Change scale_cuda to scale_npp #2851
Conversation
leszko
force-pushed
the
rafal/scale-npp
branch
from
250197a
to
d9eb1cc
Compare
leszko
force-pushed
the
rafal/scale-npp
branch
from
d9eb1cc
to
d725d0e
Compare
| # Base CUDA Develop image which contains CUDA SDK libs for the following architectures: linux amd64, linux arm64, windows amd64 | ||
| # | ||
| # To build this image you need the following steps: | ||
| # 1. Download NVIDIA CUDA SDK for ARM64, extract, and copy into cuda/arm64/usr/local/cuda/ |
There was a problem hiding this comment.
you can use the ADD directive in Dockerfile. https://stackoverflow.com/a/52069435/1190388
There was a problem hiding this comment.
Ohhh nice. I didn't know about it! I will leave it as it is right know and if at some point we decide to automate building cuda image, then I'll use this ADD + do cleanup of not needed files.
| COPY cuda/arm64/usr/local/cuda/lib64/libnp* /usr/local/cuda_arm64/lib64/ | ||
| COPY cuda/arm64/usr/local/cuda/include /usr/local/cuda_arm64/include | ||
|
|
||
| RUN ln -s /usr/local/cuda /usr/local/cuda_amd64 |
There was a problem hiding this comment.
should we be building/pushing this regularly as well? or does it need to be manual? (could add a comment in the Dockerfile if it needs to be manual always
There was a problem hiding this comment.
I'd keep this step manual because :
- We never change the cuda version, so we don't need to rebuild this image
- It'd be relatively difficult to automate, because of exe files
Codecov Report
@@ Coverage Diff @@
## master #2851 +/- ##
===================================================
- Coverage 56.49462% 56.48581% -0.00881%
===================================================
Files 88 88
Lines 19239 19242 +3
===================================================
Hits 10869 10869
- Misses 7771 7774 +3
Partials 599 599
Continue to review full report in Codecov by Sentry.
|
leszko
force-pushed
the
rafal/scale-npp
branch
2 times, most recently
from
733bdfa
to
80f0fe9
Compare
Build seprate binaries for cpu and gpu Build seprate binaries for cpu and gpu Build seprate binaries for cpu and gpu Test build only cpu-linux-arm Test build only cpu-linux-arm Test build only cpu-linux-arm Test build only cpu-linux-arm Test build only cpu-linux-arm Test build only cpu-linux-arm Build seprate binaries for cpu and gpu Revert "Test build only cpu-linux-arm" This reverts commit 99b3acf. Set seperate build environment for CPU and GPU Set seperate build environment for CPU and GPU
leszko
force-pushed
the
rafal/scale-npp
branch
from
80f0fe9
to
b4236fc
Compare
Change ffmpeg installation to use CUDA SDK (needed to use the
scale_nppfilter).It also removes tensorflow build, I believe we don't need it anymore.
Related work: