexagon/qnn/global_avg_pool2d.py
create mode 100755 python/tvm/topi/hexagon/slice_ops/global_avg_pool2d.py
copy tests/python/contrib/test_hexagon/topi/{test_adaptive_avg_pool1d.py =>
slice_op/test_global_avg_pool2d.py} (60%)
