Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Update adaptive tests #81

Open
wants to merge 13 commits into
base: master
Choose a base branch
from

Conversation

rihardsk
Copy link

Created a Python script to automatically generate the expected outputs for self-adaptive Marian and updated said outputs.

@snukky
Copy link
Member

snukky commented Jan 26, 2022

Can you provide information how did you compile Marian (CMake command and GCC and CUDA versions) and what GPU devices you run the tests on when generating the test outputs?

@rihardsk
Copy link
Author

The CMake command is cmake .. -DCMAKE_BUILD_TYPE=Debug -DCOMPILE_ADAPTIVE=ON -DCOMPILE_SERVER=ON (i'm not entirely sure whether it was a debug build, though), but i don't remember the GPU device or software versions i was using. I could just re-generate the outputs, i guess.

CMake command: cmake .. -DCMAKE_BUILD_TYPE=Debug -DCOMPILE_ADAPTIVE=ON -DCOMPILE_SERVER=ON
GCC version: 7.5.0
CUDA version: 10.1
GPU model: Quadro RTX 6000 (Turing)
@rihardsk rihardsk force-pushed the update-adaptive-tests branch from c75acf7 to bc7efdc Compare February 1, 2022 09:06
@rihardsk
Copy link
Author

rihardsk commented Feb 1, 2022

@snukky, i regenerated the outputs. This time i used:

  • CMake command: cmake .. -DCMAKE_BUILD_TYPE=Debug -DCOMPILE_ADAPTIVE=ON -DCOMPILE_SERVER=ON
  • GCC version: 7.5.0
  • CUDA version: 10.1
  • GPU model: Quadro RTX 6000 (Turing)

The -DCOMPILE_ADAPTIVE=ON -DCOMPILE_SERVER=ON flags shouldn't be needed for the expected output generation (with regenerate-expected-outputs.sh). They're only needed to compiling self-adaptive marian to run the tests.

# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants