TensorFlowSettings Class

TensorFlow job settings.

All required parameters must be populated in order to send to Azure.

Inheritance

Model

TensorFlowSettings

Constructor

TensorFlowSettings(*, python_script_file_path: str, python_interpreter_path: Optional[str] = None, master_command_line_args: Optional[str] = None, worker_command_line_args: Optional[str] = None, parameter_server_command_line_args: Optional[str] = None, worker_count: Optional[int] = None, parameter_server_count: Optional[int] = None, **kwargs)

Parameters

python_script_file_path

str

Required

Required. The python script to execute.

python_interpreter_path

str

Required

The path to the Python interpreter.

master_command_line_args

str

Required

Command line arguments that need to be passed to the python script for the master task.

worker_command_line_args

str

Required

Command line arguments that need to be passed to the python script for the worker task. Optional for single process jobs.

parameter_server_command_line_args

str

Required

Command line arguments that need to be passed to the python script for the parameter server. Optional for single process jobs.

worker_count

int

Required

The number of worker tasks. If specified, the value must be less than or equal to (nodeCount * numberOfGPUs per VM). If not specified, the default value is equal to nodeCount. This property can be specified only for distributed TensorFlow training.

parameter_server_count

int

Required

The number of parameter server tasks. If specified, the value must be less than or equal to nodeCount. If not specified, the default value is equal to 1 for distributed TensorFlow training. This property can be specified only for distributed TensorFlow training.