def listen(self, max_seconds=30):
"""Waits to receive up to two bytes for up to max_seconds"""
if not self.connection:
self.connect()
start = time.time()
conn, _, err = select([self.connection], [], [self.connection],
max_seconds)
try:
if len(err) > 0:
raise socket.error("Couldn't open socket")
message = b''
while True:
if time.time() - start > max_seconds:
raise socket.error(
"Timeout of %s seconds waiting for W&B process" %
max_seconds)
res = self.connection.recv(1024)
term = res.find(b'\0')
if term != -1:
message += res[:term]
break
else:
message += res
message = json.loads(message.decode('utf8'))
if message['status'] == 'done':
return True, None
elif message['status'] == 'ready':
return True, message
elif message['status'] == 'launch_error':
return False, None
else:
raise socket.error("Invalid status: %s" % message['status'])
except (socket.error, ValueError) as e:
util.sentry_exc(e)
return False, None
def _thread_except_body(self):
# TODO: Consolidate with internal_util.ExceptionThread
try:
self._thread_body()
except Exception as e:
exc_info = sys.exc_info()
self._exc_info = exc_info
logger.exception("generic exception in filestream thread")
util.sentry_exc(exc_info, delay=True)
raise e
def wandb_stream_read(fd):
# print("start reading", file=sys.stderr)
while True:
try:
data = os.read(fd, 200)
except OSError as e:
sentry_exc(e)
# print("problem", e, file=sys.stderr)
return
if len(data) == 0:
break
def init(
job_type = None,
dir=None,
config = None,
project = None,
entity = None,
reinit = None,
tags = None,
group = None,
name = None,
notes = None,
magic = None,
config_exclude_keys=None,
config_include_keys=None,
anonymous = None,
mode = None,
allow_val_change = None,
resume = None,
force = None,
tensorboard=None, # alias for sync_tensorboard
sync_tensorboard=None,
monitor_gym=None,
save_code=None,
id=None,
settings = None,
):
"""Initialize W&B
Spawns a new process to start or resume a run locally and communicate with a
wandb server. Should be called before any calls to wandb.log.
Arguments:
job_type (str, optional): The type of job running, defaults to 'train'
dir (str, optional): An absolute path to a directory where metadata will
be stored.
config (dict, argparse, or absl.flags, str, optional):
Sets the config parameters (typically hyperparameters) to store with the
run. See also wandb.config.
If dict, argparse or absl.flags: will load the key value pairs into
the runs config object.
If str: will look for a yaml file that includes config parameters and
load them into the run's config object.
project (str, optional): W&B Project.
entity (str, optional): W&B Entity.
reinit (bool, optional): Allow multiple calls to init in the same process.
tags (list, optional): A list of tags to apply to the run.
group (str, optional): A unique string shared by all runs in a given group.
name (str, optional): A display name for the run which does not have to be
unique.
notes (str, optional): A multiline string associated with the run.
magic (bool, dict, or str, optional): magic configuration as bool, dict,
json string, yaml filename.
config_exclude_keys (list, optional): string keys to exclude storing in W&B
when specifying config.
config_include_keys (list, optional): string keys to include storing in W&B
when specifying config.
anonymous (str, optional): Can be "allow", "must", or "never". Controls
whether anonymous logging is allowed. Defaults to never.
mode (str, optional): Can be "online", "offline" or "disabled". Defaults to
online.
allow_val_change (bool, optional): allow config values to be changed after
setting. Defaults to true in jupyter and false otherwise.
resume (bool, str, optional): Sets the resuming behavior. Should be one of:
"allow", "must", "never", "auto" or None. Defaults to None.
Cases:
- "auto" (or True): automatically resume the previous run on the same machine.
if the previous run crashed, otherwise starts a new run.
- "allow": if id is set with init(id="UNIQUE_ID") or WANDB_RUN_ID="UNIQUE_ID"
and it is identical to a previous run, wandb will automatically resume the
run with the id. Otherwise wandb will start a new run.
- "never": if id is set with init(id="UNIQUE_ID") or WANDB_RUN_ID="UNIQUE_ID"
and it is identical to a previous run, wandb will crash.
