Defaults to "info" or "debug", depending on `debug`_ option. When debug mode is enabled, Loglevel is set to "debug", otherwise, Loglevel is set to "info".
Most modules of NNI will be affected by this value, including NNI manager, tuner, training service, etc.
The exception is trial, whose logging level is directly managed by trial code.
For Python modules, "trace" acts as logging level 0 and "fatal" acts as ``logging.CRITICAL``.
experimentWorkingDirectory
--------------------------
Specify the :ref:`directory <path>` to place log, checkpoint, metadata, and other run-time stuff.
type: ``Optional[str]``
By default uses ``~/nni-experiments``.
NNI will create a subdirectory named by experiment ID, so it is safe to use the same directory for multiple experiments.
tunerGpuIndices
---------------
Limit the GPUs visible to tuner, assessor, and advisor.
type: ``Optional[list[int] | str | int]``
This will be the ``CUDA_VISIBLE_DEVICES`` environment variable of tuner process.
Because tuner, assessor, and advisor run in the same process, this option will affect them all.
tuner
-----
Specify the tuner.
type: Optional `AlgorithmConfig`_
The built-in tuners can be found `here <../builtin_tuner.rst>`__ and you can follow `this tutorial <../Tuner/CustomizeTuner.rst>`__ to customize a new tuner.
assessor
--------
Specify the assessor.
type: Optional `AlgorithmConfig`_
The built-in assessors can be found `here <../builtin_assessor.rst>`__ and you can follow `this tutorial <../Assessor/CustomizeAssessor.rst>`__ to customize a new assessor.
advisor
-------
Specify the advisor.
type: Optional `AlgorithmConfig`_
NNI provides two built-in advisors: `BOHB <../Tuner/BohbAdvisor.rst>`__ and `Hyperband <../Tuner/HyperbandAdvisor.rst>`__, and you can follow `this tutorial <../Tuner/CustomizeAdvisor.rst>`__ to customize a new advisor.
trainingService
---------------
Specify the `training service <../TrainingService/Overview.rst>`__.
type: `TrainingServiceConfig`_
sharedStorage
-------------
Configure the shared storage, detailed usage can be found `here <../Tutorial/HowToUseSharedStorage.rst>`__.
type: Optional `SharedStorageConfig`_
.. list-table::
:widths: 10 10 80
:header-rows: 1
* - Field Name
- Type
- Description
* - experimentName
- ``Optional[str]``
- Mnemonic name of the experiment, which will be shown in WebUI and nnictl.
* - searchSpaceFile
- ``Optional[str]``
- Path_ to the JSON file containing the search space.
Search space format is determined by tuner. The common format for built-in tuners is documented `here <../Tutorial/SearchSpaceSpec.rst>`__.
Mutually exclusive to ``searchSpace``.
* - searchSpace
- ``Optional[JSON]``
- Search space object.
The format is determined by tuner. Common format for built-in tuners is documented `here <../Tutorial/SearchSpaceSpec.rst>`__.
Note that ``None`` means "no such field" so empty search space should be written as ``{}``.
Mutually exclusive to ``searchSpaceFile``.
* - trialCommand
- ``str``
- Command to launch trial.
The command will be executed in bash on Linux and macOS, and in PowerShell on Windows.
Note that using ``python3`` on Linux and macOS, and using ``python`` on Windows.
* - trialCodeDirectory
- ``str``
- `Path`_ to the directory containing trial source files.
default: ``"."``.
All files in this directory will be sent to the training machine, unless in the ``.nniignore`` file.
(See :ref:`nniignore <nniignore>` for details.)
* - trialConcurrency
- ``int``
- Specify how many trials should be run concurrently.
The real concurrency also depends on hardware resources and may be less than this value.
* - trialGpuNumber
- ``Optional[int]``
- This field might have slightly different meanings for various training services,
especially when set to ``0`` or ``None``.
See `training service's document <../training_services.rst>`__ for details.
In local mode, setting the field to ``0`` will prevent trials from accessing GPU (by empty ``CUDA_VISIBLE_DEVICES``).
