This is an automated email from the ASF dual-hosted git repository.
masahi pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from b972877 [microNPU] Refactor base address determination to codegen
(#9929)
add ffff8dd Add FP requantize flow. Set float32 flow by default for llvm
x86 targets with (#9637)
No new revisions were added by this update.
Summary of changes:
include/tvm/relay/qnn/attrs.h | 8 +-
.../_quantize.py => qnn/op/_requantize.py} | 4 +-
python/tvm/relay/qnn/op/qnn.py | 102 ++-
python/tvm/topi/x86/utils.py | 22 +
src/relay/qnn/op/requantize.cc | 223 +++++-
src/relay/qnn/op/requantize_config.cc | 93 +++
src/relay/qnn/op/requantize_config.h | 126 ++++
src/relay/qnn/utils.cc | 16 +
src/relay/qnn/utils.h | 14 +-
src/relay/transforms/pattern_utils.h | 25 +
tests/python/relay/test_op_qnn_requantize.py | 750 ++++++++++++---------
11 files changed, 1041 insertions(+), 342 deletions(-)
copy python/tvm/relay/{quantize/_quantize.py => qnn/op/_requantize.py} (89%)
create mode 100644 src/relay/qnn/op/requantize_config.cc
create mode 100644 src/relay/qnn/op/requantize_config.h
