wandb 0.17.0rc1__py3-none-win32.whl

wandb/sdk/wandb_settings.py

@@ -0,0 +1,1973 @@
+import collections.abc
+import configparser
+import enum
+import getpass
+import json
+import logging
+import multiprocessing
+import os
+import platform
+import re
+import shutil
+import socket
+import sys
+import tempfile
+import time
+from dataclasses import dataclass
+from datetime import datetime
+from distutils.util import strtobool
+from functools import reduce
+from typing import (
+    Any,
+    Callable,
+    Dict,
+    FrozenSet,
+    ItemsView,
+    Iterable,
+    Mapping,
+    Optional,
+    Sequence,
+    Set,
+    Tuple,
+    Union,
+    no_type_check,
+)
+from urllib.parse import quote, unquote, urlencode, urlparse, urlsplit
+
+from google.protobuf.wrappers_pb2 import BoolValue, DoubleValue, Int32Value, StringValue
+
+import wandb
+import wandb.env
+from wandb import util
+from wandb.apis.internal import Api
+from wandb.errors import UsageError
+from wandb.proto import wandb_settings_pb2
+from wandb.sdk.internal.system.env_probe_helpers import is_aws_lambda
+from wandb.sdk.lib import filesystem
+from wandb.sdk.lib._settings_toposort_generated import SETTINGS_TOPOLOGICALLY_SORTED
+from wandb.sdk.lib.run_moment import RunMoment
+from wandb.sdk.wandb_setup import _EarlyLogger
+
+from .lib import apikey
+from .lib.gitlib import GitRepo
+from .lib.ipython import _get_python_type
+from .lib.runid import generate_id
+
+if sys.version_info >= (3, 8):
+    from typing import get_args, get_origin, get_type_hints
+else:
+    from typing_extensions import get_args, get_origin, get_type_hints
+
+
+class SettingsPreprocessingError(UsageError):
+    """Raised when the value supplied to a wandb.Settings() setting does not pass preprocessing."""
+
+
+class SettingsValidationError(UsageError):
+    """Raised when the value supplied to a wandb.Settings() setting does not pass validation."""
+
+
+class SettingsUnexpectedArgsError(UsageError):
+    """Raised when unexpected arguments are passed to wandb.Settings()."""
+
+
+def _get_wandb_dir(root_dir: str) -> str:
+    """Get the full path to the wandb directory.
+
+    The setting exposed to users as `dir=` or `WANDB_DIR` is the `root_dir`.
+    We add the `__stage_dir__` to it to get the full `wandb_dir`
+    """
+    # We use the hidden version if it already exists, otherwise non-hidden.
+    if os.path.exists(os.path.join(root_dir, ".wandb")):
+        __stage_dir__ = ".wandb" + os.sep
+    else:
+        __stage_dir__ = "wandb" + os.sep
+
+    path = os.path.join(root_dir, __stage_dir__)
+    if not os.access(root_dir or ".", os.W_OK):
+        wandb.termwarn(
+            f"Path {path} wasn't writable, using system temp directory.",
+            repeat=False,
+        )
+        path = os.path.join(tempfile.gettempdir(), __stage_dir__ or ("wandb" + os.sep))
+
+    return os.path.expanduser(path)
+
+
+def _str_as_bool(val: Union[str, bool]) -> bool:
+    """Parse a string as a bool."""
+    if isinstance(val, bool):
+        return val
+    try:
+        ret_val = bool(strtobool(str(val)))
+        return ret_val
+    except (AttributeError, ValueError):
+        pass
+
+    raise UsageError(f"Could not parse value {val} as a bool.")
+
+
+def _str_as_json(val: Union[str, Dict[str, Any]]) -> Any:
+    """Parse a string as a json object."""
+    if not isinstance(val, str):
+        return val
+    try:
+        return json.loads(val)
+    except (AttributeError, ValueError):
+        pass
+
+    raise UsageError(f"Could not parse value {val} as JSON.")
+
+
+def _str_as_tuple(val: Union[str, Sequence[str]]) -> Tuple[str, ...]:
+    """Parse a (potentially comma-separated) string as a tuple."""
+    if isinstance(val, str):
+        return tuple(val.split(","))
+    return tuple(val)
+
+
+def _datetime_as_str(val: Union[datetime, str]) -> str:
+    """Parse a datetime object as a string."""
+    if isinstance(val, datetime):
+        return datetime.strftime(val, "%Y%m%d_%H%M%S")
+    return val
+
+
+def _redact_dict(
+    d: Dict[str, Any],
+    unsafe_keys: Union[Set[str], FrozenSet[str]] = frozenset({"api_key"}),
+    redact_str: str = "***REDACTED***",
+) -> Dict[str, Any]:
+    """Redact a dict of unsafe values specified by their key."""
+    if not d or unsafe_keys.isdisjoint(d):
+        return d
+    safe_dict = d.copy()
+    safe_dict.update({k: redact_str for k in unsafe_keys.intersection(d)})
+    return safe_dict
+
+
+def _get_program() -> Optional[str]:
+    program = os.getenv(wandb.env.PROGRAM)
+    if program is not None:
+        return program
+    try:
+        import __main__
+
+        if __main__.__spec__ is None:
+            return __main__.__file__
+        # likely run as `python -m ...`
+        return f"-m {__main__.__spec__.name}"
+    except (ImportError, AttributeError):
+        return None
+
+
+def _runmoment_preprocessor(val: Any) -> Optional[RunMoment]:
+    if isinstance(val, RunMoment) or val is None:
+        return val
+    elif isinstance(val, str):
+        return RunMoment.from_uri(val)
+    raise UsageError(f"Could not parse value {val} as a RunMoment.")
+
+
+def _get_program_relpath(
+    program: str, root: Optional[str] = None, _logger: Optional[_EarlyLogger] = None
+) -> Optional[str]:
+    if not program:
+        if _logger is not None:
+            _logger.warning("Empty program passed to get_program_relpath")
+        return None
+
+    root = root or os.getcwd()
+    if not root:
+        return None
+
+    full_path_to_program = os.path.join(
+        root, os.path.relpath(os.getcwd(), root), program
+    )
+    if os.path.exists(full_path_to_program):
+        relative_path = os.path.relpath(full_path_to_program, start=root)
+        if "../" in relative_path:
+            if _logger is not None:
+                _logger.warning(f"Could not save program above cwd: {program}")
+            return None
+        return relative_path
+
+    if _logger is not None:
+        _logger.warning(f"Could not find program at {program}")
+    return None
+
+
+def is_instance_recursive(obj: Any, type_hint: Any) -> bool:  # noqa: C901
+    if type_hint is Any:
+        return True
+
+    origin = get_origin(type_hint)
+    args = get_args(type_hint)
+
+    if origin is None:
+        return isinstance(obj, type_hint)
+
+    if origin is Union:
+        return any(is_instance_recursive(obj, arg) for arg in args)
+
+    if issubclass(origin, collections.abc.Mapping):
+        if not isinstance(obj, collections.abc.Mapping):
+            return False
+        key_type, value_type = args
+
+        for key, value in obj.items():
+            if not is_instance_recursive(key, key_type) or not is_instance_recursive(
+                value, value_type
+            ):
+                return False
+
+        return True
+
+    if issubclass(origin, collections.abc.Sequence):
+        if not isinstance(obj, collections.abc.Sequence) or isinstance(
+            obj, (str, bytes, bytearray)
+        ):
+            return False
+
+        if len(args) == 1 and args[0] != ...:
+            (item_type,) = args
+            for item in obj:
+                if not is_instance_recursive(item, item_type):
+                    return False
+        elif len(args) == 2 and args[-1] == ...:
+            item_type = args[0]
+            for item in obj:
+                if not is_instance_recursive(item, item_type):
+                    return False
+        elif len(args) == len(obj):
+            for item, item_type in zip(obj, args):
+                if not is_instance_recursive(item, item_type):
+                    return False
+        else:
+            return False
+
+        return True
+
+    if issubclass(origin, collections.abc.Set):
+        if not isinstance(obj, collections.abc.Set):
+            return False
+
+        (item_type,) = args
+        for item in obj:
+            if not is_instance_recursive(item, item_type):
+                return False
+
+        return True
+
+    return False
+
+
+@enum.unique
+class Source(enum.IntEnum):
+    OVERRIDE: int = 0
+    BASE: int = 1  # todo: audit this
+    ORG: int = 2
+    ENTITY: int = 3
+    PROJECT: int = 4
+    USER: int = 5
+    SYSTEM: int = 6
+    WORKSPACE: int = 7
+    ENV: int = 8
+    SETUP: int = 9
+    LOGIN: int = 10
+    INIT: int = 11
+    SETTINGS: int = 12
+    ARGS: int = 13
+    RUN: int = 14
+
+
+ConsoleValue = {
+    "auto",
+    "off",
+    "wrap",
+    "redirect",
+    # internal console states
+    "wrap_raw",
+    "wrap_emu",
+}
+
+
+@dataclass()
+class SettingsData:
+    """Settings for the W&B SDK."""
