PyTorch 1.3.1 #13

New issue

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

Sign up for GitHub

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

Already on GitHub? Sign in to your account

Jump to bottom

Merged

akartsky merged 14 commits into aws:master from akartsky:master

Nov 20, 2019

Contributor

akartsky commented Oct 31, 2019 •

edited

Loading

Added PyTorch 1.3.1 support.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

Collaborator

sagemaker-bot commented Oct 31, 2019

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

nikhil-sk reviewed

View reviewed changes

docker/1.3.0/py3/Dockerfile.gpu Outdated

+              ARG PYTORCH_VERSION=1.3.0
+              ARG TORCHVISION_VERSION=0.4.1
+              ARG MMS_VERSION=1.0.7

Contributor

nikhil-sk Nov 13, 2019

Please update this to the latest 1.0.8 everywhere.

Contributor Author

akartsky Nov 19, 2019

done

nikhil-sk reviewed

View reviewed changes

docker/1.3.0/py3/Dockerfile.gpu Outdated

+              	 /opt/conda/bin/conda clean -ya
+              ENV PATH /opt/conda/bin:$PATH
+              ARG PYTORCH_VERSION=1.3.0

Contributor

nikhil-sk Nov 13, 2019 •

edited

Loading

Please update to PyTorch 1.3.1 everywhere as discussed, thank you.

Contributor Author

akartsky Nov 19, 2019

done

akartsky force-pushed the master branch from d776371 to 0114b59 Compare

November 13, 2019 21:58


          added pytorch 1.3.0 inference files

e8de4ed

akartsky force-pushed the master branch from 0114b59 to 29def8f Compare

November 13, 2019 22:08

akartsky added 3 commits

November 13, 2019 14:24


          Updated torchvision to 0.4.1 and added ENV path

11cd077


          added changes for pytorch 1.3.1

70bc3e8


          changed MMS version to 1.0.8

c97988f

akartsky force-pushed the master branch from 29def8f to c97988f Compare

November 13, 2019 22:25


          updated torchvision to 0.4.2

ff96c42

Collaborator

sagemaker-bot commented Nov 14, 2019

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

akartsky changed the title ~~PyTorch 1.3.0~~ PyTorch 1.3.1

ghost suggested changes

View reviewed changes

docker/1.3.1/py2/Dockerfile.cpu Outdated

+                  curl \
+                  git \
+                  wget \
+                  vim \

ghost Nov 17, 2019

Please sort apt-get libraries

Contributor Author

akartsky Nov 19, 2019

done

docker/1.3.1/py2/Dockerfile.cpu Outdated


		# See http://bugs.python.org/issue19846
		ENV LANG C.UTF-8

ghost Nov 17, 2019

Move all ARGs and ENVs to the top. ARGs before ENV. Open MP envs are an exception.

Contributor Author

akartsky Nov 19, 2019

done

docker/1.3.1/py2/Dockerfile.cpu Outdated

+              # Add arguments to achieve the version, python and url
+              ARG PYTHON_VERSION=2.7
+              RUN curl -o ~/miniconda.sh -O  https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh  && \

ghost Nov 17, 2019

Use && in the beginning of the next line (applies to all multi-line RUNs in all docker files)

Contributor Author

akartsky Nov 19, 2019

done

docker/1.3.1/py2/Dockerfile.cpu Outdated

+                   mkl==2019.4 \
+                   mkl-include==2019.4 \
+                   cython==0.29.12 \
+                   typing==3.7.4 && \

ghost Nov 17, 2019

Update all libraries to latest versions

Contributor

saimidu Nov 17, 2019

Easiest way to do this is to unpin all versions, run the build, if it's successful, pin versions according to the ones that come up in a pip freeze | grep <package> within the container.

docker/1.3.1/py2/Dockerfile.cpu Outdated

+                                   pillow==6.1.0 \
+                                   h5py==2.9.0 \
+                                   requests==2.22.0 && \
+                  conda install pytorch==$PYTORCH_VERSION torchvision==$TORCHVISION_VERSION cpuonly -c pytorch && \

ghost Nov 17, 2019

Align according to best practices doc

Contributor Author

akartsky Nov 19, 2019

done

docker/1.3.1/py2/Dockerfile.gpu Outdated

		@@ -0,0 +1,101 @@
		FROM nvidia/cuda:10.0-cudnn7-devel-ubuntu16.04

ghost Nov 17, 2019

Please make same changes as CPU to GPU.