- "must": if id is set with init(id="UNIQUE_ID") or WANDB_RUN_ID="UNIQUE_ID"
and it is identical to a previous run, wandb will automatically resume the
run with the id. Otherwise wandb will crash.
- None: never resumes - if a run has a duplicate run_id the previous run is
overwritten.
See https://docs.wandb.com/library/advanced/resuming for more detail.
force (bool, optional): If true, will cause script to crash if user can't or isn't
logged in to a wandb server. If false, will cause script to run in offline
modes if user can't or isn't logged in to a wandb server. Defaults to false.
sync_tensorboard (bool, optional): Synchronize wandb logs from tensorboard or
tensorboardX and saves the relevant events file. Defaults to false.
monitor_gym: (bool, optional): automatically logs videos of environment when
using OpenAI Gym (see https://docs.wandb.com/library/integrations/openai-gym)
Defaults to false.
save_code (bool, optional): Save the entrypoint or jupyter session history
source code.
id (str, optional): A globally unique (per project) identifier for the run. This
is primarily used for resuming.
Examples:
Basic usage
```
wandb.init()
```
Launch multiple runs from the same script
```
for x in range(10):
with wandb.init(project="my-projo") as run:
for y in range(100):
run.log({"metric": x+y})
```
Raises:
Exception: if problem.
Returns:
A `Run` object.
"""
assert not wandb._IS_INTERNAL_PROCESS
kwargs = dict(locals())
error_seen = None
except_exit = None
try:
wi = _WandbInit()
wi.setup(kwargs)
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):
sentry_exc(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 UsageError:
raise
except KeyboardInterrupt as e:
assert logger
logger.warning("interrupted", exc_info=e)
raise e
except Exception as e:
error_seen = e
traceback.print_exc()
assert logger
logger.error("error", 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.
sentry_exc(e, delay=True)
# reraise(*sys.exc_info())
# six.raise_from(Exception("problem"), e)
finally:
if error_seen:
wandb.termerror("Abnormal program exit")
if except_exit:
os._exit(-1)
six.raise_from(Exception("problem"), error_seen)
return run
def init(
job_type = None,
dir=None,
config = None,
project = None,
entity = None,
reinit = None,
tags = None,
group = None,
name = None,
notes = None,
magic = None,
config_exclude_keys=None,
config_include_keys=None,
anonymous = None,
mode = None,
allow_val_change = None,
resume = None,
force = None,
tensorboard=None, # alias for sync_tensorboard
sync_tensorboard=None,
monitor_gym=None,
save_code=None,
id=None,
settings = None,
):
"""
Start a new tracked run with `wandb.init()`.
In an ML training pipeline, you could add `wandb.init()`
to the beginning of your training script as well as your evaluation
script, and each piece would be tracked as a run in W&B.
`wandb.init()` spawns a new background process to log data to a run, and it
also syncs data to wandb.ai by default so you can see live visualizations.
Call `wandb.init()` to start a run before logging data with `wandb.log()`.
`wandb.init()` returns a run object, and you can also access the run object
with wandb.run.
Arguments:
project: (str, optional) The name of the project where you're sending
the new run. If the project is not specified, the run is put in an
"Uncategorized" project.
entity: (str, optional) An entity is a username or team name where
you're sending runs. This entity must exist before you can send runs
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
in [Settings](wandb.ai/settings) under "default location to create
new projects".
config: (dict, argparse, absl.flags, str, optional)
This sets wandb.config, a dictionary-like object for saving inputs
to your job, like hyperparameters for a model or settings for a data
preprocessing job. The config will show up in a table in the UI that
you can use to group, filter, and sort runs. Keys should not contain
`.` in their names, and values should be under 10 MB.
If dict, argparse or absl.flags: will load the key value pairs into
the wandb.config object.
If str: will look for a yaml file by that name, and load config from
that file into the wandb.config object.
save_code: (bool, optional) Turn this on to save the main script or
notebook to W&B. This is valuable for improving experiment
reproducibility and to diff code across experiments in the UI. By
default this is off, but you can flip the default behavior to "on"
in [Settings](wandb.ai/settings).
group: (str, optional) Specify a group to organize individual runs into
a larger experiment. For example, you might be doing cross
validation, or you might have multiple jobs that train and evaluate
a model against different test sets. Group gives you a way to
organize runs together into a larger whole, and you can toggle this
on and off in the UI. For more details, see
[Grouping](docs.wandb.com/library/grouping).
job_type: (str, optional) Specify the type of run, which is useful when
you're grouping runs together into larger experiments using group.
For example, you might have multiple jobs in a group, with job types
like train and eval. Setting this makes it easy to filter and group
similar runs together in the UI so you can compare apples to apples.
tags: (list, optional) A list of strings, which will populate the list
of tags on this run in the UI. Tags are useful for organizing runs
together, or applying temporary labels like "baseline" or
"production". It's easy to add and remove tags in the UI, or filter
down to just runs with a specific tag.
name: (str, optional) A short display name for this run, which is how
you'll identify this run in the UI. By default we generate a random
two-word name that lets you easily cross-reference runs from the
table to charts. Keeping these run names short makes the chart
legends and tables easier to read. If you're looking for a place to
save your hyperparameters, we recommend saving those in config.
notes: (str, optional) A longer description of the run, like a -m commit
message in git. This helps you remember what you were doing when you
ran this run.
dir: (str, optional) An absolute path to a directory where metadata will
be stored. When you call download() on an artifact, this is the
directory where downloaded files will be saved. By default this is
the ./wandb directory.
resume (bool, str, optional): Sets the resuming behavior. Options:
"allow", "must", "never", "auto" or None. Defaults to None.
Cases:
- None (default): If the new run has the same ID as a previous run,
this run overwrites that data.
- "auto" (or True): if the preivous run on this machine crashed,
automatically resume it. Otherwise, start a new run.
- "allow": if id is set with init(id="UNIQUE_ID") or
WANDB_RUN_ID="UNIQUE_ID" and it is identical to a previous run,
wandb will automatically resume the run with that id. Otherwise,
wandb will start a new run.
- "never": if id is set with init(id="UNIQUE_ID") or
WANDB_RUN_ID="UNIQUE_ID" and it is identical to a previous run,
wandb will crash.
- "must": if id is set with init(id="UNIQUE_ID") or
WANDB_RUN_ID="UNIQUE_ID" and it is identical to a previous run,
wandb will automatically resume the run with the id. Otherwise
wandb will crash.
See https://docs.wandb.com/library/advanced/resuming for more.
reinit: (bool, optional) Allow multiple wandb.init() calls in the same
process. (default: False)
magic: (bool, dict, or str, optional) The bool controls whether we try to
auto-instrument your script, capturing basic details of your run
without you having to add more wandb code. (default: False)
You can also pass a dict, json string, or yaml filename.
config_exclude_keys: (list, optional) string keys to exclude from
`wandb.config`.
config_include_keys: (list, optional) string keys to include in
wandb.config.
anonymous: (str, optional) Controls anonymous data logging. Options:
- "never" (default): requires you to link your W&B account before
tracking the run so you don't accidentally create an anonymous
run.
- "allow": lets a logged-in user track runs with their account, but
lets someone who is running the script without a W&B account see
the charts in the UI.
- "must": sends the run to an anonymous account instead of to a
signed-up user account.
mode: (str, optional) Can be "online", "offline" or "disabled". Defaults to
online.
allow_val_change: (bool, optional) Whether to allow config values to
change after setting the keys once. By default we throw an exception
if a config value is overwritten. If you want to track something
like a varying learning_rate at multiple times during training, use
wandb.log() instead. (default: False in scripts, True in Jupyter)
force: (bool, optional) If True, this crashes the script if a user isn't
logged in to W&B. If False, this will let the script run in offline
mode if a user isn't logged in to W&B. (default: False)
sync_tensorboard: (bool, optional) Synchronize wandb logs from tensorboard or
tensorboardX and saves the relevant events file. (default: False)
monitor_gym: (bool, optional) automatically logs videos of environment when
using OpenAI Gym. (default: False)
See https://docs.wandb.com/library/integrations/openai-gym
id: (str, optional) A unique ID for this run, used for Resuming. It must
be unique in the project, and if you delete a run you can't reuse
the ID. Use the name field for a short descriptive name, or config
for saving hyperparameters to compare across runs. The ID cannot
contain special characters.
See https://docs.wandb.com/library/resuming
Examples:
Basic usage
```
wandb.init()
```
Launch multiple runs from the same script
```
for x in range(10):
with wandb.init(project="my-projo") as run:
for y in range(100):
run.log({"metric": x+y})
```
Raises:
Exception: if problem.
Returns:
A `Run` object.
"""
wandb._assert_is_user_process()
kwargs = dict(locals())
error_seen = None
except_exit = None
try:
wi = _WandbInit()
wi.setup(kwargs)
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):
sentry_exc(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 UsageError:
raise
except KeyboardInterrupt as e:
assert logger
logger.warning("interrupted", exc_info=e)
raise e
except Exception as e:
error_seen = e
traceback.print_exc()
assert logger
logger.error("error", 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.
sentry_exc(e, delay=True)
# reraise(*sys.exc_info())
# six.raise_from(Exception("problem"), e)
finally:
if error_seen:
wandb.termerror("Abnormal program exit")
if except_exit:
os._exit(-1)
six.raise_from(Exception("problem"), error_seen)
return run
def wandb_internal(
settings,
record_q,
result_q,
):
"""Internal process function entrypoint.
Read from record queue and dispatch work to various threads.
Arguments:
settings: dictionary of configuration parameters.
record_q: records to be handled
result_q: for sending results back
"""
# mark this process as internal
wandb._set_internal_process()
started = time.time()
# register the exit handler only when wandb_internal is called, not on import
@atexit.register
def handle_exit(*args):
logger.info("Internal process exited")
# Lets make sure we dont modify settings so use a static object
_settings = settings_static.SettingsStatic(settings)
if _settings.log_internal:
configure_logging(_settings.log_internal, _settings._log_level)
parent_pid = os.getppid()
pid = os.getpid()
logger.info(
"W&B internal server running at pid: %s, started at: %s",
pid,
datetime.fromtimestamp(started),
)
publish_interface = interface.BackendSender(record_q=record_q)
stopped = threading.Event()
threads = []
send_record_q = queue.Queue()
record_sender_thread = SenderThread(
settings=_settings,
record_q=send_record_q,
result_q=result_q,
stopped=stopped,
interface=publish_interface,
debounce_interval_ms=30000,
)
threads.append(record_sender_thread)
write_record_q = queue.Queue()
record_writer_thread = WriterThread(
settings=_settings,
record_q=write_record_q,
result_q=result_q,
stopped=stopped,
writer_q=write_record_q,
)
threads.append(record_writer_thread)
record_handler_thread = HandlerThread(
settings=_settings,
record_q=record_q,
result_q=result_q,
stopped=stopped,
sender_q=send_record_q,
writer_q=write_record_q,
interface=publish_interface,
)
threads.append(record_handler_thread)
process_check = ProcessCheck(settings=_settings, pid=parent_pid)
for thread in threads:
thread.start()
interrupt_count = 0
while not stopped.is_set():
try:
# wait for stop event
while not stopped.is_set():
time.sleep(1)
if process_check.is_dead():
logger.error("Internal process shutdown.")
stopped.set()
except KeyboardInterrupt:
interrupt_count += 1
logger.warning(
"Internal process interrupt: {}".format(interrupt_count))
finally:
if interrupt_count >= 2:
logger.error("Internal process interrupted.")
stopped.set()
for thread in threads:
thread.join()
for thread in threads:
exc_info = thread.get_exception()
if exc_info:
logger.error("Thread {}:".format(thread.name), exc_info=exc_info)
print("Thread {}:".format(thread.name), file=sys.stderr)
traceback.print_exception(*exc_info)
sentry_exc(exc_info, delay=True)
wandb.termerror("Internal wandb error: file data was not synced")
sys.exit(-1)
def wandb_internal(settings, record_q, result_q):
"""Internal process function entrypoint.
Read from record queue and dispatch work to various threads.