+
+    _args: Sequence[str]
+    _aws_lambda: bool
+    _async_upload_concurrency_limit: int
+    _cli_only_mode: bool  # Avoid running any code specific for runs
+    _code_path_local: str
+    _colab: bool
+    # _config_dict: Config
+    _cuda: str
+    _disable_meta: bool  # Do not collect system metadata
+    _disable_service: (
+        bool  # Disable wandb-service, spin up internal process the old way
+    )
+    _disable_setproctitle: bool  # Do not use setproctitle on internal process
+    _disable_stats: bool  # Do not collect system metrics
+    _disable_update_check: bool  # Disable version check
+    _disable_viewer: bool  # Prevent early viewer query
+    _disable_machine_info: bool  # Disable automatic machine info collection
+    _except_exit: bool
+    _executable: str
+    _extra_http_headers: Mapping[str, str]
+    # file stream retry client configuration
+    _file_stream_retry_max: int  # max number of retries
+    _file_stream_retry_wait_min_seconds: float  # min wait time between retries
+    _file_stream_retry_wait_max_seconds: float  # max wait time between retries
+    _file_stream_timeout_seconds: float  # timeout for individual HTTP requests
+    # file transfer retry client configuration
+    _file_transfer_retry_max: int
+    _file_transfer_retry_wait_min_seconds: float
+    _file_transfer_retry_wait_max_seconds: float
+    _file_transfer_timeout_seconds: float
+    _flow_control_custom: bool
+    _flow_control_disabled: bool
+    # graphql retry client configuration
+    _graphql_retry_max: int
+    _graphql_retry_wait_min_seconds: float
+    _graphql_retry_wait_max_seconds: float
+    _graphql_timeout_seconds: float
+    _internal_check_process: float
+    _internal_queue_timeout: float
+    _ipython: bool
+    _jupyter: bool
+    _jupyter_name: str
+    _jupyter_path: str
+    _jupyter_root: str
+    _kaggle: bool
+    _live_policy_rate_limit: int
+    _live_policy_wait_time: int
+    _log_level: int
+    _network_buffer: int
+    _noop: bool
+    _notebook: bool
+    _offline: bool
+    _sync: bool
+    _os: str
+    _platform: str
+    _proxies: Mapping[str, str]  # dedicated global proxy servers [scheme -> url]
+    _python: str
+    _runqueue_item_id: str
+    _require_core: bool
+    _save_requirements: bool
+    _service_transport: str
+    _service_wait: float
+    _shared: bool
+    _start_datetime: str
+    _start_time: float
+    _stats_pid: int  # (internal) base pid for system stats
+    _stats_sample_rate_seconds: float
+    _stats_samples_to_average: int
+    _stats_join_assets: (
+        bool  # join metrics from different assets before sending to backend
+    )
+    _stats_neuron_monitor_config_path: (
+        str  # path to place config file for neuron-monitor (AWS Trainium)
+    )
+    _stats_open_metrics_endpoints: Mapping[str, str]  # open metrics endpoint names/urls
+    # open metrics filters in one of the two formats:
+    # - {"metric regex pattern, including endpoint name as prefix": {"label": "label value regex pattern"}}
+    # - ("metric regex pattern 1", "metric regex pattern 2", ...)
+    _stats_open_metrics_filters: Union[Sequence[str], Mapping[str, Mapping[str, str]]]
+    _stats_disk_paths: Sequence[str]  # paths to monitor disk usage
+    _stats_buffer_size: int  # number of consolidated samples to buffer before flushing, available in run obj
+    _tmp_code_dir: str
+    _tracelog: str
+    _unsaved_keys: Sequence[str]
+    _windows: bool
+    allow_val_change: bool
+    anonymous: str
+    api_key: str
+    azure_account_url_to_access_key: Dict[str, str]
+    base_url: str  # The base url for the wandb api
+    code_dir: str
+    colab_url: str
+    config_paths: Sequence[str]
+    console: str
+    deployment: str
+    disable_code: bool
+    disable_git: bool
+    disable_hints: bool
+    disable_job_creation: bool
+    disabled: bool  # Alias for mode=dryrun, not supported yet
+    docker: str
+    email: str
+    entity: str
+    files_dir: str
+    force: bool
+    fork_from: Optional[RunMoment]
+    git_commit: str
+    git_remote: str
+    git_remote_url: str
+    git_root: str
+    heartbeat_seconds: int
+    host: str
+    ignore_globs: Tuple[str]
+    init_timeout: float
+    is_local: bool
+    job_name: str
+    job_source: str
+    label_disable: bool
+    launch: bool
+    launch_config_path: str
+    log_dir: str
+    log_internal: str
+    log_symlink_internal: str
+    log_symlink_user: str
+    log_user: str
+    login_timeout: float
+    # magic: Union[str, bool, dict]  # never used in code, deprecated
+    mode: str
+    notebook_name: str
+    problem: str
+    program: str
+    program_abspath: str
+    program_relpath: str
+    project: str
+    project_url: str
+    quiet: bool
+    reinit: bool
+    relogin: bool
+    # todo: add a preprocessing step to convert this to string
+    resume: Union[str, bool]
+    resume_fname: str
+    resumed: bool  # indication from the server about the state of the run (different from resume - user provided flag)
+    root_dir: str
+    run_group: str
+    run_id: str
+    run_job_type: str
+    run_mode: str
+    run_name: str
+    run_notes: str
+    run_tags: Tuple[str]
+    run_url: str
+    sagemaker_disable: bool
+    save_code: bool
+    settings_system: str
+    settings_workspace: str
+    show_colors: bool
+    show_emoji: bool
+    show_errors: bool
+    show_info: bool
+    show_warnings: bool
+    silent: bool
+    start_method: str
+    strict: bool
+    summary_errors: int
+    summary_timeout: int
+    summary_warnings: int
+    sweep_id: str
+    sweep_param_path: str
+    sweep_url: str
+    symlink: bool
+    sync_dir: str
+    sync_file: str
+    sync_symlink_latest: str
+    system_sample: int
+    system_sample_seconds: int
+    table_raise_on_max_row_limit_exceeded: bool
+    timespec: str
+    tmp_dir: str
+    username: str
+    wandb_dir: str
+
+
+class Property:
+    """A class to represent attributes (individual settings) of the Settings object.
+
+    - Encapsulates the logic of how to preprocess and validate values of settings
+    throughout the lifetime of a class instance.
+    - Allows for runtime modification of settings with hooks, e.g. in the case when
+    a setting depends on another setting.
+    - The update() method is used to update the value of a setting.
+    - The `is_policy` attribute determines the source priority when updating the property value.
+    E.g. if `is_policy` is True, the smallest `Source` value takes precedence.
+    """
+
+    def __init__(  # pylint: disable=unused-argument
+        self,
+        name: str,
+        value: Optional[Any] = None,
+        preprocessor: Union[Callable, Sequence[Callable], None] = None,
+        # validators allow programming by contract
+        validator: Union[Callable, Sequence[Callable], None] = None,
+        # runtime converter (hook): properties can be e.g. tied to other properties
+        hook: Union[Callable, Sequence[Callable], None] = None,
+        # always apply hook even if value is None. can be used to replace @property's
+        auto_hook: bool = False,
+        is_policy: bool = False,
+        frozen: bool = False,
+        source: int = Source.BASE,
+        **kwargs: Any,
+    ):
+        self.name = name
+        self._preprocessor = preprocessor
+        self._validator = validator
+        self._hook = hook
+        self._auto_hook = auto_hook
+        self._is_policy = is_policy
+        self._source = source
+
+        # preprocess and validate value
+        self._value = self._validate(self._preprocess(value))
+
+        self.__frozen = frozen
+
+    @property
+    def value(self) -> Any:
+        """Apply the runtime modifier(s) (if any) and return the value."""
+        _value = self._value
+        if (_value is not None or self._auto_hook) and self._hook is not None:
+            _hook = [self._hook] if callable(self._hook) else self._hook
+            for h in _hook:
+                _value = h(_value)
+        return _value
+
+    @property
+    def is_policy(self) -> bool:
+        return self._is_policy
+
+    @property
+    def source(self) -> int:
+        return self._source
+
+    def _preprocess(self, value: Any) -> Any:
+        if value is not None and self._preprocessor is not None:
+            _preprocessor = (
+                [self._preprocessor]
+                if callable(self._preprocessor)
+                else self._preprocessor
+            )
+            for p in _preprocessor:
+                try:
+                    value = p(value)
+                except Exception:
+                    raise SettingsPreprocessingError(
+                        f"Unable to preprocess value for property {self.name}: {value}."
+                    )
+        return value
+
+    def _validate(self, value: Any) -> Any:
+        if value is not None and self._validator is not None:
+            _validator = (
+                [self._validator] if callable(self._validator) else self._validator
+            )
+            for v in _validator:
+                if not v(value):
+                    # failed validation will likely cause a downstream error
+                    # when trying to convert to protobuf, so we raise a hard error
+                    raise SettingsValidationError(
+                        f"Invalid value for property {self.name}: {value}."
+                    )
+        return value
+
+    def update(self, value: Any, source: int = Source.OVERRIDE) -> None:
+        """Update the value of the property."""
+        if self.__frozen:
+            raise TypeError("Property object is frozen")
+        # - always update value if source == Source.OVERRIDE
+        # - if not previously overridden:
+        #   - update value if source is lower than or equal to current source and property is policy
+        #   - update value if source is higher than or equal to current source and property is not policy
+        if (
+            (source == Source.OVERRIDE)
+            or (
+                self._is_policy
+                and self._source != Source.OVERRIDE
+                and source <= self._source
+            )
+            or (
+                not self._is_policy
+                and self._source != Source.OVERRIDE
+                and source >= self._source
+            )
+        ):
+            # self.__dict__["_value"] = self._validate(self._preprocess(value))
+            self._value = self._validate(self._preprocess(value))
+            self._source = source
+
+    def __setattr__(self, key: str, value: Any) -> None:
+        if "_Property__frozen" in self.__dict__ and self.__frozen:
+            raise TypeError(f"Property object {self.name} is frozen")
+        if key == "value":
+            raise AttributeError("Use update() to update property value")
+        self.__dict__[key] = value
+
+    def __str__(self) -> str:
+        return f"{self.value!r}" if isinstance(self.value, str) else f"{self.value}"
+
+    def __repr__(self) -> str:
+        return (
+            f"<Property {self.name}: value={self.value} "
+            f"_value={self._value} source={self._source} is_policy={self._is_policy}>"
+        )
+        # return f"<Property {self.name}: value={self.value}>"
+        # return self.__dict__.__repr__()
+
+
+class Settings(SettingsData):
+    """A class to represent modifiable settings."""
+
+    def _default_props(self) -> Dict[str, Dict[str, Any]]:
+        """Initialize instance attributes (individual settings) as Property objects.
+
+        Helper method that is used in `__init__` together with the class attributes.
+        Note that key names must be the same as the class attribute names.
+        """
+        props: Dict[str, Dict[str, Any]] = dict(
+            _async_upload_concurrency_limit={
+                "preprocessor": int,
+                "validator": self._validate__async_upload_concurrency_limit,
+            },
+            _aws_lambda={
+                "hook": lambda _: is_aws_lambda(),
+                "auto_hook": True,
+            },
+            _code_path_local={
+                "hook": lambda _: _get_program_relpath(self.program),
+                "auto_hook": True,
+            },
+            _colab={
+                "hook": lambda _: "google.colab" in sys.modules,
+                "auto_hook": True,
+            },
+            _disable_machine_info={
+                "value": False,
+                "preprocessor": _str_as_bool,
+            },
+            _disable_meta={
+                "value": False,
+                "preprocessor": _str_as_bool,
+                "hook": lambda x: self._disable_machine_info or x,
+            },
+            _disable_service={
+                "value": False,
+                "preprocessor": _str_as_bool,
+                "is_policy": True,
+            },
+            _disable_setproctitle={"value": False, "preprocessor": _str_as_bool},
+            _disable_stats={
+                "value": False,
+                "preprocessor": _str_as_bool,
+                "hook": lambda x: self._disable_machine_info or x,
+            },
+            _disable_update_check={"preprocessor": _str_as_bool},
+            _disable_viewer={"preprocessor": _str_as_bool},
+            _extra_http_headers={"preprocessor": _str_as_json},
+            # Retry filestream requests for 2 hours before dropping chunk (how do we recover?)
+            # retry_count = seconds_in_2_hours / max_retry_time + num_retries_until_max_60_sec
+            #             = 7200 / 60 + ceil(log2(60/2))
+            #             = 120 + 5
+            _file_stream_retry_max={"value": 125, "preprocessor": int},
+            _file_stream_retry_wait_min_seconds={"value": 2, "preprocessor": float},
+            _file_stream_retry_wait_max_seconds={"value": 60, "preprocessor": float},
+            # A 3 minute timeout for all filestream post requests
+            _file_stream_timeout_seconds={"value": 180, "preprocessor": float},
+            _file_transfer_retry_max={"value": 20, "preprocessor": int},
+            _file_transfer_retry_wait_min_seconds={"value": 2, "preprocessor": float},
+            _file_transfer_retry_wait_max_seconds={"value": 60, "preprocessor": float},
+            _file_transfer_timeout_seconds={"value": 0, "preprocessor": float},
+            _flow_control_disabled={
+                "hook": lambda _: self._network_buffer == 0,
+                "auto_hook": True,
+            },
+            _flow_control_custom={
+                "hook": lambda _: bool(self._network_buffer),
+                "auto_hook": True,
+            },
+            _graphql_retry_max={"value": 20, "preprocessor": int},
+            _graphql_retry_wait_min_seconds={"value": 2, "preprocessor": float},
+            _graphql_retry_wait_max_seconds={"value": 60, "preprocessor": float},
+            _graphql_timeout_seconds={"value": 30.0, "preprocessor": float},
+            _internal_check_process={"value": 8, "preprocessor": float},
+            _internal_queue_timeout={"value": 2, "preprocessor": float},
+            _ipython={
+                "hook": lambda _: _get_python_type() == "ipython",
+                "auto_hook": True,
+            },
+            _jupyter={
+                "hook": lambda _: _get_python_type() == "jupyter",
+                "auto_hook": True,
+            },
+            _kaggle={"hook": lambda _: util._is_likely_kaggle(), "auto_hook": True},
+            _log_level={"value": logging.DEBUG},
+            _network_buffer={"preprocessor": int},
+            _noop={"hook": lambda _: self.mode == "disabled", "auto_hook": True},
+            _notebook={
+                "hook": lambda _: self._ipython
+                or self._jupyter
+                or self._colab
+                or self._kaggle,
+                "auto_hook": True,
+            },
+            _offline={
+                "hook": (
+                    lambda _: True
+                    if self.disabled or (self.mode in ("dryrun", "offline"))
+                    else False
+                ),
+                "auto_hook": True,
+            },
+            _platform={"value": util.get_platform_name()},
+            _proxies={
+                "preprocessor": _str_as_json,
+            },
+            _require_core={"value": False, "preprocessor": _str_as_bool},
+            _save_requirements={"value": True, "preprocessor": _str_as_bool},
+            _service_wait={
+                "value": 30,
+                "preprocessor": float,
+                "validator": self._validate__service_wait,
+            },
+            _shared={
+                "hook": lambda _: self.mode == "shared",
+                "auto_hook": True,
+            },
+            _start_datetime={"preprocessor": _datetime_as_str},
+            _stats_sample_rate_seconds={
+                "value": 2.0,
+                "preprocessor": float,
+                "validator": self._validate__stats_sample_rate_seconds,
+            },
+            _stats_samples_to_average={
+                "value": 15,
+                "preprocessor": int,
+                "validator": self._validate__stats_samples_to_average,
+            },
+            _stats_join_assets={"value": True, "preprocessor": _str_as_bool},
+            _stats_neuron_monitor_config_path={
+                "hook": lambda x: self._path_convert(x),
+            },
+            _stats_open_metrics_endpoints={
+                "preprocessor": _str_as_json,
+            },
+            _stats_open_metrics_filters={
+                # capture all metrics on all endpoints by default
+                "value": (".*",),
+                "preprocessor": _str_as_json,
+            },
+            _stats_disk_paths={
+                "value": ("/",),
+                "preprocessor": _str_as_json,
+            },
+            _stats_buffer_size={
+                "value": 0,
+                "preprocessor": int,
+            },
+            _sync={"value": False},
+            _tmp_code_dir={
+                "value": "code",
+                "hook": lambda x: self._path_convert(self.tmp_dir, x),
+            },
+            _windows={
+                "hook": lambda _: platform.system() == "Windows",
+                "auto_hook": True,
+            },
+            anonymous={"validator": self._validate_anonymous},
+            api_key={"validator": self._validate_api_key},
+            base_url={
+                "value": "https://api.wandb.ai",
+                "preprocessor": lambda x: str(x).strip().rstrip("/"),
+                "validator": self._validate_base_url,
+            },
+            colab_url={
+                "hook": lambda _: self._get_colab_url(),
+                "auto_hook": True,
+            },
+            config_paths={"preprocessor": _str_as_tuple},
+            console={
+                "value": "auto",
+                "validator": self._validate_console,
+                "hook": lambda x: self._convert_console(x),
+                "auto_hook": True,
+            },
+            deployment={
+                "hook": lambda _: "local" if self.is_local else "cloud",
+                "auto_hook": True,
+            },
+            disable_code={
+                "value": False,
+                "preprocessor": _str_as_bool,
+                "hook": lambda x: self._disable_machine_info or x,
+            },
+            disable_hints={"preprocessor": _str_as_bool},
+            disable_git={
+                "value": False,
+                "preprocessor": _str_as_bool,
+                "hook": lambda x: self._disable_machine_info or x,
+            },
+            disable_job_creation={
+                "value": False,
+                "preprocessor": _str_as_bool,
+                "hook": lambda x: self._disable_machine_info or x,
+            },
+            disabled={"value": False, "preprocessor": _str_as_bool},
+            files_dir={
+                "value": "files",
+                "hook": lambda x: self._path_convert(
+                    self.wandb_dir, f"{self.run_mode}-{self.timespec}-{self.run_id}", x
+                ),
+            },
+            force={"preprocessor": _str_as_bool},
+            fork_from={
+                "value": None,
+                "preprocessor": _runmoment_preprocessor,
+            },
+            git_remote={"value": "origin"},
+            heartbeat_seconds={"value": 30},
+            ignore_globs={
+                "value": tuple(),
+                "preprocessor": lambda x: tuple(x) if not isinstance(x, tuple) else x,
+            },
+            init_timeout={"value": 90, "preprocessor": lambda x: float(x)},
+            is_local={
+                "hook": (
+                    lambda _: self.base_url != "https://api.wandb.ai"
+                    if self.base_url is not None
+                    else False
+                ),
+                "auto_hook": True,
+            },
+            job_name={"preprocessor": str},
+            job_source={"validator": self._validate_job_source},
+            label_disable={"preprocessor": _str_as_bool},
+            launch={"preprocessor": _str_as_bool},
+            log_dir={
+                "value": "logs",
+                "hook": lambda x: self._path_convert(
+                    self.wandb_dir, f"{self.run_mode}-{self.timespec}-{self.run_id}", x
+                ),
+            },
+            log_internal={
+                "value": "debug-internal.log",
+                "hook": lambda x: self._path_convert(self.log_dir, x),
+            },
+            log_symlink_internal={
+                "value": "debug-internal.log",
+                "hook": lambda x: self._path_convert(self.wandb_dir, x),
+            },
+            log_symlink_user={
+                "value": "debug.log",
+                "hook": lambda x: self._path_convert(self.wandb_dir, x),
+            },
+            log_user={
+                "value": "debug.log",
+                "hook": lambda x: self._path_convert(self.log_dir, x),
+            },
+            login_timeout={"preprocessor": lambda x: float(x)},
+            mode={"value": "online", "validator": self._validate_mode},
+            problem={"value": "fatal", "validator": self._validate_problem},
+            program={
+                "hook": lambda x: self._get_program(x),
+            },
+            project={"validator": self._validate_project},
+            project_url={"hook": lambda _: self._project_url(), "auto_hook": True},
+            quiet={"preprocessor": _str_as_bool},
+            reinit={"preprocessor": _str_as_bool},
+            relogin={"preprocessor": _str_as_bool},
+            # todo: hack to make to_proto() always happy
+            resume={"preprocessor": lambda x: None if x is False else x},
+            resume_fname={
+                "value": "wandb-resume.json",
+                "hook": lambda x: self._path_convert(self.wandb_dir, x),
+            },
+            resumed={"value": "False", "preprocessor": _str_as_bool},
+            root_dir={
+                "preprocessor": lambda x: str(x),
+                "value": os.path.abspath(os.getcwd()),
+            },
+            run_id={
+                "validator": self._validate_run_id,
+            },
+            run_mode={
+                "hook": lambda _: "offline-run" if self._offline else "run",
+                "auto_hook": True,
+            },
+            run_tags={
+                "preprocessor": lambda x: tuple(x) if not isinstance(x, tuple) else x,
+            },
+            run_url={"hook": lambda _: self._run_url(), "auto_hook": True},
+            sagemaker_disable={"preprocessor": _str_as_bool},
+            save_code={"preprocessor": _str_as_bool},
+            settings_system={
+                "value": os.path.join("~", ".config", "wandb", "settings"),
+                "hook": lambda x: self._path_convert(x),
+            },
+            settings_workspace={
+                "value": "settings",
+                "hook": lambda x: self._path_convert(self.wandb_dir, x),
+            },
+            show_colors={"preprocessor": _str_as_bool},
+            show_emoji={"preprocessor": _str_as_bool},
+            show_errors={"value": "True", "preprocessor": _str_as_bool},
+            show_info={"value": "True", "preprocessor": _str_as_bool},
+            show_warnings={"value": "True", "preprocessor": _str_as_bool},
+            silent={"value": "False", "preprocessor": _str_as_bool},
+            start_method={"validator": self._validate_start_method},
+            strict={"preprocessor": _str_as_bool},
+            summary_timeout={"value": 60, "preprocessor": lambda x: int(x)},
+            summary_warnings={
+                "value": 5,
+                "preprocessor": lambda x: int(x),
+                "is_policy": True,
+            },
+            sweep_url={"hook": lambda _: self._sweep_url(), "auto_hook": True},
+            symlink={"preprocessor": _str_as_bool},
+            sync_dir={
+                "hook": [
+                    lambda _: self._path_convert(
+                        self.wandb_dir, f"{self.run_mode}-{self.timespec}-{self.run_id}"
+                    )
+                ],
+                "auto_hook": True,
+            },
+            sync_file={
+                "hook": lambda _: self._path_convert(
+                    self.sync_dir, f"run-{self.run_id}.wandb"
+                ),
+                "auto_hook": True,
+            },
+            sync_symlink_latest={
+                "value": "latest-run",
+                "hook": lambda x: self._path_convert(self.wandb_dir, x),
+            },
+            system_sample={"value": 15},
+            system_sample_seconds={"value": 2},
+            table_raise_on_max_row_limit_exceeded={
+                "value": False,
+                "preprocessor": _str_as_bool,
+            },
+            timespec={
+                "hook": lambda _: self._start_datetime,
+                "auto_hook": True,
+            },
+            tmp_dir={
+                "value": "tmp",
+                "hook": lambda x: (
+                    self._path_convert(
+                        self.wandb_dir,
+                        f"{self.run_mode}-{self.timespec}-{self.run_id}",
+                        x,
+                    )
+                    or tempfile.gettempdir()
+                ),
+            },
+            wandb_dir={
+                "hook": lambda _: _get_wandb_dir(self.root_dir or ""),
+                "auto_hook": True,
+            },
+        )
+        return props
+
+    # helper methods for validating values
+    @staticmethod
+    def _validator_factory(hint: Any) -> Callable[[Any], bool]:  # noqa: C901
+        """Return a factory for setting type validators."""
+
+        def helper(value: Any) -> bool:
+            try:
+                is_valid = is_instance_recursive(value, hint)
+            except Exception:
+                # instance check failed, but let's not crash and only print a warning
+                is_valid = False
+
+            return is_valid
+
+        return helper
+
+    @staticmethod
+    def _validate_mode(value: str) -> bool:
+        choices: Set[str] = {"dryrun", "run", "offline", "online", "disabled", "shared"}
+        if value not in choices:
+            raise UsageError(f"Settings field `mode`: {value!r} not in {choices}")
+        return True
+
+    @staticmethod
+    def _validate_project(value: Optional[str]) -> bool:
+        invalid_chars_list = list("/\\#?%:")
+        if value is not None:
+            if len(value) > 128:
+                raise UsageError(
+                    f"Invalid project name {value!r}: exceeded 128 characters"
+                )
+            invalid_chars = {char for char in invalid_chars_list if char in value}
+            if invalid_chars:
+                raise UsageError(
+                    f"Invalid project name {value!r}: "
+                    f"cannot contain characters {','.join(invalid_chars_list)!r}, "
+                    f"found {','.join(invalid_chars)!r}"
+                )
+        return True
+
+    @staticmethod
+    def _validate_start_method(value: str) -> bool:
+        available_methods = ["thread"]
+        if hasattr(multiprocessing, "get_all_start_methods"):
+            available_methods += multiprocessing.get_all_start_methods()
+        if value not in available_methods:
+            raise UsageError(
+                f"Settings field `start_method`: {value!r} not in {available_methods}"
+            )
+        return True
+
+    @staticmethod
+    def _validate_console(value: str) -> bool:
+        choices = ConsoleValue
+        if value not in choices:
+            # do not advertise internal console states
+            choices -= {"wrap_emu", "wrap_raw"}
+            raise UsageError(f"Settings field `console`: {value!r} not in {choices}")
+        return True
+
+    @staticmethod
+    def _validate_problem(value: str) -> bool:
+        choices: Set[str] = {"fatal", "warn", "silent"}
+        if value not in choices:
+            raise UsageError(f"Settings field `problem`: {value!r} not in {choices}")
+        return True
+
+    @staticmethod
+    def _validate_anonymous(value: str) -> bool:
+        choices: Set[str] = {"allow", "must", "never", "false", "true"}
+        if value not in choices:
+            raise UsageError(f"Settings field `anonymous`: {value!r} not in {choices}")
+        return True
+
+    @staticmethod
+    def _validate_run_id(value: str) -> bool:
+        # if len(value) > len(value.strip()):
+        #     raise UsageError("Run ID cannot start or end with whitespace")
+        return bool(value.strip())
+
+    @staticmethod
+    def _validate_api_key(value: str) -> bool:
+        if len(value) > len(value.strip()):
+            raise UsageError("API key cannot start or end with whitespace")
+
+        # todo: move this check to the post-init validation step
+        # if value.startswith("local") and not self.is_local:
+        #     raise UsageError(
+        #         "Attempting to use a local API key to connect to https://api.wandb.ai"
+        #     )
+        # todo: move here the logic from sdk/lib/apikey.py
+
+        return True
+
+    @staticmethod
+    def _validate_base_url(value: Optional[str]) -> bool:
+        """Validate the base url of the wandb server.