And when set to ``None``, trials will be created and scheduled as if they did not use GPU,
but they can still use all GPU resources if they want.
* - maxExperimentDuration
- ``Optional[str]``
- Limit the duration of this experiment if specified.
format: ``number + s|m|h|d``
examples: ``"10m"``, ``"0.5h"``
When time runs out, the experiment will stop creating trials but continue to serve WebUI.
* - maxTrialNumber
- ``Optional[int]``
- Limit the number of trials to create if specified.
When the budget runs out, the experiment will stop creating trials but continue to serve WebUI.
* - maxTrialDuration
- ``Optional[str]``
- Limit the duration of trial job if specified.
format: ``number + s|m|h|d``
examples: ``"10m"``, ``"0.5h"``
When time runs out, the current trial job will stop.
* - nniManagerIp
- ``Optional[str]``
- IP of the current machine, used by training machines to access NNI manager. Not used in local mode.
If not specified, IPv4 address of ``eth0`` will be used.
Except for the local mode, it is highly recommended to set this field manually.
Defaults to "info" or "debug", depending on ``debug`` option. When debug mode is enabled, Loglevel is set to "debug", otherwise, Loglevel is set to "info".
Most modules of NNI will be affected by this value, including NNI manager, tuner, training service, etc.
The exception is trial, whose logging level is directly managed by trial code.
For Python modules, "trace" acts as logging level 0 and "fatal" acts as ``logging.CRITICAL``.
* - experimentWorkingDirectory
- ``Optional[str]``
- Specify the :ref:`directory <path>` to place log, checkpoint, metadata, and other run-time stuff.
By default uses ``~/nni-experiments``.
NNI will create a subdirectory named by experiment ID, so it is safe to use the same directory for multiple experiments.
* - tunerGpuIndices
- ``Optional[list[int] | str | int]``
- Limit the GPUs visible to tuner, assessor, and advisor.
This will be the ``CUDA_VISIBLE_DEVICES`` environment variable of tuner process.
Because tuner, assessor, and advisor run in the same process, this option will affect them all.
* - tuner
- ``Optional[AlgorithmConfig]``
- Specify the tuner.
The built-in tuners can be found `here <../builtin_tuner.rst>`__ and you can follow `this tutorial <../Tuner/CustomizeTuner.rst>`__ to customize a new tuner.
* - assessor
- ``Optional[AlgorithmConfig]``
- Specify the assessor.
The built-in assessors can be found `here <../builtin_assessor.rst>`__ and you can follow `this tutorial <../Assessor/CustomizeAssessor.rst>`__ to customize a new assessor.
* - advisor
- ``Optional[AlgorithmConfig]``
- Specify the advisor.
NNI provides two built-in advisors: `BOHB <../Tuner/BohbAdvisor.rst>`__ and `Hyperband <../Tuner/HyperbandAdvisor.rst>`__, and you can follow `this tutorial <../Tuner/CustomizeAdvisor.rst>`__ to customize a new advisor.
* - trainingService
- ``TrainingServiceConfig``
- Specify the `training service <../TrainingService/Overview.rst>`__.
* - sharedStorage
- ``Optional[SharedStorageConfig]``
- Configure the shared storage, detailed usage can be found `here <../Tutorial/HowToUseSharedStorage.rst>`__.
AlgorithmConfig
^^^^^^^^^^^^^^^
...
...
@@ -363,42 +250,34 @@ For customized algorithms, there are two ways to describe them:
2. Specify code directory and class name directly.
.. list-table::
:widths: 10 10 80
:header-rows: 1
name
----
* - Field Name
- Type
- Description
Name of the built-in or registered algorithm.
* - name
- ``Optional[str]``
- Name of the built-in or registered algorithm.
``str`` for the built-in and registered algorithm, ``None`` for other customized algorithms.
type: ``str`` for the built-in and registered algorithm, ``None`` for other customized algorithms.
* - className
- ``Optional[str]``
- Qualified class name of not registered customized algorithm.
``None`` for the built-in and registered algorithm, ``str`` for other customized algorithms.
example: ``"my_tuner.MyTuner"``
* - codeDirectory
- ``Optional[str]``
- `Path`_ to the directory containing the customized algorithm class.
