Utils
zenml.utils
special
Initialization of the utils module.
The utils module contains utility functions handling analytics, reading and
writing YAML data as well as other general purpose functions.
artifact_utils
Util functions for artifact handling.
get_artifacts_of_pipeline_run(pipeline_run, only_produced=False)
Get all artifacts produced during a pipeline run.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
pipeline_run |
PipelineRunResponseModel |
The pipeline run. |
required |
only_produced |
bool |
If only artifacts produced by the pipeline run should be returned or also cached artifacts. |
False |
Returns:
| Type | Description |
|---|---|
List[ArtifactResponseModel] |
A list of all artifacts produced during the pipeline run. |
Source code in zenml/utils/artifact_utils.py
def get_artifacts_of_pipeline_run(
pipeline_run: "PipelineRunResponseModel", only_produced: bool = False
) -> List["ArtifactResponseModel"]:
"""Get all artifacts produced during a pipeline run.
Args:
pipeline_run: The pipeline run.
only_produced: If only artifacts produced by the pipeline run should be
returned or also cached artifacts.
Returns:
A list of all artifacts produced during the pipeline run.
"""
artifacts: List["ArtifactResponseModel"] = []
for step in pipeline_run.steps.values():
if not only_produced or step.status == ExecutionStatus.COMPLETED:
artifacts.extend(step.outputs.values())
return artifacts
get_producer_step_of_artifact(artifact)
Get the step run that produced a given artifact.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
artifact |
ArtifactResponseModel |
The artifact. |
required |
Returns:
| Type | Description |
|---|---|
StepRunResponseModel |
The step run that produced the artifact. |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If the run that created the artifact no longer exists. |
Source code in zenml/utils/artifact_utils.py
def get_producer_step_of_artifact(
artifact: "ArtifactResponseModel",
) -> "StepRunResponseModel":
"""Get the step run that produced a given artifact.
Args:
artifact: The artifact.
Returns:
The step run that produced the artifact.
Raises:
RuntimeError: If the run that created the artifact no longer exists.
"""
if not artifact.producer_step_run_id:
raise RuntimeError(
f"The run that produced the artifact with id '{artifact.id}' no "
"longer exists. This can happen if the run was deleted."
)
return Client().get_run_step(artifact.producer_step_run_id)
load_artifact(artifact)
Load the given artifact into memory.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
artifact |
ArtifactResponseModel |
The artifact to load. |
required |
Returns:
| Type | Description |
|---|---|
Any |
The artifact loaded into memory. |
Source code in zenml/utils/artifact_utils.py
def load_artifact(artifact: "ArtifactResponseModel") -> Any:
"""Load the given artifact into memory.
Args:
artifact: The artifact to load.
Returns:
The artifact loaded into memory.
"""
artifact_store_loaded = False
if artifact.artifact_store_id:
try:
artifact_store_model = Client().get_stack_component(
component_type=StackComponentType.ARTIFACT_STORE,
name_id_or_prefix=artifact.artifact_store_id,
)
_ = StackComponent.from_model(artifact_store_model)
artifact_store_loaded = True
except KeyError:
pass
if not artifact_store_loaded:
logger.warning(
"Unable to restore artifact store while trying to load artifact "
"`%s`. If this artifact is stored in a remote artifact store, "
"this might lead to issues when trying to load the artifact.",
artifact.id,
)
return _load_artifact(
materializer=artifact.materializer,
data_type=artifact.data_type,
uri=artifact.uri,
)
load_artifact_visualization(artifact, index=0, zen_store=None, encode_image=False)
Load a visualization of the given artifact.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
artifact |
ArtifactResponseModel |
The artifact to visualize. |
required |
index |
int |
The index of the visualization to load. |
0 |
zen_store |
Optional[BaseZenStore] |
The ZenStore to use for finding the artifact store. If not provided, the client's ZenStore will be used. |
None |
encode_image |
bool |
Whether to base64 encode image visualizations. |
False |
Returns:
| Type | Description |
|---|---|
LoadedVisualizationModel |
The loaded visualization. |
Exceptions:
| Type | Description |
|---|---|
DoesNotExistException |
If the artifact does not have the requested visualization or if the visualization was not found in the artifact store. |
Source code in zenml/utils/artifact_utils.py
def load_artifact_visualization(
artifact: "ArtifactResponseModel",
index: int = 0,
zen_store: Optional["BaseZenStore"] = None,
encode_image: bool = False,
) -> LoadedVisualizationModel:
"""Load a visualization of the given artifact.
