You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by pt...@apache.org on 2020/09/15 23:01:01 UTC
[incubator-mxnet] branch v1.x updated (9dfac79 -> 606933f)
This is an automated email from the ASF dual-hosted git repository.
ptrendx pushed a change to branch v1.x
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git.
from 9dfac79 [1.x] Backport Fix for duplicate subgraph inputs/outputs (#16131) (#19112)
add 606933f TensorRT: add int8 with calibration (#19011)
No new revisions were added by this update.
Summary of changes:
ci/docker/runtime_functions.sh | 16 ++
ci/jenkins/Jenkins_steps.groovy | 18 ++
ci/jenkins/Jenkinsfile_unix_gpu | 1 +
python/mxnet/contrib/__init__.py | 1 -
python/mxnet/contrib/tensorrt.py | 67 -------
src/operator/subgraph/tensorrt/onnx_to_tensorrt.cc | 82 ++++++++-
src/operator/subgraph/tensorrt/onnx_to_tensorrt.h | 14 +-
src/operator/subgraph/tensorrt/tensorrt-inl.h | 121 ++++++++++--
src/operator/subgraph/tensorrt/tensorrt.cc | 45 ++++-
src/operator/subgraph/tensorrt/tensorrt.cu | 27 ++-
.../subgraph/tensorrt/tensorrt_int8_calibrator.cc | 149 +++++++++++++++
.../subgraph/tensorrt/tensorrt_int8_calibrator.h | 105 +++++++++++
{tools => tests/python/tensorrt}/rec2idx.py | 7 +-
tests/python/tensorrt/test_tensorrt.py | 202 +++++++++++++++++++++
14 files changed, 751 insertions(+), 104 deletions(-)
delete mode 100644 python/mxnet/contrib/tensorrt.py
create mode 100644 src/operator/subgraph/tensorrt/tensorrt_int8_calibrator.cc
create mode 100644 src/operator/subgraph/tensorrt/tensorrt_int8_calibrator.h
copy {tools => tests/python/tensorrt}/rec2idx.py (98%)
create mode 100644 tests/python/tensorrt/test_tensorrt.py