search

chainer / onnx-chainer

Add-on package for ONNX format support in Chainer
MIT License
85 stars 24 forks source link

Build computational graph manually and not use backward with FunctionHook #162

Closed disktnk closed 5 years ago

disktnk commented 5 years ago

~WIP~

Currently ONNX-Chainer uses FunctionHook's backward hook to make computational graph, this cause a bit difficulty to investigate errors and implicit graph construction. This PR will introduce manually making a computational graph. Along with moving convert logic from hook to making graph, this PR also provides clearer code on exporting.

codecov-io commented 5 years ago

Codecov Report

Merging #162 into master will increase coverage by 0.3%. The diff coverage is 95.83%.

Impacted file tree graph

@@            Coverage Diff            @@
##           master     #162     +/-   ##
=========================================
+ Coverage   86.37%   86.67%   +0.3%     
=========================================
  Files          23       24      +1     
  Lines        1218     1276     +58     
=========================================
+ Hits         1052     1106     +54     
- Misses        166      170      +4
Impacted Files Coverage Δ
onnx_chainer/export.py 91.32% <100%> (-0.89%) :arrow_down:
onnx_chainer/graph.py 95.68% <95.68%> (ø)
onnx_chainer/replace_func.py 88.5% <0%> (-3.45%) :arrow_down:
onnx_chainer/functions/__init__.py 100% <0%> (ø) :arrow_up:
onnx_chainer/mapping.py 90% <0%> (ø) :arrow_up:
onnx_chainer/functions/array.py 85.88% <0%> (+1.74%) :arrow_up:

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 235fb96...e7c2e4e. Read the comment docs.

disktnk commented 5 years ago

ChainerCI, test this please

pfn-ci-bot commented 5 years ago

Successfully created a job for commit e7c2e4e: