Fix multi-tensor LAMB reduction to be deterministic #6028
suffiank
merged
23 commits into
master
from
sukha/lambdeterminism
define ordering of reduction across blocks
647ac2fd
Merge branch 'master' of github.com:microsoft/onnxruntime into sukha/…
e40ea726
Merge branch 'master' of github.com:microsoft/onnxruntime into sukha/…
b18f3733
save state
8ec12204
remove debug code
118da700
remove debug code
a58b336a
resolve merge conflict
000ee80b
ke1337
commented
on 2020-12-04
wschin
dismissed these changes
on 2020-12-04
wschin
commented
on 2020-12-04
review comments
4ce75ba4
Merge branch 'master' of github.com:microsoft/onnxruntime into sukha/…
eabe1ff4
edgchen1
commented
on 2020-12-09
edgchen1
commented
on 2020-12-09
edgchen1
commented
on 2020-12-09
edgchen1
commented
on 2020-12-09
edgchen1
commented
on 2020-12-09
significant correction for reduction only over blocks on same tensor
0b923e61
addressing ocmments
23f06f08
Merge branch 'master' of github.com:microsoft/onnxruntime into sukha/…
cd8bd92f
update rocm/lamb.cc to build as well
e1cbe7c7
remove times 2048*size in multitensor test until threshold error in r…
33fcf06e
edgchen1
commented
on 2020-12-09
edgchen1
commented
on 2020-12-09
Merge branch 'sukha/lambdeterminism' of github.com:microsoft/onnxrunt…
9683a2f7
convert tuple => struct as per recomendation
b18e1b96
update comment
79de8c6b
apply perfect forwarding for launch_multitensor to permit passing ref…
10fbb674
remove excess template arguments from rocm lamb.cc launch_multitensor…
fdc8952e
fixes for AMD build
e490a96a
wschin
commented
on 2020-12-10
wschin
commented
on 2020-12-10
wschin
commented
on 2020-12-10
wschin
commented
on 2020-12-10
wschin
commented
on 2020-12-10
pr comments
8dd26b12
run formatter from vscode
7bf462fd
formatter on cuda files
b84e9e01
wschin
approved these changes
on 2020-12-11
edgchen1
approved these changes
on 2020-12-11
Login to write a write a comment.
Login via GitHub