Args:
settings: dictionary of configuration parameters.
record_q: records to be handled
result_q: for sending results back
"""
# mark this process as internal
wandb._IS_INTERNAL_PROCESS = True
# Lets make sure we dont modify settings so use a static object
settings = settings_static.SettingsStatic(settings)
if settings.log_internal:
configure_logging(settings.log_internal, settings._log_level)
parent_pid = os.getppid()
pid = os.getpid()
logger.info("W&B internal server running at pid: %s", pid)
publish_interface = interface.BackendSender(record_q=record_q)
stopped = threading.Event()
send_record_q = queue.Queue()
record_sender_thread = SenderThread(
settings=settings,
record_q=send_record_q,
result_q=result_q,
stopped=stopped,
interface=publish_interface,
)
threads = [record_sender_thread]
write_record_q = queue.Queue()
record_writer_thread = WriterThread(
settings=settings,
record_q=write_record_q,
result_q=result_q,
stopped=stopped,
writer_q=write_record_q,
)
threads.append(record_writer_thread)
record_handler_thread = HandlerThread(
settings=settings,
record_q=record_q,
result_q=result_q,
stopped=stopped,
sender_q=send_record_q,
writer_q=write_record_q,
interface=publish_interface,
)
threads.append(record_handler_thread)
process_check = ProcessCheck(settings=settings, pid=parent_pid)
for thread in threads:
thread.start()
interrupt_count = 0
while not stopped.isSet():
try:
# wait for stop event
while not stopped.isSet():
time.sleep(1)
if process_check.is_dead():
logger.error("Internal process shutdown.")
stopped.set()
except KeyboardInterrupt:
interrupt_count += 1
logger.warning(
"Internal process interrupt: {}".format(interrupt_count))
finally:
if interrupt_count >= 2:
logger.error("Internal process interrupted.")
stopped.set()
for thread in threads:
thread.join()
for thread in threads:
if exc_info := thread.get_exception():
logger.error("Thread {}:".format(thread.name), exc_info=exc_info)
print("Thread {}:".format(thread.name), file=sys.stderr)
traceback.print_exception(*exc_info)
sentry_exc(exc_info, delay=True)
sys.exit(-1)
def init(
job_type = None,
dir=None,
config = None, # TODO(jhr): type is a union for argparse/absl
project = None,
entity = None,
reinit = None,
tags = None,
group = None,
name = None,
notes = None,
magic = None, # TODO(jhr): type is union
config_exclude_keys=None,
config_include_keys=None,
anonymous = None,
mode = None,
allow_val_change = None,
resume = None,
force = None,
tensorboard=None, # alias for sync_tensorboard
sync_tensorboard=None,
monitor_gym=None,
save_code=None,
id=None,
settings = None,
):
"""Initialize a wandb Run.
Args:
entity: alias for team.
team: personal user or team to use for Run.
project: project name for the Run.
Raises:
Exception: if problem.
Returns:
wandb Run object
"""
assert not wandb._IS_INTERNAL_PROCESS
kwargs = dict(locals())
error_seen = None
except_exit = None
try:
wi = _WandbInit()
wi.setup(kwargs)
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):
sentry_exc(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 UsageError:
raise
except KeyboardInterrupt as e:
assert logger
logger.warning("interrupted", exc_info=e)
raise e
except Exception as e:
error_seen = e
traceback.print_exc()
assert logger
logger.error("error", 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.
sentry_exc(e, delay=True)
# reraise(*sys.exc_info())
# six.raise_from(Exception("problem"), e)
finally:
if error_seen:
wandb.termerror("Abnormal program exit")
if except_exit:
os._exit(-1)
six.raise_from(Exception("problem"), error_seen)
return run
def init(
settings=None,
entity=None,
team=None,
project=None,
mode=None,
group=None,
job_type=None,
tags=None,
name=None,
config=None, # TODO(jhr): type is a union for argparse/absl
notes=None,
magic=None, # TODO(jhr): type is union
config_exclude_keys=None,
config_include_keys=None,
reinit=None,
anonymous=None,
dir=None,
allow_val_change=None,
resume=None,
force=None,
tensorboard=None,
sync_tensorboard=None,
monitor_gym=None,
id=None,
):
"""Initialize a wandb Run.
Args:
entity: alias for team.
team: personal user or team to use for Run.
project: project name for the Run.
Raises:
Exception: if problem.
Returns:
wandb Run object
"""
assert not wandb._IS_INTERNAL_PROCESS
kwargs = locals()
try:
wi = _WandbInit()
wi.setup(kwargs)
try:
run = wi.init()
except (KeyboardInterrupt, Exception) as e:
if not isinstance(e, KeyboardInterrupt):
sentry_exc(e)
getcaller()
assert logger
logger.exception("we got issues")
wi._atexit_cleanup()
if wi.settings.problem == "fatal":
raise
if wi.settings.problem == "warn":
pass
run = RunDummy()
except KeyboardInterrupt as e:
assert logger
logger.warning("interupted", exc_info=e)
raise_from(Exception("interrupted"), e)
except Exception as e:
assert logger
logger.error("error", 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.
sentry_exc(e, delay=True)
raise_from(Exception("problem"), e)
return run
def init(
job_type: Optional[str] = None,
dir=None,
config: Union[Dict,
None] = None, # TODO(jhr): type is a union for argparse/absl
project: Optional[str] = None,
entity: Optional[str] = None,
reinit: bool = None,
tags: Optional[List] = None,
team: Optional[str] = None,
group: Optional[str] = None,
name: Optional[str] = None,
notes: Optional[str] = None,
magic: bool = None, # TODO(jhr): type is union
config_exclude_keys=None,
config_include_keys=None,
anonymous: Optional[str] = None,
disable: bool = None,
offline: bool = None,
allow_val_change: bool = None,
resume: Optional[Union[bool, str]] = None,
force=None,
tensorboard=None, # alias for sync_tensorboard
sync_tensorboard=None,
monitor_gym=None,
id=None,
settings: Union[Settings, Dict[str, Any], str, None] = None,
) -> Run:
"""Initialize a wandb Run.
Args:
entity: alias for team.
team: personal user or team to use for Run.
project: project name for the Run.
Raises:
Exception: if problem.
Returns:
wandb Run object
"""
assert not wandb._IS_INTERNAL_PROCESS
kwargs = locals()
try:
wi = _WandbInit()
wi.setup(kwargs)
try:
run = wi.init()
except (KeyboardInterrupt, Exception) as e:
if not isinstance(e, KeyboardInterrupt):
sentry_exc(e)
getcaller()
assert logger
if wi.settings.problem == "fatal":
raise
if wi.settings.problem == "warn":
pass
run = RunDummy()
except UsageError:
raise
except KeyboardInterrupt as e:
assert logger
logger.warning("interrupted", exc_info=e)
raise_from(Exception("interrupted"), e)
except Exception as e:
assert logger
logger.error("error", 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.
sentry_exc(e, delay=True)
reraise(*sys.exc_info())
# raise_from(Exception("problem"), e)
return run
def init(
job_type: Optional[str] = None,
dir=None,
config: Union[Dict, str, None] = None,
project: Optional[str] = None,
entity: Optional[str] = None,
reinit: bool = None,
tags: Optional[Sequence] = None,
group: Optional[str] = None,
name: Optional[str] = None,
notes: Optional[str] = None,
magic: Union[dict, str, bool] = None,
config_exclude_keys=None,
config_include_keys=None,
anonymous: Optional[str] = None,
mode: Optional[str] = None,
allow_val_change: Optional[bool] = None,
resume: Optional[Union[bool, str]] = None,
force: Optional[bool] = None,
tensorboard=None, # alias for sync_tensorboard
sync_tensorboard=None,
monitor_gym=None,
save_code=None,
id=None,
settings: Union[Settings, Dict[str, Any], None] = None,
) -> Union[Run, RunDisabled, None]:
"""Starts a new run to track and log to W&B.
In an ML training pipeline, you could add `wandb.init()`
to the beginning of your training script as well as your evaluation
script, and each piece would be tracked as a run in W&B.
`wandb.init()` spawns a new background process to log data to a run, and it
also syncs data to wandb.ai by default so you can see live visualizations.
Call `wandb.init()` to start a run before logging data with `wandb.log()`:
<!--yeadoc-test:init-method-log-->
```python
import wandb
wandb.init()
# ... calculate metrics, generate media
wandb.log({"accuracy": 0.9})
```
`wandb.init()` returns a run object, and you can also access the run object
via `wandb.run`:
<!--yeadoc-test:init-and-assert-global-->
```python
import wandb
run = wandb.init()
assert run is wandb.run
```
At the end of your script, we will automatically call `wandb.finish` to
finalize and cleanup the run. However, if you call `wandb.init` from a
child process, you must explicitly call `wandb.finish` at the end of the
child process.
For more on using `wandb.init()`, including detailed examples, check out our
[guide and FAQs](https://docs.wandb.ai/guides/track/launch).
Arguments:
project: (str, optional) The name of the project where you're sending
the new run. If the project is not specified, the run is put in an
"Uncategorized" project.
entity: (str, optional) An entity is a username or team name where
you're sending runs. This entity must exist before you can send runs
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
in [your settings](https://wandb.ai/settings) under "default location
to create new projects".
config: (dict, argparse, absl.flags, str, optional)
This sets `wandb.config`, a dictionary-like object for saving inputs
to your job, like hyperparameters for a model or settings for a data
preprocessing job. The config will show up in a table in the UI that
you can use to group, filter, and sort runs. Keys should not contain
`.` in their names, and values should be under 10 MB.
If dict, argparse or absl.flags: will load the key value pairs into
the `wandb.config` object.
If str: will look for a yaml file by that name, and load config from
that file into the `wandb.config` object.
save_code: (bool, optional) Turn this on to save the main script or
notebook to W&B. This is valuable for improving experiment
reproducibility and to diff code across experiments in the UI. By
default this is off, but you can flip the default behavior to on
in [your settings page](https://wandb.ai/settings).
group: (str, optional) Specify a group to organize individual runs into
a larger experiment. For example, you might be doing cross
validation, or you might have multiple jobs that train and evaluate
a model against different test sets. Group gives you a way to
organize runs together into a larger whole, and you can toggle this
on and off in the UI. For more details, see our
[guide to grouping runs](https://docs.wandb.com/library/grouping).
job_type: (str, optional) Specify the type of run, which is useful when
you're grouping runs together into larger experiments using group.
For example, you might have multiple jobs in a group, with job types
like train and eval. Setting this makes it easy to filter and group
similar runs together in the UI so you can compare apples to apples.
tags: (list, optional) A list of strings, which will populate the list
of tags on this run in the UI. Tags are useful for organizing runs
together, or applying temporary labels like "baseline" or
"production". It's easy to add and remove tags in the UI, or filter
down to just runs with a specific tag.
name: (str, optional) A short display name for this run, which is how
you'll identify this run in the UI. By default we generate a random
two-word name that lets you easily cross-reference runs from the
table to charts. Keeping these run names short makes the chart
legends and tables easier to read. If you're looking for a place to
save your hyperparameters, we recommend saving those in config.
notes: (str, optional) A longer description of the run, like a `-m` commit
message in git. This helps you remember what you were doing when you
ran this run.
dir: (str, optional) An absolute path to a directory where metadata will
be stored. When you call `download()` on an artifact, this is the
directory where downloaded files will be saved. By default this is
the `./wandb` directory.
resume: (bool, str, optional) Sets the resuming behavior. Options:
`"allow"`, `"must"`, `"never"`, `"auto"` or `None`. Defaults to `None`.
Cases:
- `None` (default): If the new run has the same ID as a previous run,
this run overwrites that data.
- `"auto"` (or `True`): if the preivous run on this machine crashed,
automatically resume it. Otherwise, start a new run.
- `"allow"`: if id is set with `init(id="UNIQUE_ID")` or
`WANDB_RUN_ID="UNIQUE_ID"` and it is identical to a previous run,
wandb will automatically resume the run with that id. Otherwise,
wandb will start a new run.
- `"never"`: if id is set with `init(id="UNIQUE_ID")` or
`WANDB_RUN_ID="UNIQUE_ID"` and it is identical to a previous run,
wandb will crash.
- `"must"`: if id is set with `init(id="UNIQUE_ID")` or
`WANDB_RUN_ID="UNIQUE_ID"` and it is identical to a previous run,
wandb will automatically resume the run with the id. Otherwise
wandb will crash.
See [our guide to resuming runs](https://docs.wandb.com/library/advanced/resuming)
for more.
reinit: (bool, optional) Allow multiple `wandb.init()` calls in the same
process. (default: `False`)
magic: (bool, dict, or str, optional) The bool controls whether we try to
auto-instrument your script, capturing basic details of your run
without you having to add more wandb code. (default: `False`)
You can also pass a dict, json string, or yaml filename.
config_exclude_keys: (list, optional) string keys to exclude from
`wandb.config`.
config_include_keys: (list, optional) string keys to include in
`wandb.config`.
anonymous: (str, optional) Controls anonymous data logging. Options:
- `"never"` (default): requires you to link your W&B account before
tracking the run so you don't accidentally create an anonymous
run.
- `"allow"`: lets a logged-in user track runs with their account, but
lets someone who is running the script without a W&B account see
the charts in the UI.
- `"must"`: sends the run to an anonymous account instead of to a
signed-up user account.
mode: (str, optional) Can be `"online"`, `"offline"` or `"disabled"`. Defaults to
online.
allow_val_change: (bool, optional) Whether to allow config values to
change after setting the keys once. By default we throw an exception
if a config value is overwritten. If you want to track something
like a varying learning rate at multiple times during training, use
`wandb.log()` instead. (default: `False` in scripts, `True` in Jupyter)
force: (bool, optional) If `True`, this crashes the script if a user isn't
logged in to W&B. If `False`, this will let the script run in offline
mode if a user isn't logged in to W&B. (default: `False`)
sync_tensorboard: (bool, optional) Synchronize wandb logs from tensorboard or
tensorboardX and save the relevant events file. (default: `False`)
monitor_gym: (bool, optional) Automatically log videos of environment when
using OpenAI Gym. (default: `False`)
See [our guide to this integration](https://docs.wandb.com/library/integrations/openai-gym).
id: (str, optional) A unique ID for this run, used for resuming. It must
be unique in the project, and if you delete a run you can't reuse
the ID. Use the name field for a short descriptive name, or config
for saving hyperparameters to compare across runs. The ID cannot
contain special characters.
See [our guide to resuming runs](https://docs.wandb.com/library/resuming).
Examples:
### Set where the run is logged
You can change where the run is logged, just like changing
the organization, repository, and branch in git:
```python
import wandb
user = "******"
project = "capsules"
display_name = "experiment-2021-10-31"
wandb.init(entity=user, project=project, name=display_name)
```
### Add metadata about the run to the config
Pass a dictionary-style object as the `config` keyword argument to add
metadata, like hyperparameters, to your run.
<!--yeadoc-test:init-set-config--->
```python
import wandb
config = {"lr": 3e-4, "batch_size": 32}
config.update({"architecture": "resnet", "depth": 34})
wandb.init(config=config)
```
Raises:
Exception: if problem.
Returns:
A `Run` object.
"""
wandb._assert_is_user_process()
if resume is True:
resume = "auto" # account for changing resume interface, True and auto should behave the same
kwargs = dict(locals())
error_seen = None
except_exit = None
try:
wi = _WandbInit()
wi.setup(kwargs)
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):
sentry_exc(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 UsageError as e:
wandb.termerror(str(e))
raise
except KeyboardInterrupt as e:
assert logger
logger.warning("interrupted", exc_info=e)
raise e
except Exception as e:
error_seen = e
traceback.print_exc()
assert logger
logger.error("error", 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.
sentry_exc(e, delay=True)
# reraise(*sys.exc_info())
# six.raise_from(Exception("problem"), e)
finally:
if error_seen:
wandb.termerror("Abnormal program exit")
if except_exit:
os._exit(-1)
six.raise_from(Exception("problem"), error_seen)
return run