+
+        param value: URL to validate
+
+        Based on the Django URLValidator, but with a few additional checks.
+
+        Copyright (c) Django Software Foundation and individual contributors.
+        All rights reserved.
+
+        Redistribution and use in source and binary forms, with or without modification,
+        are permitted provided that the following conditions are met:
+
+            1. Redistributions of source code must retain the above copyright notice,
+               this list of conditions and the following disclaimer.
+
+            2. Redistributions in binary form must reproduce the above copyright
+               notice, this list of conditions and the following disclaimer in the
+               documentation and/or other materials provided with the distribution.
+
+            3. Neither the name of Django nor the names of its contributors may be used
+               to endorse or promote products derived from this software without
+               specific prior written permission.
+
+        THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
+        ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+        WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+        DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
+        ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+        (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+        LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
+        ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+        (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+        SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+        """
+        if value is None:
+            return True
+
+        ul = "\u00a1-\uffff"  # Unicode letters range (must not be a raw string).
+
+        # IP patterns
+        ipv4_re = (
+            r"(?:0|25[0-5]|2[0-4][0-9]|1[0-9]?[0-9]?|[1-9][0-9]?)"
+            r"(?:\.(?:0|25[0-5]|2[0-4][0-9]|1[0-9]?[0-9]?|[1-9][0-9]?)){3}"
+        )
+        ipv6_re = r"\[[0-9a-f:.]+\]"  # (simple regex, validated later)
+
+        # Host patterns
+        hostname_re = (
+            r"[a-z" + ul + r"0-9](?:[a-z" + ul + r"0-9-]{0,61}[a-z" + ul + r"0-9])?"
+        )
+        # Max length for domain name labels is 63 characters per RFC 1034 sec. 3.1
+        domain_re = r"(?:\.(?!-)[a-z" + ul + r"0-9-]{1,63}(?<!-))*"
+        tld_re = (
+            r"\."  # dot
+            r"(?!-)"  # can't start with a dash
+            r"(?:[a-z" + ul + "-]{2,63}"  # domain label
+            r"|xn--[a-z0-9]{1,59})"  # or punycode label
+            r"(?<!-)"  # can't end with a dash
+            r"\.?"  # may have a trailing dot
+        )
+        # host_re = "(" + hostname_re + domain_re + tld_re + "|localhost)"
+        # todo?: allow hostname to be just a hostname (no tld)?
+        host_re = "(" + hostname_re + domain_re + f"({tld_re})?" + "|localhost)"
+
+        regex = re.compile(
+            r"^(?:[a-z0-9.+-]*)://"  # scheme is validated separately
+            r"(?:[^\s:@/]+(?::[^\s:@/]*)?@)?"  # user:pass authentication
+            r"(?:" + ipv4_re + "|" + ipv6_re + "|" + host_re + ")"
+            r"(?::[0-9]{1,5})?"  # port
+            r"(?:[/?#][^\s]*)?"  # resource path
+            r"\Z",
+            re.IGNORECASE,
+        )
+        schemes = {"http", "https"}
+        unsafe_chars = frozenset("\t\r\n")
+
+        scheme = value.split("://")[0].lower()
+        split_url = urlsplit(value)
+        parsed_url = urlparse(value)
+
+        if re.match(r".*wandb\.ai[^\.]*$", value) and "api." not in value:
+            # user might guess app.wandb.ai or wandb.ai is the default cloud server
+            raise UsageError(
+                f"{value} is not a valid server address, did you mean https://api.wandb.ai?"
+            )
+        elif re.match(r".*wandb\.ai[^\.]*$", value) and scheme != "https":
+            raise UsageError("http is not secure, please use https://api.wandb.ai")
+        elif parsed_url.netloc == "":
+            raise UsageError(f"Invalid URL: {value}")
+        elif unsafe_chars.intersection(value):
+            raise UsageError("URL cannot contain unsafe characters")
+        elif scheme not in schemes:
+            raise UsageError("URL must start with `http(s)://`")
+        elif not regex.search(value):
+            raise UsageError(f"{value} is not a valid server address")
+        elif split_url.hostname is None or len(split_url.hostname) > 253:
+            raise UsageError("hostname is invalid")
+
+        return True
+
+    @staticmethod
+    def _validate__service_wait(value: float) -> bool:
+        if value <= 0:
+            raise UsageError("_service_wait must be a positive number")
+        return True
+
+    @staticmethod
+    def _validate__stats_sample_rate_seconds(value: float) -> bool:
+        if value < 0.1:
+            raise UsageError("_stats_sample_rate_seconds must be >= 0.1")
+        return True
+
+    @staticmethod
+    def _validate__stats_samples_to_average(value: int) -> bool:
+        if value < 1 or value > 30:
+            raise UsageError("_stats_samples_to_average must be between 1 and 30")
+        return True
+
+    @staticmethod
+    def _validate__async_upload_concurrency_limit(value: int) -> bool:
+        if value <= 0:
+            raise UsageError("_async_upload_concurrency_limit must be positive")
+
+        try:
+            import resource  # not always available on Windows
+
+            file_limit = resource.getrlimit(resource.RLIMIT_NOFILE)[0]
+        except Exception:
+            # Couldn't get the open-file-limit for some reason,
+            # probably very platform-specific. Not a problem,
+            # we just won't use it to cap the concurrency.
+            pass
+        else:
+            if value > file_limit:
+                wandb.termwarn(
+                    (
+                        "_async_upload_concurrency_limit setting of"
+                        f" {value} exceeds this process's limit"
+                        f" on open files ({file_limit}); may cause file-upload failures."
+                        " Try decreasing _async_upload_concurrency_limit,"
+                        " or increasing your file limit with `ulimit -n`."
+                    ),
+                    repeat=False,
+                )
+
+        return True
+
+    @staticmethod
+    def _validate_job_source(value: str) -> bool:
+        valid_sources = ["repo", "artifact", "image"]
+        if value not in valid_sources:
+            raise UsageError(
+                f"Settings field `job_source`: {value!r} not in {valid_sources}"
+            )
+        return True
+
+    # other helper methods
+    @staticmethod
+    def _path_convert(*args: str) -> str:
+        """Join path and apply os.path.expanduser to it."""
+        return os.path.expanduser(os.path.join(*args))
+
+    def _convert_console(self, console: str) -> str:
+        if console == "auto":
+            if (
+                self._jupyter
+                or (self.start_method == "thread")
+                or not self._disable_service
+                or self._windows
+            ):
+                console = "wrap"
+            else:
+                console = "redirect"
+        return console
+
+    def _get_colab_url(self) -> Optional[str]:
+        if not self._colab:
+            return None
+        if self._jupyter_path and self._jupyter_path.startswith("fileId="):
+            unescaped = unquote(self._jupyter_path)
+            return "https://colab.research.google.com/notebook#" + unescaped
+        return None
+
+    def _get_program(self, program: Optional[str]) -> Optional[str]:
+        if program is not None and program != "<python with no main file>":
+            return program
+
+        if not self._jupyter:
+            return program
+
+        if self.notebook_name:
+            return self.notebook_name
+
+        if not self._jupyter_path:
+            return program
+
+        if self._jupyter_path.startswith("fileId="):
+            return self._jupyter_name
+        else:
+            return self._jupyter_path
+
+    def _get_url_query_string(self) -> str:
+        # TODO(settings) use `wandb_setting` (if self.anonymous != "true":)
+        if Api().settings().get("anonymous") != "true":
+            return ""
+
+        api_key = apikey.api_key(settings=self)
+
+        return f"?{urlencode({'apiKey': api_key})}"
+
+    def _project_url_base(self) -> str:
+        if not all([self.entity, self.project]):
+            return ""
+
+        app_url = wandb.util.app_url(self.base_url)
+        return f"{app_url}/{quote(self.entity)}/{quote(self.project)}"
+
+    def _project_url(self) -> str:
+        project_url = self._project_url_base()
+        if not project_url:
+            return ""
+
+        query = self._get_url_query_string()
+
+        return f"{project_url}{query}"
+
+    def _run_url(self) -> str:
+        """Return the run url."""
+        project_url = self._project_url_base()
+        if not all([project_url, self.run_id]):
+            return ""
+
+        query = self._get_url_query_string()
+        return f"{project_url}/runs/{quote(self.run_id)}{query}"
+
+    def _set_run_start_time(self, source: int = Source.BASE) -> None:
+        """Set the time stamps for the settings.
+
+        Called once the run is initialized.
+        """
+        time_stamp: float = time.time()
+        datetime_now: datetime = datetime.fromtimestamp(time_stamp)
+        datetime_now_str = _datetime_as_str(datetime_now)
+        object.__setattr__(self, "_Settings_start_datetime", datetime_now_str)
+        object.__setattr__(self, "_Settings_start_time", time_stamp)
+        self.update(
+            _start_datetime=datetime_now_str,
+            _start_time=time_stamp,
+            source=source,
+        )
+
+    def _sweep_url(self) -> str:
+        """Return the sweep url."""
+        project_url = self._project_url_base()
+        if not all([project_url, self.sweep_id]):
+            return ""
+
+        query = self._get_url_query_string()
+        return f"{project_url}/sweeps/{quote(self.sweep_id)}{query}"
+
+    def __init__(self, **kwargs: Any) -> None:
+        self.__frozen: bool = False
+        self.__initialized: bool = False
+
+        self.__modification_order = SETTINGS_TOPOLOGICALLY_SORTED
+
+        # Set default settings values
+        # We start off with the class attributes and `default_props` dicts
+        # and then create Property objects.
+        # Once initialized, attributes are to only be updated using the `update` method
+        default_props = self._default_props()
+
+        # Init instance attributes as Property objects.
+        # Type hints of class attributes are used to generate a type validator function
+        # for runtime checks for each attribute.
+        # These are defaults, using Source.BASE for non-policy attributes and Source.RUN for policies.
+        for prop, type_hint in get_type_hints(SettingsData).items():
+            validators = [self._validator_factory(type_hint)]
+
+            if prop in default_props:
+                validator = default_props[prop].pop("validator", [])
+                # Property validator could be either Callable or Sequence[Callable]
+                if callable(validator):
+                    validators.append(validator)
+                elif isinstance(validator, Sequence):
+                    validators.extend(list(validator))
+                object.__setattr__(
+                    self,
+                    prop,
+                    Property(
+                        name=prop,
+                        **default_props[prop],
+                        validator=validators,
+                        # todo: double-check this logic:
+                        source=Source.RUN
+                        if default_props[prop].get("is_policy", False)
+                        else Source.BASE,
+                    ),
+                )
+            else:
+                object.__setattr__(
+                    self,
+                    prop,
+                    Property(
+                        name=prop,
+                        validator=validators,
+                        source=Source.BASE,
+                    ),
+                )
+
+        # update overridden defaults from kwargs
+        unexpected_arguments = [k for k in kwargs.keys() if k not in self.__dict__]
+        # allow only explicitly defined arguments
+        if unexpected_arguments:
+            raise SettingsUnexpectedArgsError(
+                f"Got unexpected arguments: {unexpected_arguments}. "
+            )
+
+        # automatically inspect setting validators and runtime hooks and topologically sort them
+        # so that we can safely update them. throw error if there are cycles.
+        for prop in self.__modification_order:
+            if prop in kwargs:
+                source = Source.RUN if self.__dict__[prop].is_policy else Source.BASE
+                self.update({prop: kwargs[prop]}, source=source)
+                kwargs.pop(prop)
+
+        for k, v in kwargs.items():
+            # todo: double-check this logic:
+            source = Source.RUN if self.__dict__[k].is_policy else Source.BASE
+            self.update({k: v}, source=source)
+
+        # setup private attributes
+        object.__setattr__(self, "_Settings_start_datetime", None)
+        object.__setattr__(self, "_Settings_start_time", None)
+
+        # done with init, use self.update() to update attributes from now on
+        self.__initialized = True
+
+        # todo? freeze settings to prevent accidental changes
+        # self.freeze()
+
+    def __str__(self) -> str:
+        # get attributes that are instances of the Property class:
+        representation = {
+            k: v.value for k, v in self.__dict__.items() if isinstance(v, Property)
+        }
+        return f"<Settings {_redact_dict(representation)}>"
+
+    def __repr__(self) -> str:
+        # private attributes
+        private = {k: v for k, v in self.__dict__.items() if k.startswith("_Settings")}
+        # get attributes that are instances of the Property class:
+        attributes = {
+            k: f"<Property value={v.value} source={v.source}>"
+            for k, v in self.__dict__.items()
+            if isinstance(v, Property)
+        }
+        representation = {**private, **attributes}
+        return f"<Settings {representation}>"
+
+    def __copy__(self) -> "Settings":
+        """Ensure that a copy of the settings object is a truly deep copy.
+
+        Note that the copied object will not be frozen  todo? why is this needed?
+        """
+        # get attributes that are instances of the Property class:
+        attributes = {k: v for k, v in self.__dict__.items() if isinstance(v, Property)}
+        new = Settings()
+        # update properties that have deps or are dependent on in the topologically-sorted order
+        for prop in self.__modification_order:
+            new.update({prop: attributes[prop]._value}, source=attributes[prop].source)
+            attributes.pop(prop)
+
+        # update the remaining attributes
+        for k, v in attributes.items():
+            # make sure to use the raw property value (v._value),
+            # not the potential result of runtime hooks applied to it (v.value)
+            new.update({k: v._value}, source=v.source)
+        new.unfreeze()
+
+        return new
+
+    def __deepcopy__(self, memo: dict) -> "Settings":
+        return self.__copy__()
+
+    # attribute access methods
+    @no_type_check  # this is a hack to make mypy happy
+    def __getattribute__(self, name: str) -> Any:
+        """Expose `attribute.value` if `attribute` is a Property."""
+        item = object.__getattribute__(self, name)
+        if isinstance(item, Property):
+            return item.value
+        return item
+
+    def __setattr__(self, key: str, value: Any) -> None:
+        if "_Settings__initialized" in self.__dict__ and self.__initialized:
+            raise TypeError(f"Please use update() to update attribute `{key}` value")
+        object.__setattr__(self, key, value)
+
+    def __iter__(self) -> Iterable:
+        return iter(self.to_dict())
+
+    def copy(self) -> "Settings":
+        return self.__copy__()
+
+    # implement the Mapping interface
+    def keys(self) -> Iterable[str]:
+        return self.to_dict().keys()
+
+    @no_type_check  # this is a hack to make mypy happy
+    def __getitem__(self, name: str) -> Any:
+        """Expose attribute.value if attribute is a Property."""
+        item = object.__getattribute__(self, name)
+        if isinstance(item, Property):
+            return item.value
+        return item
+
+    def update(
+        self,
+        settings: Optional[Union[Dict[str, Any], "Settings"]] = None,
+        source: int = Source.OVERRIDE,
+        **kwargs: Any,
+    ) -> None:
+        """Update individual settings."""
+        if "_Settings__frozen" in self.__dict__ and self.__frozen:
+            raise TypeError("Settings object is frozen")
+
+        if isinstance(settings, Settings):
+            # If a Settings object is passed, detect the settings that differ
+            # from defaults, collect them into a dict, and apply them using `source`.
+            # This comes up in `wandb.init(settings=wandb.Settings(...))` and
+            # seems like the behavior that the user would expect when calling init that way.
