wandb 0.17.0rc2__py3-none-win_amd64.whl → 0.17.1__py3-none-win_amd64.whl

wandb/sdk/wandb_init.py

@@ -15,7 +15,6 @@ import os
 import platform
 import sys
 import tempfile
-import traceback
 from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Union
 
 import wandb
@@ -555,7 +554,7 @@ class _WandbInit:
         percent_done = handle.percent_done
         self.printer.progress_update(line, percent_done=percent_done)
 
-    def init(self) -> Union[Run, RunDisabled, None]:  # noqa: C901
+    def init(self) -> Union[Run, RunDisabled]:  # noqa: C901
         if logger is None:
             raise RuntimeError("Logger not initialized")
         logger.info("calling init triggers")
@@ -654,9 +653,6 @@ class _WandbInit:
             if self.settings.launch:
                 tel.feature.launch = True
 
-            if self.settings._async_upload_concurrency_limit:
-                tel.feature.async_uploads = True
-
             for module_name in telemetry.list_telemetry_imports(only_imported=True):
                 setattr(tel.imports_init, module_name, True)
 
@@ -843,13 +839,6 @@ class _WandbInit:
         return run
 
 
-def getcaller() -> None:
-    if not logger:
-        return None
-    src, line, func, stack = logger.findCaller(stack_info=True)
-    print("Problem at:", src, line, func)
-
-
 def _attach(
     attach_id: Optional[str] = None,
     run_id: Optional[str] = None,
@@ -957,8 +946,9 @@ def init(
     save_code: Optional[bool] = None,
     id: Optional[str] = None,
     fork_from: Optional[str] = None,
+    resume_from: Optional[str] = None,
     settings: Union[Settings, Dict[str, Any], None] = None,
-) -> Union[Run, RunDisabled, None]:
+) -> Union[Run, RunDisabled]:
     r"""Start a new run to track and log to W&B.
 
     In an ML training pipeline, you could add `wandb.init()`
@@ -1006,7 +996,7 @@ def init(
             there, so make sure to create your account or team in the UI before
             starting to log runs.
             If you don't specify an entity, the run will be sent to your default
-            entity, which is usually your username. Change your default entity
+            entity. Change your default entity
             in [your settings](https://wandb.ai/settings) under "default location
             to create new projects".
         config: (dict, argparse, absl.flags, str, optional)
@@ -1164,57 +1154,33 @@ def init(
     wandb._assert_is_user_process()
 
     kwargs = dict(locals())
-    error_seen = None
-    except_exit = None
-    run: Optional[Union[Run, RunDisabled]] = None
 
-    # convert fork_from into a version that can be passed to settings
-    if fork_from is not None and resume is not None:
-        raise ValueError("Cannot specify both `fork_from` and `resume`")
+    num_resume_options_set = (
+        (fork_from is not None)  # wrap
+        + (resume is not None)
+        + (resume_from is not None)
+    )
+    if num_resume_options_set > 1:
+        raise ValueError(
+            "You cannot specify more than one of `fork_from`, `resume`, or `resume_from`"
+        )
 
