Swift: GCP image

Created on 27 May 2020  路  9Comments  路  Source: tensorflow/swift

Hi Team,

Since swift for tensorflow has a new release every couple of months, it would be great if you could provide a GCP disk image alongside like python TensorFlow. This is very helpful for those of us who use google cloud platform.

Thanks!

enhancement infrastructure

Most helpful comment

A set of S4TF experimental DLVM images are now live 馃殌 To find them, run:

gcloud compute images list --project deeplearning-platform-release --no-standard-images | grep swift

As an example of how to create a small GCE instance using the cpu image:

gcloud compute instances create s4tf-ubuntu \
  --image-project=deeplearning-platform-release \
  --image-family=swift-latest-cpu-ubuntu-1804 \
  --maintenance-policy=TERMINATE \
  --machine-type=n1-standard-2 \
  --boot-disk-size=256GB

I will add this to our docs in a bit more detail.

All 9 comments

Hi Sina! Just to confirm your request - are you referring to a machine image like these, the Deep Learning VMs?

Hi Michelle. Yes that鈥檚 it. Thanks.

I agree that having an S4TF version of these images would be great! I've added this request to the DLVM team's queue. In the meantime, you can use this script to install the S4TF toolchain on a GCE instance with a stock Ubuntu 18.04 image.

You could automate setup with something like this:

mkdir -p ~/repos/tensorflow
git clone https://github.com/tensorflow/swift.git ~/repos/tensorflow/swift
sudo ~/repos/tensorflow/swift/utils/install-ubuntu1804.sh \
  --toolchain-url https://storage.googleapis.com/swift-tensorflow-artifacts/releases/v0.9/rc1/swift-tensorflow-RELEASE-0.9-ubuntu18.04.tar.gz
export PATH="${HOME}/swift-toolchain/usr/bin:$PATH"

Thank you very much!

Good news! A DLVM image with S4TF for CPU and CUDA 11 will be released as experimental next week 馃帀

I will update this thread with the gcloud command once it's live.

A set of S4TF experimental DLVM images are now live 馃殌 To find them, run:

gcloud compute images list --project deeplearning-platform-release --no-standard-images | grep swift

As an example of how to create a small GCE instance using the cpu image:

gcloud compute instances create s4tf-ubuntu \
  --image-project=deeplearning-platform-release \
  --image-family=swift-latest-cpu-ubuntu-1804 \
  --maintenance-policy=TERMINATE \
  --machine-type=n1-standard-2 \
  --boot-disk-size=256GB

I will add this to our docs in a bit more detail.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

wangtz picture wangtz  路  5Comments

elefantel picture elefantel  路  5Comments

RogerS49 picture RogerS49  路  5Comments

frogcjn picture frogcjn  路  3Comments

ifsheldon picture ifsheldon  路  4Comments