+            defaults = Settings()
+            settings_dict = dict()
+            for k, v in settings.__dict__.items():
+                if isinstance(v, Property):
+                    if v._value != defaults.__dict__[k]._value:
+                        settings_dict[k] = v._value
+            # replace with the generated dict
+            settings = settings_dict
+
+        # add kwargs to settings
+        settings = settings or dict()
+        # explicit kwargs take precedence over settings
+        settings = {**settings, **kwargs}
+        unknown_properties = []
+        for key in settings.keys():
+            # only allow updating known Properties
+            if key not in self.__dict__ or not isinstance(self.__dict__[key], Property):
+                unknown_properties.append(key)
+        if unknown_properties:
+            raise KeyError(f"Unknown settings: {unknown_properties}")
+        # only if all keys are valid, update them
+
+        # store settings to be updated in a dict to preserve stats on preprocessing and validation errors
+        settings.copy()
+
+        # update properties that have deps or are dependent on in the topologically-sorted order
+        for key in self.__modification_order:
+            if key in settings:
+                self.__dict__[key].update(settings.pop(key), source=source)
+
+        # update the remaining properties
+        for key, value in settings.items():
+            self.__dict__[key].update(value, source)
+
+    def items(self) -> ItemsView[str, Any]:
+        return self.to_dict().items()
+
+    def get(self, key: str, default: Optional[Any] = None) -> Any:
+        return self.to_dict().get(key, default)
+
+    def freeze(self) -> None:
+        object.__setattr__(self, "_Settings__frozen", True)
+
+    def unfreeze(self) -> None:
+        object.__setattr__(self, "_Settings__frozen", False)
+
+    def is_frozen(self) -> bool:
+        return self.__frozen
+
+    def to_dict(self) -> Dict[str, Any]:
+        """Return a dict representation of the settings."""
+        # get attributes that are instances of the Property class:
+        attributes = {
+            k: v.value for k, v in self.__dict__.items() if isinstance(v, Property)
+        }
+        return attributes
+
+    def to_proto(self) -> wandb_settings_pb2.Settings:
+        """Generate a protobuf representation of the settings."""
+        from dataclasses import fields
+
+        settings = wandb_settings_pb2.Settings()
+        for field in fields(SettingsData):
+            k = field.name
+            v = getattr(self, k)
+            # special case for _stats_open_metrics_filters
+            if k == "_stats_open_metrics_filters":
+                if isinstance(v, (list, set, tuple)):
+                    setting = getattr(settings, k)
+                    setting.sequence.value.extend(v)
+                elif isinstance(v, dict):
+                    setting = getattr(settings, k)
+                    for key, value in v.items():
+                        for kk, vv in value.items():
+                            setting.mapping.value[key].value[kk] = vv
+                else:
+                    raise TypeError(f"Unsupported type {type(v)} for setting {k}")
+                continue
+
+            if isinstance(v, bool):
+                getattr(settings, k).CopyFrom(BoolValue(value=v))
+            elif isinstance(v, int):
+                getattr(settings, k).CopyFrom(Int32Value(value=v))
+            elif isinstance(v, float):
+                getattr(settings, k).CopyFrom(DoubleValue(value=v))
+            elif isinstance(v, str):
+                getattr(settings, k).CopyFrom(StringValue(value=v))
+            elif isinstance(v, (list, set, tuple)):
+                # we only support sequences of strings for now
+                sequence = getattr(settings, k)
+                sequence.value.extend(v)
+            elif isinstance(v, dict):
+                mapping = getattr(settings, k)
+                for key, value in v.items():
+                    # we only support dicts with string values for now
+                    mapping.value[key] = value
+            elif isinstance(v, RunMoment):
+                getattr(settings, k).CopyFrom(
+                    wandb_settings_pb2.RunMoment(
+                        run=v.run,
+                        value=v.value,
+                        metric=v.metric,
+                    )
+                )
+            elif v is None:
+                # None is the default value for all settings, so we don't need to set it,
+                # i.e. None means that the value was not set.
+                pass
+            else:
+                raise TypeError(f"Unsupported type {type(v)} for setting {k}")
+        # TODO: store property sources in the protobuf so that we can reconstruct the
+        #  settings object from the protobuf
+        return settings
+
+    # apply settings from different sources
+    # TODO(dd): think about doing some|all of that at init
+    def _apply_settings(
+        self,
+        settings: "Settings",
+        _logger: Optional[_EarlyLogger] = None,
+    ) -> None:
+        """Apply settings from a Settings object."""
+        if _logger is not None:
+            _logger.info(f"Applying settings from {settings}")
+        attributes = {
+            k: v for k, v in settings.__dict__.items() if isinstance(v, Property)
+        }
+        # update properties that have deps or are dependent on in the topologically-sorted order
+        for prop in self.__modification_order:
+            self.update({prop: attributes[prop]._value}, source=attributes[prop].source)
+            attributes.pop(prop)
+        # update the remaining properties
+        for k, v in attributes.items():
+            # note that only the same/higher priority settings are propagated
+            self.update({k: v._value}, source=v.source)
+
+    @staticmethod
+    def _load_config_file(file_name: str, section: str = "default") -> dict:
+        parser = configparser.ConfigParser()
+        parser.add_section(section)
+        parser.read(file_name)
+        config: Dict[str, Any] = dict()
+        for k in parser[section]:
+            config[k] = parser[section][k]
+            # TODO (cvp): we didn't do this in the old cli, but it seems necessary
+            if k == "ignore_globs":
+                config[k] = config[k].split(",")
+        return config
+
+    def _apply_base(self, pid: int, _logger: Optional[_EarlyLogger] = None) -> None:
+        if _logger is not None:
+            _logger.info(f"Current SDK version is {wandb.__version__}")
+            _logger.info(f"Configure stats pid to {pid}")
+        self.update({"_stats_pid": pid}, source=Source.SETUP)
+
+    def _apply_config_files(self, _logger: Optional[_EarlyLogger] = None) -> None:
+        # TODO(jhr): permit setting of config in system and workspace
+        if self.settings_system is not None:
+            if _logger is not None:
+                _logger.info(f"Loading settings from {self.settings_system}")
+            self.update(
+                self._load_config_file(self.settings_system),
+                source=Source.SYSTEM,
+            )
+        if self.settings_workspace is not None:
+            if _logger is not None:
+                _logger.info(f"Loading settings from {self.settings_workspace}")
+            self.update(
+                self._load_config_file(self.settings_workspace),
+                source=Source.WORKSPACE,
+            )
+
+    def _apply_env_vars(
+        self,
+        environ: Mapping[str, Any],
+        _logger: Optional[_EarlyLogger] = None,
+    ) -> None:
+        env_prefix: str = "WANDB_"
+        special_env_var_names = {
+            "WANDB_TRACELOG": "_tracelog",
+            "WANDB_DISABLE_SERVICE": "_disable_service",
+            "WANDB_SERVICE_TRANSPORT": "_service_transport",
+            "WANDB_DIR": "root_dir",
+            "WANDB_NAME": "run_name",
+            "WANDB_NOTES": "run_notes",
+            "WANDB_TAGS": "run_tags",
+            "WANDB_JOB_TYPE": "run_job_type",
+            "WANDB_HTTP_TIMEOUT": "_graphql_timeout_seconds",
+            "WANDB_FILE_PUSHER_TIMEOUT": "_file_transfer_timeout_seconds",
+            "WANDB_USER_EMAIL": "email",
+        }
+        env = dict()
+        for setting, value in environ.items():
+            if not setting.startswith(env_prefix):
+                continue
+
+            if setting in special_env_var_names:
+                key = special_env_var_names[setting]
+            else:
+                # otherwise, strip the prefix and convert to lowercase
+                key = setting[len(env_prefix) :].lower()
+
+            if key in self.__dict__:
+                if key in ("ignore_globs", "run_tags"):
+                    value = value.split(",")
+                env[key] = value
+            elif _logger is not None:
+                _logger.warning(f"Unknown environment variable: {setting}")
+
+        if _logger is not None:
+            _logger.info(
+                f"Loading settings from environment variables: {_redact_dict(env)}"
+            )
+        self.update(env, source=Source.ENV)
+
+    def _infer_settings_from_environment(
+        self, _logger: Optional[_EarlyLogger] = None
+    ) -> None:
+        """Modify settings based on environment (for runs and cli)."""
+        settings: Dict[str, Union[bool, str, Sequence, None]] = dict()
+        # disable symlinks if on windows (requires admin or developer setup)
+        settings["symlink"] = True
+        if self._windows:
+            settings["symlink"] = False
+
+        # TODO(jhr): this needs to be moved last in setting up settings ?
+        #  (dd): loading order does not matter as long as source is set correctly
+
+        # For code saving, only allow env var override if value from server is true, or
+        # if no preference was specified.
