search

GoogleCloudPlatform / cloudml-samples

Cloud ML Engine repo. Please visit the new Vertex AI samples repo at https://github.com/GoogleCloudPlatform/vertex-ai-samples
https://cloud.google.com/ai-platform/docs/
Apache License 2.0
1.52k stars 857 forks source link

census/tf-keras training tensorflow2 #476

Closed wutayng closed 4 years ago

wutayng commented 4 years ago

Is your feature request related to a problem? Please describe. The tutorial here works correctly for the following command:

gcloud ai-platform local train \
  --package-path trainer \
  --module-name trainer.task \
  --job-dir local-training-output

Model training occurs and epochs are shown in the console.

However, I need this pipeline to work in tensorflow2. When I change requirements.txt to tensorflow 2.1.0 and re-install, I get the following output for the gcloud command above

<subprocess.Popen object at 0x106cdb650>

Nothing else appears in the console and no training occurs.

Describe the solution you'd like A (probably small) code change to make this tutorial work correctly for tensorflow 2

gogasca commented 4 years ago

I was able to reproduce issue, let me look into it and get back to you

gogasca commented 4 years ago

We opened internal issue. Looking at it now.

kpe commented 4 years ago

I have the same problem - no idea what it is

gogasca commented 4 years ago

We found the issue, we will provide an update about the fix soon. Thanks

gogasca commented 4 years ago

Issue was submitted it may make it for tomorrow SDK new version I will confirm it

gogasca commented 4 years ago

Issue is fixed in latest SDK version, thanks for reporting SDK must be 287 or higher.

gcloud --version