Contributor Author

akartsky Nov 19, 2019

done

docker/1.3.1/py2/Dockerfile.gpu Outdated

+              ARG PYTHON_VERSION=2.7
+              # Install OpenSSH, Allow OpenSSH to talk to containers without asking for confirmation
+              RUN apt-get install -y --no-install-recommends openssh-client openssh-server && \

ghost Nov 17, 2019

Move each install to one line.

Contributor Author

akartsky Nov 19, 2019

done

docker/1.3.1/py2/mms-entrypoint.py Outdated

+              import sys
+              import os.path
+              from sagemaker_pytorch_serving_container import serving

ghost Nov 17, 2019

Should this be moved inside the if condition? Keep this consistent with the mms-entrypoint for mxnet.

ghost Nov 19, 2019

docker/1.3.1/py3/Dockerfile.cpu

		@@ -0,0 +1,81 @@
		FROM ubuntu:16.04

ghost Nov 17, 2019

Same as previous docker files.

docker/1.3.1/py3/Dockerfile.gpu Outdated

		@@ -0,0 +1,93 @@
		FROM nvidia/cuda:10.0-cudnn7-devel-ubuntu16.04

ghost Nov 17, 2019

Same as previous docker file.

Contributor Author

akartsky Nov 19, 2019

done

akartsky added 4 commits

November 18, 2019 16:07


          docker best practies changes

03f91d3


          updated cuda

648e9e1


          fixed typo

9e5ff2a


          fixed typo

ec3620c

nikhil-sk previously approved these changes

View reviewed changes

Collaborator

sagemaker-bot commented Nov 19, 2019

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository


          line break changes

6abf671

akartsky dismissed nikhil-sk’s stale review via

6abf671

November 19, 2019 03:20

Collaborator

sagemaker-bot commented Nov 19, 2019

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

akartsky added 3 commits

November 19, 2019 14:21


          fixed typo

03c682a


          fixed typo


          moved import serving inside if

5f1dc15

Collaborator

sagemaker-bot commented Nov 20, 2019

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

ghost previously approved these changes

View reviewed changes

docker/1.2.0/py2/Dockerfile.cpu Outdated

                                    pillow==6.1.0 \
                                    h5py==2.9.0 \
                                    requests==2.22.0 && \

ghost Nov 19, 2019

Remove empty line

docker/1.2.0/py2/Dockerfile.cpu Outdated

                                    pillow==6.1.0 \
                                    h5py==2.9.0 \
                                    requests==2.22.0 && \

ghost Nov 19, 2019

&& should be in the next line. This needs to be fixed across the file.

docker/1.2.0/py2/Dockerfile.cpu Outdated

		@@ -50,6 +50,7 @@ RUN conda install -c conda-forge awscli==1.16.210 opencv==4.0.1 && \
		pillow==6.1.0 \
		h5py==2.9.0 \

ghost Nov 19, 2019

Alignment changes required. The names of the libraries must be aligned with conda...

docker/1.2.0/py2/Dockerfile.cpu Outdated

		@@ -50,6 +50,7 @@ RUN conda install -c conda-forge awscli==1.16.210 opencv==4.0.1 && \
		pillow==6.1.0 \

ghost Nov 19, 2019

Lines 28~41 alignment is off by one space.

docker/1.3.1/py2/mms-entrypoint.py Outdated

+              import sys
+              import os.path
+              from sagemaker_pytorch_serving_container import serving

ghost Nov 19, 2019


          fixed typo

f333e2c

akartsky dismissed ghost ’s stale review via

f333e2c

November 20, 2019 01:27

ghost approved these changes

View reviewed changes

nikhil-sk reviewed

View reviewed changes

docker/1.3.1/py3/Dockerfile.gpu

+                  pillow==5.4.1 \
+                  h5py==2.9.0 \
+                  requests==2.22.0 \
+               && conda install -c \

Contributor

nikhil-sk Nov 20, 2019

You can combine all commands of '-c pytorch' together i.e. magma-cuda100 is also being fetched from pytorch channel.

nikhil-sk approved these changes

View reviewed changes

Collaborator

sagemaker-bot commented Nov 20, 2019

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

akartsky merged commit d0b02e2 into aws:master

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet