Search code examples
gitdockerjupyter-notebookazure-machine-learning-serviceopenai-gym

Azure ML Notebook GIT_DISCOVERY_ACROSS_FILESYSTEM


I have created a resource group, a compute instance, a CPU cluster and a notebook as described in Cartpool notebook.

However, when the command ray_env_build_details.wait_for_completion(show_output=True) is executed I get the following error:

Environment version is set. Attempting to register desired version. To auto-version, reset version to None.
Image Build Status: Queued

2022/11/06 17:46:18 Downloading source code...
2022/11/06 17:46:19 Finished downloading source code
2022/11/06 17:46:20 Creating Docker network: acb_default_network, driver: 'bridge'
2022/11/06 17:46:20 Successfully set up Docker network: acb_default_network
2022/11/06 17:46:20 Setting up Docker configuration...
2022/11/06 17:46:21 Successfully set up Docker configuration
2022/11/06 17:46:21 Logging in to registry: ca98820ccecd46c29745d8e4c35dca23.azurecr.io
2022/11/06 17:46:22 Successfully logged into ca98820ccecd46c29745d8e4c35dca23.azurecr.io
2022/11/06 17:46:22 Executing step ID: acb_step_0. Timeout(sec): 5400, Working directory: '', Network: 'acb_default_network'
2022/11/06 17:46:22 Scanning for dependencies...
2022/11/06 17:46:22 Output from dependency scanning: fatal: not a git repository (or any parent up to mount point /)
Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).
fatal: not a git repository (or any parent up to mount point /)
Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).
unexpected dockerfile format
failed to run step ID: acb_step_0: failed to scan dependencies: exit status 1

Run ID: chc failed after 5s. Error: failed during run, err: exit status 1
Image Build Status: Failed

<azureml.core.environment.ImageBuildDetails at 0x7f27449317c0>

I have tried running git init in a terminal window for the Notebook before running the command, but that didn't help.

Any guidance?


Solution

  • wait_for_completion() – stops the execution of the process of creation of the environment, until all the dependencies are satisfies. If the dependencies are not satisfactory, it will throw the error. Separate the dependencies from the creation process. Then we can avoid the method overlapping.

    enter image description here

    I tried to reproduce the issue. Solved the error.