``None`` for the built-in and registered algorithm, ``str`` for other customized algorithms.
className
---------
Qualified class name of not registered customized algorithm.
type: ``None`` for the built-in and registered algorithm, ``str`` for other customized algorithms.
example: ``"my_tuner.MyTuner"``
codeDirectory
-------------
`Path`_ to the directory containing the customized algorithm class.
type: ``None`` for the built-in and registered algorithm, ``str`` for other customized algorithms.
classArgs
---------
Keyword arguments passed to algorithm class' constructor.
type: ``Optional[dict[str, Any]]``
See algorithm's document for supported value.
* - classArgs
- ``Optional[dict[str, Any]]``
- Keyword arguments passed to algorithm class' constructor.
See algorithm's document for supported value.
TrainingServiceConfig
^^^^^^^^^^^^^^^^^^^^^
...
...
@@ -407,635 +286,421 @@ One of the following:
- `LocalConfig`_
- `RemoteConfig`_
- :ref:`OpenpaiConfig <openpai-class>`
- `OpenpaiConfig`_
- `AmlConfig`_
- `DlcConfig`_
- `HybridConfig`_
For `Kubeflow <../TrainingService/KubeflowMode.rst>`_, `FrameworkController <../TrainingService/FrameworkControllerMode.rst>`_, and `AdaptDL <../TrainingService/AdaptDLMode.rst>`_ training platforms, it is suggested to use `v1 config schema <../Tutorial/ExperimentConfig.rst>`_ for now.
LocalConfig
-----------
Detailed usage can be found `here <../TrainingService/LocalMode.rst>`__.
platform
""""""""
Constant string ``"local"``.
.. list-table::
:widths: 10 10 80
:header-rows: 1
* - Field Name
- Type
- Description
useActiveGpu
""""""""""""
* - platform
- ``"local"``
-
Specify whether NNI should submit trials to GPUs occupied by other tasks.
type: ``Optional[bool]``
Must be set when `trialGpuNumber`_ greater than zero.
Following processes can make GPU "active":
* - useActiveGpu
- ``Optional[bool]``
- Specify whether NNI should submit trials to GPUs occupied by other tasks.
Must be set when ``trialGpuNumber`` greater than zero.
Following processes can make GPU "active":
- non-NNI CUDA programs
- graphical desktop
- trials submitted by other NNI instances, if you have more than one NNI experiments running at same time
- other users' CUDA programs, if you are using a shared server
If you are using a graphical OS like Windows 10 or Ubuntu desktop, set this field to ``True``, otherwise, the GUI will prevent NNI from launching any trial.
When you create multiple NNI experiments and ``useActiveGpu`` is set to ``True``, they will submit multiple trials to the same GPU(s) simultaneously.
maxTrialNumberPerGpu
""""""""""""""""""""
If you are using a graphical OS like Windows 10 or Ubuntu desktop, set this field to ``True``, otherwise, the GUI will prevent NNI from launching any trial.
When you create multiple NNI experiments and ``useActiveGpu`` is set to ``True``, they will submit multiple trials to the same GPU(s) simultaneously.
Specify how many trials can share one GPU.
type: ``int``
default: ``1``
gpuIndices
""""""""""
Limit the GPUs visible to trial processes.
type: ``Optional[list[int] | str | int]``
If `trialGpuNumber`_ is less than the length of this value, only a subset will be visible to each trial.
This will be used as ``CUDA_VISIBLE_DEVICES`` environment variable.
* - maxTrialNumberPerGpu
- ``int``
- Specify how many trials can share one GPU.
default: ``1``
* - gpuIndices
- ``Optional[list[int] | str | int]``
- Limit the GPUs visible to trial processes.
If ``trialGpuNumber`` is less than the length of this value, only a subset will be visible to each trial.
This will be used as ``CUDA_VISIBLE_DEVICES`` environment variable.
RemoteConfig
------------
Detailed usage can be found `here <../TrainingService/RemoteMachineMode.rst>`__.
