search

microsoftarchive / BatchAI

Repo for publishing code Samples and CLI samples for BatchAI service
MIT License
125 stars 62 forks source link

Environment variable AZ_BATCHAI_NUM_GPUS is returning empty string #40

Closed karolzak closed 6 years ago

karolzak commented 6 years ago

Hi, while trying to setup a distributed learning job with TF I found out that AZ_BATCHAI_NUM_GPUS variable (described here) doesnt exist and it returns an empty string:

Part of my job.json config:

...
"masterCommandLineArgs": "--job_name=worker --num_gpus=$AZ_BATCHAI_NUM_GPUS
...

Error msg:

 error: argument --num_gpus: invalid int value: ''

Either support that variable or fix the docs 😀 Thanks!

CC: @AlexanderYukhanov

AlexanderYukhanov commented 6 years ago

thank you for reporting the issue, working on the fix