     try:
         wi = _WandbInit()
         wi.setup(kwargs)
-        assert wi.settings
-        except_exit = wi.settings._except_exit
-        try:
-            run = wi.init()
-            except_exit = wi.settings._except_exit
-        except (KeyboardInterrupt, Exception) as e:
-            if not isinstance(e, KeyboardInterrupt):
-                wandb._sentry.exception(e)
-            if not (
-                wandb.wandb_agent._is_running() and isinstance(e, KeyboardInterrupt)
-            ):
-                getcaller()
-            assert logger
-            if wi.settings.problem == "fatal":
-                raise
-            if wi.settings.problem == "warn":
-                pass
-            # TODO(jhr): figure out how to make this RunDummy
-            run = None
-    except Error as e:
-        if logger is not None:
-            logger.exception(str(e))
-        raise e
+        return wi.init()
+
     except KeyboardInterrupt as e:
-        assert logger
-        logger.warning("interrupted", exc_info=e)
-        raise e
+        if logger is not None:
+            logger.warning("interrupted", exc_info=e)
+
+        raise
+
     except Exception as e:
-        error_seen = e
-        traceback.print_exc()
-        assert logger
-        logger.error("error", exc_info=e)
+        if logger is not None:
+            logger.exception("error in wandb.init()", exc_info=e)
+
         # Need to build delay into this sentry capture because our exit hooks
         # mess with sentry's ability to send out errors before the program ends.
-        wandb._sentry.exception(e)
-        # reraise(*sys.exc_info())
-    finally:
-        if error_seen:
-            if except_exit:
-                wandb.termerror("Abnormal program exit")
-                os._exit(1)
-            raise Error("An unexpected error occurred") from error_seen
-    return run
+        wandb._sentry.reraise(e)
+        raise AssertionError()  # unreachable

wandb/sdk/wandb_login.py

@@ -6,7 +6,7 @@ This authenticates your machine to log data to your account.
 import enum
 import os
 import sys
-from typing import Dict, Optional, Tuple
+from typing import Optional, Tuple
 
 if sys.version_info >= (3, 8):
     from typing import Literal
@@ -112,11 +112,10 @@ class ApiKeyStatus(enum.Enum):
 
 class _WandbLogin:
     def __init__(self):
-        self.kwargs: Optional[Dict] = None
         self._settings: Optional[Settings] = None
         self._backend = None
-        self._silent = None
-        self._entity = None
+        self._silent: Optional[bool] = None
+        self._entity: Optional[str] = None
         self._wl = None
         self._key = None
         self._relogin = None
@@ -130,7 +129,8 @@ class _WandbLogin:
         host: Optional[str] = None,
         force: Optional[bool] = None,
         timeout: Optional[int] = None,
-    ):
+    ) -> None:
+        """Updates login-related settings on the global setup object."""
         self._relogin = relogin
 
         # built up login settings
@@ -152,19 +152,24 @@ class _WandbLogin:
         self._wl = wandb.setup(settings=login_settings)
         self._settings = self._wl.settings
 
-    def is_apikey_configured(self):
+    def is_apikey_configured(self) -> bool:
+        """Returns whether an API key is set or can be inferred."""
         return apikey.api_key(settings=self._settings) is not None
 
     def set_backend(self, backend):
         self._backend = backend
 
-    def set_silent(self, silent: bool):
+    def set_silent(self, silent: bool) -> None:
         self._silent = silent
 
-    def set_entity(self, entity: str):
+    def set_entity(self, entity: str) -> None:
         self._entity = entity
 
-    def login(self):
+    def login(self) -> bool:
+        """Returns whether the user is logged in (i.e. an API key exists).
+
+        If the user is logged in, this also prints an informational message.
+        """
         apikey_configured = self.is_apikey_configured()
         if self._settings.relogin or self._relogin:
             apikey_configured = False
@@ -172,11 +177,12 @@ class _WandbLogin:
             return False
 
         if not self._silent:
-            self.login_display()
+            self._print_logged_in_message()
 
         return apikey_configured
 
-    def login_display(self):
+    def _print_logged_in_message(self) -> None:
+        """Prints a message telling the user they are logged in."""
         username = self._wl._get_username()
 
         if username:
@@ -200,7 +206,8 @@ class _WandbLogin:
             repeat=False,
         )
 
-    def configure_api_key(self, key):
+    def configure_api_key(self, key: str) -> None:
+        """Saves the API key and updates the the global setup object."""
         if self._settings._notebook and not self._settings.silent:
             wandb.termwarn(
                 "If you're specifying your api key in code, ensure this "
@@ -213,8 +220,14 @@ class _WandbLogin:
         self._key = key
 
     def update_session(
-        self, key: Optional[str], status: ApiKeyStatus = ApiKeyStatus.VALID
+        self,
+        key: Optional[str],
+        status: ApiKeyStatus = ApiKeyStatus.VALID,
     ) -> None:
+        """Updates mode and API key settings on the global setup object.
+
+        If we're online, this also pulls in user settings from the server.
+        """
         _logger = wandb.setup()._get_logger()
         login_settings = dict()
         if status == ApiKeyStatus.OFFLINE:
@@ -252,7 +265,8 @@ class _WandbLogin:
                 return None, ApiKeyStatus.OFFLINE
             return key, ApiKeyStatus.VALID
 
-    def prompt_api_key(self):
+    def prompt_api_key(self) -> None:
+        """Updates the global API key by prompting the user."""
         key, status = self._prompt_api_key()
         if status == ApiKeyStatus.NOTTY:
             directive = (
@@ -265,14 +279,6 @@ class _WandbLogin:
         self.update_session(key, status=status)
         self._key = key
 
-    def propogate_login(self):
-        # TODO(jhr): figure out if this is really necessary
-        if self._backend:
-            # TODO: calling this twice is gross, this deserves a refactor
-            # Make sure our backend picks up the new creds
-            # _ = self._backend.interface.communicate_login(key, anonymous)
-            pass
-
 
 def _login(
     *,
@@ -333,7 +339,4 @@ def _login(
     if not key:
         wlogin.prompt_api_key()
 
-    # make sure login credentials get to the backend
-    wlogin.propogate_login()
-
     return wlogin._key or False

wandb/sdk/wandb_run.py

@@ -207,6 +207,15 @@ class RunStatusChecker:
         self._network_status_thread.start()
         self._internal_messages_thread.start()
 
+    @staticmethod
+    def _abandon_status_check(
+        lock: threading.Lock,
+        handle: Optional[MailboxHandle],
+    ):
+        with lock:
+            if handle:
+                handle.abandon()
+
     def _loop_check_status(
         self,
         *,
@@ -247,7 +256,7 @@ class RunStatusChecker:
                 local_handle = None
 
             time_elapsed = time.monotonic() - time_probe
-            wait_time = max(self._stop_polling_interval - time_elapsed, 0)
+            wait_time = max(timeout - time_elapsed, 0)
             join_requested = self._join_event.wait(timeout=wait_time)
 
     def check_network_status(self) -> None:
@@ -265,13 +274,19 @@ class RunStatusChecker:
                         )
                     )
 
-        self._loop_check_status(
-            lock=self._network_status_lock,
-            set_handle=lambda x: setattr(self, "_network_status_handle", x),
-            timeout=self._retry_polling_interval,
-            request=self._interface.deliver_network_status,
-            process=_process_network_status,
-        )
+        try:
+            self._loop_check_status(
+                lock=self._network_status_lock,
+                set_handle=lambda x: setattr(self, "_network_status_handle", x),
+                timeout=self._retry_polling_interval,
+                request=self._interface.deliver_network_status,
+                process=_process_network_status,
+            )
+        except BrokenPipeError:
+            self._abandon_status_check(
+                self._network_status_lock,
+                self._network_status_handle,
+            )
 
     def check_stop_status(self) -> None:
         def _process_stop_status(result: Result) -> None:
@@ -283,13 +298,19 @@ class RunStatusChecker:
                     thread.interrupt_main()
                     return
 
-        self._loop_check_status(
-            lock=self._stop_status_lock,
-            set_handle=lambda x: setattr(self, "_stop_status_handle", x),
-            timeout=self._stop_polling_interval,
-            request=self._interface.deliver_stop_status,
-            process=_process_stop_status,
-        )
+        try:
+            self._loop_check_status(
+                lock=self._stop_status_lock,
+                set_handle=lambda x: setattr(self, "_stop_status_handle", x),
+                timeout=self._stop_polling_interval,
+                request=self._interface.deliver_stop_status,
+                process=_process_stop_status,
+            )
+        except BrokenPipeError:
+            self._abandon_status_check(
+                self._stop_status_lock,
+                self._stop_status_handle,
+            )
 
     def check_internal_messages(self) -> None:
         def _process_internal_messages(result: Result) -> None:
@@ -297,25 +318,34 @@ class RunStatusChecker:
             for msg in internal_messages.messages.warning:
                 wandb.termwarn(msg)
 
-        self._loop_check_status(
-            lock=self._internal_messages_lock,
-            set_handle=lambda x: setattr(self, "_internal_messages_handle", x),
-            timeout=1,
-            request=self._interface.deliver_internal_messages,
-            process=_process_internal_messages,
-        )
+        try:
+            self._loop_check_status(
+                lock=self._internal_messages_lock,
+                set_handle=lambda x: setattr(self, "_internal_messages_handle", x),
+                timeout=1,
+                request=self._interface.deliver_internal_messages,
+                process=_process_internal_messages,
+            )
+        except BrokenPipeError:
+            self._abandon_status_check(
+                self._internal_messages_lock,
+                self._internal_messages_handle,
+            )
 
     def stop(self) -> None:
         self._join_event.set()
-        with self._stop_status_lock:
-            if self._stop_status_handle:
-                self._stop_status_handle.abandon()
-        with self._network_status_lock:
-            if self._network_status_handle:
-                self._network_status_handle.abandon()
-        with self._internal_messages_lock:
-            if self._internal_messages_handle:
-                self._internal_messages_handle.abandon()
+        self._abandon_status_check(
+            self._stop_status_lock,
+            self._stop_status_handle,
+        )
+        self._abandon_status_check(
+            self._network_status_lock,
+            self._network_status_handle,
+        )
+        self._abandon_status_check(
+            self._internal_messages_lock,
+            self._internal_messages_handle,
+        )
 
     def join(self) -> None:
         self.stop()
@@ -676,6 +706,12 @@ class Run:
                 "step": self._settings.fork_from.value,
             }
 
+        if self._settings.resume_from is not None:
+            config[wandb_key]["branch_point"] = {
+                "run_id": self._settings.resume_from.run,
+                "step": self._settings.resume_from.value,
+            }
+
         self._config._update(config, ignore_locked=True)
 
         if sweep_config:
@@ -1400,6 +1436,10 @@ class Run:
         # self._printer.display(line)
 
     def _summary_get_current_summary_callback(self) -> Dict[str, Any]:
+        if self._is_finished:
+            # TODO: WB-18420: fetch summary from backend and stage it before run is finished
+            wandb.termwarn("Summary data not available in finished run")
+            return {}
         if not self._backend or not self._backend.interface:
             return {}
         handle = self._backend.interface.deliver_get_summary()
@@ -1792,7 +1832,7 @@ class Run:
             import wandb
 
             run = wandb.init()
-            run.log({"pr": wandb.plots.precision_recall(y_test, y_probas, labels)})
+            run.log({"pr": wandb.plot.pr_curve(y_test, y_probas, labels)})
             ```
 
             ### 3D Object
@@ -2443,6 +2483,8 @@ class Run:
         self._telemetry_obj_active = True
         self._telemetry_flush()
 
+        self._detect_and_apply_job_inputs()
+
         # object is about to be returned to the user, don't let them modify it
         self._freeze()
 
@@ -2450,6 +2492,12 @@ class Run:
             if os.path.exists(self._settings.resume_fname):
                 os.remove(self._settings.resume_fname)
 
+    def _detect_and_apply_job_inputs(self) -> None:
+        """If the user has staged launch inputs, apply them to the run."""
+        from wandb.sdk.launch.inputs.internal import StagedLaunchInputs
+
+        StagedLaunchInputs().apply(self)
+
     def _make_job_source_reqs(self) -> Tuple[List[str], Dict[str, Any], Dict[str, Any]]:
         from wandb.util import working_set
 
@@ -2555,6 +2603,14 @@ class Run:
         self._console_stop()  # TODO: there's a race here with jupyter console logging
 
         assert self._backend and self._backend.interface
+
+        # get the server info before starting the defer state machine as
+        # it will stop communication with the server
+        server_info_handle = self._backend.interface.deliver_request_server_info()
+        result = server_info_handle.wait(timeout=-1)
+        assert result
+        self._server_info_response = result.response.server_info_response
+
         exit_handle = self._backend.interface.deliver_exit(self._exit_code)
         exit_handle.add_probe(on_probe=self._on_probe_exit)
 
@@ -2563,6 +2619,7 @@ class Run:
         #     self._exit_code, settings=self._settings, printer=self._printer
         # )
 
+        # wait for the exit to complete
         _ = exit_handle.wait(timeout=-1, on_progress=self._on_progress_exit)
 
         poll_exit_handle = self._backend.interface.deliver_poll_exit()
@@ -2580,16 +2637,11 @@ class Run:
 
         # dispatch all our final requests
 
-        server_info_handle = self._backend.interface.deliver_request_server_info()
         final_summary_handle = self._backend.interface.deliver_get_summary()
         sampled_history_handle = (
             self._backend.interface.deliver_request_sampled_history()
         )
 
-        result = server_info_handle.wait(timeout=-1)
-        assert result
-        self._server_info_response = result.response.server_info_response
-
         result = sampled_history_handle.wait(timeout=-1)
         assert result
         self._sampled_history = result.response.sampled_history_response
@@ -2708,12 +2760,23 @@ class Run:
                 if i not in valid:
                     raise wandb.Error(f"Unhandled define_metric() arg: summary op: {i}")
                 summary_ops.append(i)
+            with telemetry.context(run=self) as tel:
+                tel.feature.metric_summary = True
         goal_cleaned: Optional[str] = None
         if goal is not None:
             goal_cleaned = goal[:3].lower()
             valid_goal = {"min", "max"}
             if goal_cleaned not in valid_goal:
                 raise wandb.Error(f"Unhandled define_metric() arg: goal: {goal}")
+            with telemetry.context(run=self) as tel:
+                tel.feature.metric_goal = True
+        if hidden:
+            with telemetry.context(run=self) as tel:
+                tel.feature.metric_hidden = True
+        if step_sync:
+            with telemetry.context(run=self) as tel:
+                tel.feature.metric_step_sync = True
+
         m = wandb_metric.Metric(
             name=name,
             step_metric=step_metric,
@@ -2898,6 +2961,7 @@ class Run:
             api.use_artifact(
                 artifact.id,
                 entity_name=r.entity,
+                project_name=r.project,
                 use_as=use_as or artifact_or_name,
             )
         else:
@@ -2960,8 +3024,7 @@ class Run:
                     - `s3://bucket/path`
                 You can also pass an Artifact object created by calling
                 `wandb.Artifact`.
-            name: (str, optional) An artifact name. May be prefixed with entity/project.
-                Valid names can be in the following forms:
+            name: (str, optional) An artifact name. Valid names can be in the following forms:
                     - name:version
                     - name:alias
                     - digest
@@ -3607,7 +3670,11 @@ class Run:
         project_url = settings.project_url
         sweep_url = settings.sweep_url
 
-        run_state_str = "Resuming run" if settings.resumed else "Syncing run"
+        run_state_str = (
+            "Resuming run"
+            if settings.resumed or settings.resume_from
+            else "Syncing run"
+        )
         run_name = settings.run_name
         if not run_name:
             return
@@ -3964,11 +4031,11 @@ class Run:
 
         # Render summary if available
         if summary:
-            final_summary = {
-                item.key: json.loads(item.value_json)
-                for item in summary.item
-                if not item.key.startswith("_")
-            }
+            final_summary = {}
+            for item in summary.item:
+                if item.key.startswith("_") or len(item.nested_key) > 0:
+                    continue
+                final_summary[item.key] = json.loads(item.value_json)
 
             logger.info("rendering summary")
             summary_rows = []