If not specified, `sshKeyFile`_ will be used instead.
sshKeyFile
**********
`Path`_ to sshKeyFile (identity file).
type: ``Optional[str]``
.. list-table::
:widths: 10 10 80
:header-rows: 1
Only used when `password`_ is not specified.
* - Field Name
- Type
- Description
* - host
- ``str``
- IP or hostname (domain name) of the machine.
sshPassphrase
*************
* - port
- ``int``
- SSH service port.
default: ``22``
Passphrase of SSH identity file.
* - user
- ``str``
- Login user name.
type: ``Optional[str]``
* - password
- ``Optional[str]``
- If not specified, ``sshKeyFile`` will be used instead.
* - sshKeyFile
- ``Optional[str]``
- `Path`_ to ``sshKeyFile`` (identity file).
Only used when ``password`` is not specified.
useActiveGpu
************
* - sshPassphrase
- ``Optional[str]``
- Passphrase of SSH identity file.
Specify whether NNI should submit trials to GPUs occupied by other tasks.
type: ``bool``
default: ``False``
Must be set when `trialGpuNumber`_ greater than zero.
Following processes can make GPU "active":
* - useActiveGpu
- ``bool``
- Specify whether NNI should submit trials to GPUs occupied by other tasks.
default: ``False``
Must be set when ``trialGpuNumber`` greater than zero.
Following processes can make GPU "active":
- non-NNI CUDA programs
- graphical desktop
- trials submitted by other NNI instances, if you have more than one NNI experiments running at same time
- other users' CUDA programs, if you are using a shared server
If your remote machine is a graphical OS like Ubuntu desktop, set this field to ``True``, otherwise, the GUI will prevent NNI from launching any trial.
When you create multiple NNI experiments and ``useActiveGpu`` is set to ``True``, they will submit multiple trials to the same GPU(s) simultaneously.
maxTrialNumberPerGpu
********************
Specify how many trials can share one GPU.
type: ``int``
default: ``1``
gpuIndices
**********
If your remote machine is a graphical OS like Ubuntu desktop, set this field to ``True``, otherwise, the GUI will prevent NNI from launching any trial.
When you create multiple NNI experiments and ``useActiveGpu`` is set to ``True``, they will submit multiple trials to the same GPU(s) simultaneously.
Limit the GPUs visible to trial processes.
* - maxTrialNumberPerGpu
- ``int``
- Specify how many trials can share one GPU.
default: ``1``
type: ``Optional[list[int] | str | int]``
* - gpuIndices
- ``Optional[list[int] | str | int]``
- Limit the GPUs visible to trial processes.
If ``trialGpuNumber`` is less than the length of this value, only a subset will be visible to each trial.
This will be used as ``CUDA_VISIBLE_DEVICES`` environment variable.
If `trialGpuNumber`_ is less than the length of this value, only a subset will be visible to each trial.
This will be used as ``CUDA_VISIBLE_DEVICES`` environment variable.
pythonPath
**********
Specify a Python environment.
type: ``Optional[str]``
This path will be inserted at the front of PATH. Here are some examples:
* - pythonPath
- ``Optional[str]``
- Specify a Python environment.
This path will be inserted at the front of PATH. Here are some examples:
- (linux) pythonPath: ``/opt/python3.7/bin``
- (windows) pythonPath: ``C:/Python37``
If you are working on Anaconda, there is some difference. On Windows, you also have to add ``../script`` and ``../Library/bin`` separated by ``;``. Examples are as below:
If you are working on Anaconda, there is some difference. On Windows, you also have to add ``../script`` and ``../Library/bin`` separated by ``;``. Examples are as below:
An example can be found `here <https://github.com/microsoft/pai/blob/master/docs/manual/cluster-user/examples/hello-world-job.yaml>`__.
AmlConfig
---------
Detailed usage can be found `here <../TrainingService/AMLMode.rst>`__.
