1.2. I did a quick twistlock scan of this image, and it continues to look good.
1.3. I used this image to run a quick pytorchjob and it looked good.
1.4 TBD - Still need to test on a real GPU
I built it with the dev target like this:
docker build --target dev --progress=plain -t fms-hf-tuning:realdev_image . -f build/Dockerfile
2.2 I did a quick twistlock scan of this image, and it finds vulnerabilities of the python, which is to be expected. (This is why we have a slimmed down release build, to avoid these vulnerabilities)
2.3 I got inside the image and it has the requested items that @anhuong was asking for:
podman run --rm -it localhost/fms-hf-tuning:realdev_image /bin/bash
and then:
[tuning@28888908315c app]$ which git
/bin/git
[tuning@28888908315c app]$ which dnf
/bin/dnf
[tuning@28888908315c app]$ which yum
/bin/yum
as well as the dev packages:
pip list |grep pack
packaging 24.1
pip list |grep ninja
ninja 1.11.1.1
pip list |grep scikit
scikit-learn 1.5.2
pip list |grep boto
boto3 1.35.55
botocore 1.35.55
### Was the PR tested
In progress
<!-- Describe how PR was tested -->
- [ ] I have added >=1 unit test(s) for every new method I have added.
- [ ] I have ensured all unit tests pass
Description of the change
This adds a dev stage to the Dockerfile which, when used with
--target dev
enables dnf, yum, git and the fms-hf-tuning dev tools.Related issue number
Closes #1147
How to verify the PR
I built it two ways -
First with the entire build that would generate the regular release image:
1.2. I did a quick twistlock scan of this image, and it continues to look good. 1.3. I used this image to run a quick pytorchjob and it looked good. 1.4 TBD - Still need to test on a real GPU
I built it with the dev target like this:
2.2 I did a quick twistlock scan of this image, and it finds vulnerabilities of the python, which is to be expected. (This is why we have a slimmed down release build, to avoid these vulnerabilities) 2.3 I got inside the image and it has the requested items that @anhuong was asking for:
and then:
as well as the dev packages:
pip list |grep ninja ninja 1.11.1.1
pip list |grep scikit scikit-learn 1.5.2
pip list |grep boto boto3 1.35.55 botocore 1.35.55