Cannot use `build_local` on 'AzureML-TensorFlow-2.0-GPU' environment

[ezwiefel]

In azureml-sdk version 1.0.79, when I try to run the following code:

from azureml.core import Workspace, Environment

ws = Workspace.from_config()

tf_env = Environment.get(ws, 'AzureML-TensorFlow-2.0-GPU')
tf_env.build_local(ws)

I get this error:

---------------------------------------------------------------------------
Exception                                 Traceback (most recent call last)
/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/core/environment.py in build_local(self, workspace)
    785         try:
--> 786             recipe = environment_client._get_recipe_for_build(name=self.name, version=self.version)
    787         except Exception as error:

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/_restclient/environment_client.py in _get_recipe_for_build(self, name, version)
    169                                                                                    response.text)
--> 170             raise Exception(message)
    171 

Exception: Error getting recipe specifications. Code: 415
: 

During handling of the above exception, another exception occurred:

UserErrorException                        Traceback (most recent call last)
<ipython-input-6-bcb6c6a77c1e> in <module>
      4 
      5 tf_env = Environment.get(ws, 'AzureML-TensorFlow-2.0-GPU')
----> 6 tf_env.build_local(ws)

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/core/environment.py in build_local(self, workspace)
    788             # Temporary solution to filter attempts to build unregistered environments out
    789             if "Code: 404" in repr(error) or "Code: 415" in repr(error):
--> 790                 raise UserErrorException("Environment '{}' is not registered".format(self.name))
    791             else:
    792                 raise error

UserErrorException: UserErrorException:
	Message: Environment 'AzureML-TensorFlow-2.0-GPU' is not registered
	InnerException None
	ErrorResponse 
{
    "error": {
        "code": "UserError",
        "message": "Environment 'AzureML-TensorFlow-2.0-GPU' is not registered"
    }
}

[ram-msft]

Hi,
Please follow the below document to register the environment.
https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-use-environments#register-environments

[ezwiefel]

@ram-msft As you can see from my code, I'm pulling a registered environment, but when I try to run build_local, I'm told the the environment is not registered.

The code you showed above isn't really relevant.

[ram-msft]

@ezwiefel It's a bug in the Environment. #in-progress

[vizhur]

@ezwiefel The environment you trying to build is global and not registered with your workspace. To build global environment locally you would need to

tf_env = Environment.get(ws, 'AzureML-TensorFlow-2.0-GPU')
# change name as AzureML prefix is reserved
tf_env.name = "my-AzureML-TensorFlow-2.0-GPU"
tf_env_registered = tf_env.register(ws)
tf_env_registered.build_local(ws)

[ezwiefel]

Thanks @vizhur. I'd tried before with a custom environment, but was getting the same error shown above (I just used that as an example here.). It looks like that's been fixed, but now I get a different error when running your code in 1.0.83.

Saving setup content into  /tmp/tmpue0g4adn
---------------------------------------------------------------------------
UnsupportedOperation                      Traceback (most recent call last)
<ipython-input-5-4d0d6cf61cf0> in <module>
      3 tf_env.name = "my-AzureML-TensorFlow-2.0-GPU"
      4 tf_env_registered = tf_env.register(ws)
----> 5 tf_env_registered.build_local(ws)

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/core/environment.py in build_local(self, workspace, platform, **kwargs)
    829                     zip_ref.extractall(setup_temp_dir)
    830 
--> 831                 self._build_environment(setup_temp_dir, check_script, build_script, environment_variables)
    832             except subprocess.CalledProcessError as ex:
    833                 print("Building environment failed with exit code : {}".format(ex.returncode))

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/core/environment.py in _build_environment(self, setup_temp_dir, check_script, build_script, environment_variables)
    984             # Check if environment exists
    985             subprocess.check_call(check_env_command, stdout=sys.stdout,
--> 986                                   stderr=sys.stderr, cwd=setup_temp_dir, env=env)
    987         except subprocess.CalledProcessError:
    988             # else build Environment

/anaconda/envs/azureml_py36/lib/python3.6/subprocess.py in check_call(*popenargs, **kwargs)
    304     check_call(["ls", "-l"])
    305     """
--> 306     retcode = call(*popenargs, **kwargs)
    307     if retcode:
    308         cmd = kwargs.get("args")

/anaconda/envs/azureml_py36/lib/python3.6/subprocess.py in call(timeout, *popenargs, **kwargs)
    285     retcode = call(["ls", "-l"])
    286     """
--> 287     with Popen(*popenargs, **kwargs) as p:
    288         try:
    289             return p.wait(timeout=timeout)

/anaconda/envs/azureml_py36/lib/python3.6/subprocess.py in __init__(self, args, bufsize, executable, stdin, stdout, stderr, preexec_fn, close_fds, shell, cwd, env, universal_newlines, startupinfo, creationflags, restore_signals, start_new_session, pass_fds, encoding, errors)
    685         (p2cread, p2cwrite,
    686          c2pread, c2pwrite,
--> 687          errread, errwrite) = self._get_handles(stdin, stdout, stderr)
    688 
    689         # We wrap OS handles *before* launching the child, otherwise a

/anaconda/envs/azureml_py36/lib/python3.6/subprocess.py in _get_handles(self, stdin, stdout, stderr)
   1202             else:
   1203                 # Assuming file-like object
-> 1204                 c2pwrite = stdout.fileno()
   1205 
   1206             if stderr is None:

UnsupportedOperation: fileno

[vizhur]

@ezwiefel I see... I'l file the issue on our side. Can you please delete ($HOME)/.azureml/envs (there should be folders named azureml_GUID) or use other environment and try again?

[ezwiefel]

@vizhur Same error.

There actually wasn't an envs folder in .azureml, but when I tried to just delete the whole .azureml folder, I'm getting the same error.