.. list-table::
:widths: 10 10 80
:header-rows: 1
platform
""""""""
Constant string ``"aml"``.
dockerImage
"""""""""""
Name and tag of docker image to run the trials.
type: ``str``
default: ``"msranni/nni:latest"``
subscriptionId
""""""""""""""
Azure subscription ID.
type: ``str``
resourceGroup
"""""""""""""
Azure resource group name.
type: ``str``
* - Field Name
- Type
- Description
workspaceName
"""""""""""""
* - platform
- ``"aml"``
-
Azure workspace name.
* - dockerImage
- ``str``
- Name and tag of docker image to run the trials.
default: ``"msranni/nni:latest"``
type: ``str``
* - subscriptionId
- ``str``
- Azure subscription ID.
* - resourceGroup
- ``str``
- Azure resource group name.
computeTarget
"""""""""""""
AML compute cluster name.
type: ``str``
* - workspaceName
- ``str``
- Azure workspace name.
* - computeTarget
- ``str``
- AML compute cluster name.
DlcConfig
---------
Detailed usage can be found `here <../TrainingService/DlcMode.rst>`__.
.. list-table::
:widths: 10 10 80
:header-rows: 1
platform
""""""""
Constant string ``"dlc"``.
type
""""
Job spec type.
type: ``str``
default: ``"worker"``
image
"""""
Name and tag of docker image to run the trials.
type: ``str``
jobType
"""""""
PAI-DLC training job type, ``"TFJob"`` or ``"PyTorchJob"``.
type: ``str``
podCount
""""""""
Pod count to run a single training job.
type: ``str``
ecsSpec
"""""""
Training server config spec string.
* - Field Name
- Type
- Description
type: ``str``
* - platform
- ``"dlc"``
-
* - type
- ``str``
- Job spec type.
default: ``"worker"``.
region
""""""
* - image
- ``str``
- Name and tag of docker image to run the trials.
The region where PAI-DLC public-cluster locates.
* - jobType
- ``str``
- PAI-DLC training job type, ``"TFJob"`` or ``"PyTorchJob"``.
type: ``str``
* - podCount
- ``str``
- Pod count to run a single training job.
* - ecsSpec
- ``str``
- Training server config spec string.
nasDataSourceId
"""""""""""""""
* - region
- ``str``
- The region where PAI-DLC public-cluster locates.
The NAS datasource id configurated in PAI-DLC side.
* - nasDataSourceId
- ``str``
- The NAS datasource id configurated in PAI-DLC side.
type: ``str``
* - accessKeyId
- ``str``
- The accessKeyId of your cloud account.
* - accessKeySecret
- ``str``
- The accessKeySecret of your cloud account.
* - localStorageMountPoint
- ``str``
- The mount point of the NAS on PAI-DSW server, default is /home/admin/workspace/.
accessKeyId
"""""""""""
The accessKeyId of your cloud account.
type: ``str``
accessKeySecret
"""""""""""""""
The accessKeySecret of your cloud account.
type: ``str``
localStorageMountPoint
""""""""""""""""""""""
The mount point of the NAS on PAI-DSW server, default is /home/admin/workspace/.
type: ``str``
containerStorageMountPoint
""""""""""""""""""""""""""
The mount point of the NAS on PAI-DLC side, default is /root/data/.
type: ``str``
* - containerStorageMountPoint
- ``str``
- The mount point of the NAS on PAI-DLC side, default is /root/data/.
HybridConfig
------------
Currently only support `LocalConfig`_, `RemoteConfig`_, :ref:`OpenpaiConfig <openpai-class>` and `AmlConfig`_ . Detailed usage can be found `here <../TrainingService/HybridMode.rst>`__.
type: list of `TrainingServiceConfig`_
Currently only support `LocalConfig`_, `RemoteConfig`_, `OpenpaiConfig`_ and `AmlConfig`_ . Detailed usage can be found `here <../TrainingService/HybridMode.rst>`__.
SharedStorageConfig
^^^^^^^^^^^^^^^^^^^
Detailed usage can be found `here <../Tutorial/HowToUseSharedStorage.rst>`__.
nfsConfig
---------
storageType
"""""""""""
Constant string ``"NFS"``.
localMountPoint
"""""""""""""""
.. list-table::
:widths: 10 10 80
:header-rows: 1
The path that the storage has been or will be mounted in the local machine.
* - Field Name
- Type
- Description
type: ``str``
* - storageType
- ``"NFS"``
-
If the path does not exist, it will be created automatically. Recommended to use an absolute path, i.e. ``/tmp/nni-shared-storage``.
* - localMountPoint
- ``str``
- The path that the storage has been or will be mounted in the local machine.
If the path does not exist, it will be created automatically. Recommended to use an absolute path, i.e. ``/tmp/nni-shared-storage``.
* - remoteMountPoint
- ``str``
- The path that the storage will be mounted in the remote machine.
If the path does not exist, it will be created automatically. Recommended to use a relative path. i.e. ``./nni-shared-storage``.
remoteMountPoint
""""""""""""""""
* - localMounted
- ``str``
- Specify the object and status to mount the shared storage.
``usermount`` means the user has already mounted this storage on localMountPoint. ``nnimount`` means NNI will try to mount this storage on localMountPoint. ``nomount`` means storage will not mount in the local machine, will support partial storages in the future.
The path that the storage will be mounted in the remote machine.
type: ``str``
If the path does not exist, it will be created automatically. Recommended to use a relative path. i.e. ``./nni-shared-storage``.
localMounted
""""""""""""
Specify the object and status to mount the shared storage.
``usermount`` means the user has already mounted this storage on localMountPoint. ``nnimount`` means NNI will try to mount this storage on localMountPoint. ``nomount`` means storage will not mount in the local machine, will support partial storages in the future.
nfsServer
"""""""""
NFS server host.
type: ``str``
exportedDirectory
"""""""""""""""""
Exported directory of NFS server, detailed `here <https://www.ibm.com/docs/en/aix/7.2?topic=system-nfs-exporting-mounting>`_.
type: ``str``
* - nfsServer
- ``str``
- NFS server host.
* - exportedDirectory
- ``str``
- Exported directory of NFS server, detailed `here <https://www.ibm.com/docs/en/aix/7.2?topic=system-nfs-exporting-mounting>`_.
azureBlobConfig
---------------
storageType
"""""""""""
Constant string ``"AzureBlob"``.
localMountPoint
"""""""""""""""
The path that the storage has been or will be mounted in the local machine.
type: ``str``
If the path does not exist, it will be created automatically. Recommended to use an absolute path, i.e. ``/tmp/nni-shared-storage``.
remoteMountPoint
""""""""""""""""
The path that the storage will be mounted in the remote machine.
type: ``str``
If the path does not exist, it will be created automatically. Recommended to use a relative path. i.e. ``./nni-shared-storage``.
Note that the directory must be empty when using AzureBlob.
localMounted
""""""""""""
Specify the object and status to mount the shared storage.
``usermount`` means the user has already mounted this storage on localMountPoint. ``nnimount`` means NNI will try to mount this storage on localMountPoint. ``nomount`` means storage will not mount in the local machine, will support partial storages in the future.
storageAccountName
""""""""""""""""""
Azure storage account name.
type: ``str``
storageAccountKey
"""""""""""""""""
Azure storage account key.
type: ``Optional[str]``
containerName
"""""""""""""
AzureBlob container name.
type: ``str``
.. list-table::
:widths: 10 10 80
:header-rows: 1
* - Field Name
- Type
- Description
* - storageType
- ``"AzureBlob"``
-
* - localMountPoint
- ``str``
- The path that the storage has been or will be mounted in the local machine.
If the path does not exist, it will be created automatically. Recommended to use an absolute path, i.e. ``/tmp/nni-shared-storage``.
* - remoteMountPoint
- ``str``
- The path that the storage will be mounted in the remote machine.
If the path does not exist, it will be created automatically. Recommended to use a relative path. i.e. ``./nni-shared-storage``.
Note that the directory must be empty when using AzureBlob.
* - localMounted
- ``str``
- Specify the object and status to mount the shared storage.
``usermount`` means the user has already mounted this storage on localMountPoint. ``nnimount`` means NNI will try to mount this storage on localMountPoint. ``nomount`` means storage will not mount in the local machine, will support partial storages in the future.
