-
Notifications
You must be signed in to change notification settings - Fork 889
Build fixes #836
Build fixes #836
Conversation
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add back the ts_scripts/README file back from earlier commit. Other comments inline. Works for GPU cu92 and cu101.
CU9.2, DLAMI 22, Ubuntu 16.04, ami-0f009bf0d2e29c5a9:
After running install_dependencies.py, python is downgraded to 3.5
After symlinking it back to 3.6 scripts succeed on 16.04
# UBUNTU 16.04: python3 points to 3.5 which doesn’t seem to have a torch==1.6.0+cu92 release. Had to upgrade to python 3.6
> sudo add-apt-repository ppa:deadsnakes/ppa
> sudo apt-get update
> sudo apt-get install python3.6 python3.6-dev
> curl https://bootstrap.pypa.io/get-pip.py | sudo python3.6
> sudo ln -sf /usr/bin/python3.6 /usr/bin/python
> python -m pip install --upgrade pip
> export PATH=$PATH:/home/ubuntu/.local/bin
> git clone https://github.com/pytorch/serve
> cd serve
> git checkout build_fixes
> pip install -U -r requirements_cu92.txt -f https://download.pytorch.org/whl/torch_stable.html
> python ./ts_scripts/install_dependencies.py
> sudo ln -sf /usr/bin/python3.6 /usr/bin/python
> python ./torchserve_sanity.py
> python ./test/regression_tests.py
docker/README.md
Outdated
| ``` | ||
|
|
||
| For creating GPU based image with older CUDA versions (ex. CUDA 10.1), make sure that the `--build-arg CUDA_VERSION=<version>` is specified. The version is in the format "cuda92", "cuda101": | ||
| For creating GPU based image with older CUDA versions (ex. CUDA 10.1), make sure that the `--build-arg CUDA_VERSION=<version>` is specified. The version is in the format "cu92", "cu101", "cu102": |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we also clarify that we need to specify the appropriate BASE_IMAGE= with the cuda version
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@harshbafna @maaquib Is it not possible to derive the Base image name from just the Cuda version? It will simplify things for people building their own custom image of TorchServe if they can just set a single CUDA_VERSION variable
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@chauhang: Enhanced the build_image.sh script and updated the documentation. Could you please review?
@maaquib: Is the readme really required? I felt it adds more confusion and not really required. We already have which scripts ' references in the required sections of the documentation. |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
|
Retested this with a fresh DLAMI v38 on g4dn.xlarge on this branch. Running into two issues:
I'm continuing to test the different configurations and will report issues as I encounter them in the interest of getting them reported quickly. The newman test failure message is also observable in a fresh Ubuntu 18.04 CPU instance. |
|
|
||
| ```bash | ||
| ./build_image.sh | ||
| ./build_image.sh -bt dev -t torchserve-dev:1.0 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please replace 1.0 with current latest version of another tag like "dev-latest"
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
|
Thanks for the pointers @maaquib. I reran the tests after using the command pr836_ubuntu1804_cpu_regression_test_2.txt |
|
|
Reverted failing newman tests in #845 |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
|
Verified working after #845 |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
Description
Please include a summary of the feature or issue being fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.
Fixes #(issue)
Type of change
Please delete options that are not relevant.
Feature/Issue validation/testing
Please describe the tests [UT/IT] that you ran to verify your changes and relevent result summary. Provide instructions so it can be reproduced.
Please also list any relevant details for your test configuration.
Test A
Test B
UT/IT execution results
Logs
Checklist: