o ci @sddlZddlmZddlmZddlmZmZmZm Z ddl Z ddl m Z mZmZddlmZddlmZddlmZmZdd lmZdd lmZerRdd lmZee Z!eGd d d eZeGdddeZeeddGdddZ"dS)N) dataclass)Path) TYPE_CHECKINGListOptionalUnion)CheckpointConfig FailureConfig ScalingConfig) RuntimeEnv) _DEPRECATED)FAIL_FAST_DEPRECATION_MESSAGE%TRAINER_RESOURCES_DEPRECATION_MESSAGE)date_str) PublicAPI) UserCallbackcs>eZdZUdZdZeeed<fddZe ddZ Z S)r aConfiguration for scaling training. Args: num_workers: The number of workers (Ray actors) to launch. Each worker will reserve 1 CPU by default. The number of CPUs reserved by each worker can be overridden with the ``resources_per_worker`` argument. use_gpu: If True, training will be done on GPUs (1 per worker). Defaults to False. The number of GPUs reserved by each worker can be overridden with the ``resources_per_worker`` argument. resources_per_worker: If specified, the resources defined in this Dict is reserved for each worker. Define the ``"CPU"`` and ``"GPU"`` keys (case-sensitive) to override the number of CPU or GPUs used by each worker. placement_strategy: The placement strategy to use for the placement group of the Ray actors. See :ref:`Placement Group Strategies ` for the possible options. accelerator_type: [Experimental] If specified, Ray Train will launch the training coordinator and workers on the nodes with the specified type of accelerators. See :ref:`the available accelerator types `. Ensure that your cluster has instances with the specified accelerator type or is able to autoscale to fulfill the request. Example: .. testcode:: from ray.train import ScalingConfig scaling_config = ScalingConfig( # Number of distributed workers. num_workers=2, # Turn on/off GPU. use_gpu=True, ) .. testoutput:: :hide: ... Ntrainer_resourcescs |jdur tttdSN)rDeprecationWarningrsuper __post_init__self __class__K/home/ubuntu/.local/lib/python3.10/site-packages/ray/train/v2/api/config.pyrMs zScalingConfig.__post_init__cCsiSrrrrrr_trainer_resources_not_noneSsz)ScalingConfig._trainer_resources_not_none) __name__ __module__ __qualname____doc__rrdict__annotations__rpropertyr __classcell__rrrrr s , r c@s.eZdZUdZeZeeefe d<ddZ dS)r aVConfiguration related to failure handling of each training run. Args: max_failures: Tries to recover a run at least this many times. Will recover from the latest checkpoint if present. Setting to -1 will lead to infinite recovery retries. Setting to 0 will disable retries. Defaults to 0. fail_fastcCs|jtkr ttdSr)r&r rr rrrrres zFailureConfig.__post_init__N) rrr r!r r&rboolstrr#rrrrrr Xs  r stable) stabilityc@seZdZUdZdZeeed<dZeeed<dZ ee j j ed<dZ eeed<dZeeed<dZeeded <dZeeeefed <eZeed <eZeed <eZeed <eZeed<eZeed<ddZdS) RunConfigaDRuntime configuration for training runs. Args: name: Name of the trial or experiment. If not provided, will be deduced from the Trainable. storage_path: [Beta] Path where all results and checkpoints are persisted. Can be a local directory or a destination on cloud storage. For multi-node training/tuning runs, this must be set to a shared storage location (e.g., S3, NFS). This defaults to the local ``~/ray_results`` directory. storage_filesystem: [Beta] A custom filesystem to use for storage. If this is provided, `storage_path` should be a path with its prefix stripped (e.g., `s3://bucket/path` -> `bucket/path`). failure_config: Failure mode configuration. checkpoint_config: Checkpointing configuration. callbacks: [DeveloperAPI] A list of callbacks that the Ray Train controller will invoke during training. worker_runtime_env: [DeveloperAPI] Runtime environment configuration for all Ray Train worker actors. Nname storage_pathstorage_filesystemfailure_configcheckpoint_configr callbacksworker_runtime_env sync_configverbosestopprogress_reporter log_to_filecsddlm}|jdur||_|jst|_|jst|_t|jtr(|j |_d}gd}|D]}t ||t kr@t | |q0|jsKdt|_|jpOg|_|jpUi|_ddlmtfdd|jDsmtd dS) Nr)DEFAULT_STORAGE_PATHaI`RunConfig({})` is deprecated. This configuration was a Ray Tune API that did not support Ray Train usage well, so we are dropping support going forward. If you heavily rely on these configurations, you can run Ray Train as a single Ray Tune trial. See this issue for more context: https://github.com/ray-project/ray/issues/49454)r3r4r5r6r7zray_train_run-RayTrainCallbackc3s|]}t|VqdSr) isinstance).0cbr9rr sz*RunConfig.__post_init__..zAll callbacks must be instances of `ray.train.UserCallback`. Passing in a Ray Tune callback is no longer supported. See this issue for more context: https://github.com/ray-project/ray/issues/49454)ray.train.constantsr8r-r/r r0rr;ras_posixgetattrr rformatr,rr1r2ray.train.v2.api.callbackr:all ValueError)rr8run_config_deprecation_messageunsupported_paramsparamrr9rrs4        zRunConfig.__post_init__)rrr r!r,rr(r#r-r.pyarrowfs FileSystemr/r r0rr1rr2rr"r r r3r4r5r6r7rrrrrr+ks       r+)#logging dataclassesrpathlibrtypingrrrr pyarrow.fsrIray.air.configrr FailureConfigV1r ScalingConfigV1ray.runtime_envr ray.train.v2._internal.constantsr &ray.train.v2._internal.migration_utilsr rray.train.v2._internal.utilrray.util.annotationsr ray.trainr getLoggerrloggerr+rrrrs*        :