+        if (self.save_code is True or self.save_code is None) and (
+            os.getenv(wandb.env.SAVE_CODE) is not None
+            or os.getenv(wandb.env.DISABLE_CODE) is not None
+        ):
+            settings["save_code"] = wandb.env.should_save_code()
+
+        settings["disable_git"] = wandb.env.disable_git()
+
+        # Attempt to get notebook information if not already set by the user
+        if self._jupyter and (self.notebook_name is None or self.notebook_name == ""):
+            meta = wandb.jupyter.notebook_metadata(self.silent)
+            settings["_jupyter_path"] = meta.get("path")
+            settings["_jupyter_name"] = meta.get("name")
+            settings["_jupyter_root"] = meta.get("root")
+        elif (
+            self._jupyter
+            and self.notebook_name is not None
+            and os.path.exists(self.notebook_name)
+        ):
+            settings["_jupyter_path"] = self.notebook_name
+            settings["_jupyter_name"] = self.notebook_name
+            settings["_jupyter_root"] = os.getcwd()
+        elif self._jupyter:
+            wandb.termwarn(
+                "WANDB_NOTEBOOK_NAME should be a path to a notebook file, "
+                f"couldn't find {self.notebook_name}.",
+            )
+
+        # host and username are populated by apply_env_vars if corresponding env
+        # vars exist -- but if they don't, we'll fill them in here
+        if self.host is None:
+            settings["host"] = socket.gethostname()  # type: ignore
+
+        if self.username is None:
+            try:  # type: ignore
+                settings["username"] = getpass.getuser()
+            except KeyError:
+                # getuser() could raise KeyError in restricted environments like
+                # chroot jails or docker containers. Return user id in these cases.
+                settings["username"] = str(os.getuid())
+
+        _executable = (
+            self._executable
+            or os.environ.get(wandb.env._EXECUTABLE)
+            or sys.executable
+            or shutil.which("python3")
+            or "python3"
+        )
+        settings["_executable"] = _executable
+
+        settings["docker"] = wandb.env.get_docker(wandb.util.image_id_from_k8s())
+
+        # TODO: we should use the cuda library to collect this
+        if os.path.exists("/usr/local/cuda/version.txt"):
+            with open("/usr/local/cuda/version.txt") as f:
+                settings["_cuda"] = f.read().split(" ")[-1].strip()
+        if not self._jupyter:
+            settings["_args"] = sys.argv[1:]
+        settings["_os"] = platform.platform(aliased=True)
+        settings["_python"] = platform.python_version()
+        # hack to make sure we don't hang on windows
+        if self._windows and self._except_exit is None:
+            settings["_except_exit"] = True  # type: ignore
+
+        if _logger is not None:
+            _logger.info(
+                f"Inferring settings from compute environment: {_redact_dict(settings)}"
+            )
+
+        self.update(settings, source=Source.ENV)
+
+    def _infer_run_settings_from_environment(
+        self,
+        _logger: Optional[_EarlyLogger] = None,
+    ) -> None:
+        """Modify settings based on environment (for runs only)."""
+        # If there's not already a program file, infer it now.
+        settings: Dict[str, Union[bool, str, None]] = dict()
+        program = self.program or _get_program()
+        if program is not None:
+            repo = GitRepo()
+            root = repo.root or os.getcwd()
+
+            program_relpath = self.program_relpath or _get_program_relpath(
+                program, repo.root, _logger=_logger
+            )
+            settings["program_relpath"] = program_relpath
+            program_abspath = os.path.abspath(
+                os.path.join(root, os.path.relpath(os.getcwd(), root), program)
+            )
+            if os.path.exists(program_abspath):
+                settings["program_abspath"] = program_abspath
+        else:
+            program = "<python with no main file>"
+
+        settings["program"] = program
+
+        if _logger is not None:
+            _logger.info(
+                f"Inferring run settings from compute environment: {_redact_dict(settings)}"
+            )
+
+        self.update(settings, source=Source.ENV)
+
+    def _apply_setup(
+        self, setup_settings: Dict[str, Any], _logger: Optional[_EarlyLogger] = None
+    ) -> None:
+        if _logger:
+            _logger.info(f"Applying setup settings: {_redact_dict(setup_settings)}")
+        self.update(setup_settings, source=Source.SETUP)
+
+    def _apply_user(
+        self, user_settings: Dict[str, Any], _logger: Optional[_EarlyLogger] = None
+    ) -> None:
+        if _logger:
+            _logger.info(f"Applying user settings: {_redact_dict(user_settings)}")
+        self.update(user_settings, source=Source.USER)
+
+    def _apply_init(self, init_settings: Dict[str, Union[str, int, None]]) -> None:
+        # pop magic from init settings
+        init_settings.pop("magic", None)
+
+        # prevent setting project, entity if in sweep
+        # TODO(jhr): these should be locked elements in the future
+        if self.sweep_id:
+            for key in ("project", "entity", "id"):
+                val = init_settings.pop(key, None)
+                if val:
+                    wandb.termwarn(
+                        f"Ignored wandb.init() arg {key} when running a sweep."
+                    )
+        if self.launch:
+            if self.project is not None and init_settings.pop("project", None):
+                wandb.termwarn(
+                    "Project is ignored when running from wandb launch context. "
+                    "Ignored wandb.init() arg project when running running from launch.",
+                )
+            for key in ("entity", "id"):
+                # Init settings cannot override launch settings.
+                if init_settings.pop(key, None):
+                    wandb.termwarn(
+                        "Project, entity and id are ignored when running from wandb launch context. "
+                        f"Ignored wandb.init() arg {key} when running running from launch.",
+                    )
+
+        # strip out items where value is None
+        param_map = dict(
+            name="run_name",
+            id="run_id",
+            tags="run_tags",
+            group="run_group",
+            job_type="run_job_type",
+            notes="run_notes",
+            dir="root_dir",
+            sweep_id="sweep_id",
+        )
+        init_settings = {
+            param_map.get(k, k): v for k, v in init_settings.items() if v is not None
+        }
+        # fun logic to convert the resume init arg
+        if init_settings.get("resume"):
+            if isinstance(init_settings["resume"], str):
+                if init_settings["resume"] not in ("allow", "must", "never", "auto"):
+                    if init_settings.get("run_id") is None:
+                        #  TODO: deprecate or don't support
+                        init_settings["run_id"] = init_settings["resume"]
+                    init_settings["resume"] = "allow"
+            elif init_settings["resume"] is True:
+                # todo: add deprecation warning, switch to literal strings for resume
+                init_settings["resume"] = "auto"
+
+        # update settings
+        self.update(init_settings, source=Source.INIT)
+
+        # handle auto resume logic
+        if self.resume == "auto":
+            if os.path.exists(self.resume_fname):
+                with open(self.resume_fname) as f:
+                    resume_run_id = json.load(f)["run_id"]
+                if self.run_id is None:
+                    self.update({"run_id": resume_run_id}, source=Source.INIT)  # type: ignore
+                elif self.run_id != resume_run_id:
+                    wandb.termwarn(
+                        "Tried to auto resume run with "
+                        f"id {resume_run_id} but id {self.run_id} is set.",
+                    )
+        self.update({"run_id": self.run_id or generate_id()}, source=Source.INIT)
+        # persist our run id in case of failure
+        # check None for mypy
+        if self.resume == "auto" and self.resume_fname is not None:
+            filesystem.mkdir_exists_ok(self.wandb_dir)
+            with open(self.resume_fname, "w") as f:
+                f.write(json.dumps({"run_id": self.run_id}))
+
+    def _apply_login(
+        self,
+        login_settings: Dict[str, Any],
+        _logger: Optional[_EarlyLogger] = None,
+    ) -> None:
+        key_map = {
+            "key": "api_key",
+            "host": "base_url",
+            "timeout": "login_timeout",
+        }
+
+        # Rename keys and keep only the non-None values.
+        #
+        # The input keys are parameters to wandb.login(), but we use different
+        # names for some of them in Settings.
+        login_settings = {
+            key_map.get(key, key): value
+            for key, value in login_settings.items()
+            if value is not None
+        }
+
+        if _logger:
+            _logger.info(f"Applying login settings: {_redact_dict(login_settings)}")
+
+        self.update(
+            login_settings,
+            source=Source.LOGIN,
+        )
+
+    def _apply_run_start(self, run_start_settings: Dict[str, Any]) -> None:
+        # This dictionary maps from the "run message dict" to relevant fields in settings
+        # Note: that config is missing
+        param_map = {
+            "run_id": "run_id",
+            "entity": "entity",
+            "project": "project",
+            "run_group": "run_group",
+            "job_type": "run_job_type",
+            "display_name": "run_name",
+            "notes": "run_notes",
+            "tags": "run_tags",
+            "sweep_id": "sweep_id",
+            "host": "host",
+            "resumed": "resumed",
+            "git.remote_url": "git_remote_url",
+            "git.commit": "git_commit",
+        }
+        run_settings = {
+            name: reduce(lambda d, k: d.get(k, {}), attr.split("."), run_start_settings)
+            for attr, name in param_map.items()
+        }
+        run_settings = {key: value for key, value in run_settings.items() if value}
+        if run_settings:
+            self.update(run_settings, source=Source.RUN)