Args:
artifact: The artifact to visualize.
index: The index of the visualization to load.
zen_store: The ZenStore to use for finding the artifact store. If not
provided, the client's ZenStore will be used.
encode_image: Whether to base64 encode image visualizations.
Returns:
The loaded visualization.
Raises:
DoesNotExistException: If the artifact does not have the requested
visualization or if the visualization was not found in the artifact
store.
"""
# Get the visualization to load
if not artifact.visualizations:
raise DoesNotExistException(
f"Artifact '{artifact.id}' has no visualizations."
)
if index < 0 or index >= len(artifact.visualizations):
raise DoesNotExistException(
f"Artifact '{artifact.id}' only has {len(artifact.visualizations)} "
f"visualizations, but index {index} was requested."
)
visualization = artifact.visualizations[index]
# Load the visualization from the artifact's artifact store
if not artifact.artifact_store_id:
raise DoesNotExistException(
f"Artifact '{artifact.id}' cannot be visualized because the "
"underlying artifact store was deleted."
)
artifact_store = _load_artifact_store(
artifact_store_id=artifact.artifact_store_id, zen_store=zen_store
)
mode = "rb" if visualization.type == VisualizationType.IMAGE else "r"
value = _load_file_from_artifact_store(
uri=visualization.uri,
artifact_store=artifact_store,
mode=mode,
)
# Encode image visualizations if requested
if visualization.type == VisualizationType.IMAGE and encode_image:
value = base64.b64encode(bytes(value))
return LoadedVisualizationModel(type=visualization.type, value=value)
load_model_from_metadata(model_uri)
Load a zenml model artifact from a json file.
This function is used to load information from a Yaml file that was created by the save_model_metadata function. The information in the Yaml file is used to load the model into memory in the inference environment.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_uri |
str |
the artifact to extract the metadata from. |
required |
Returns:
| Type | Description |
|---|---|
Any |
The ML model object loaded into memory. |
Source code in zenml/utils/artifact_utils.py
def load_model_from_metadata(model_uri: str) -> Any:
"""Load a zenml model artifact from a json file.
This function is used to load information from a Yaml file that was created
by the save_model_metadata function. The information in the Yaml file is
used to load the model into memory in the inference environment.
Args:
model_uri: the artifact to extract the metadata from.
Returns:
The ML model object loaded into memory.
"""
# Load the model from its metadata
with fileio.open(
os.path.join(model_uri, MODEL_METADATA_YAML_FILE_NAME), "r"
) as f:
metadata = read_yaml(f.name)
data_type = metadata[METADATA_DATATYPE]
materializer = metadata[METADATA_MATERIALIZER]
model = _load_artifact(
materializer=materializer, data_type=data_type, uri=model_uri
)
# Switch to eval mode if the model is a torch model
try:
import torch.nn as nn
if isinstance(model, nn.Module):
model.eval()
except ImportError:
pass
return model
save_model_metadata(model_artifact)
Save a zenml model artifact metadata to a YAML file.
This function is used to extract and save information from a zenml model artifact such as the model type and materializer. The extracted information will be the key to loading the model into memory in the inference environment.
datatype: the model type. This is the path to the model class. materializer: the materializer class. This is the path to the materializer class.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_artifact |
ArtifactResponseModel |
the artifact to extract the metadata from. |
required |
Returns:
| Type | Description |
|---|---|
str |
The path to the temporary file where the model metadata is saved |
Source code in zenml/utils/artifact_utils.py
def save_model_metadata(model_artifact: "ArtifactResponseModel") -> str:
"""Save a zenml model artifact metadata to a YAML file.
This function is used to extract and save information from a zenml model artifact
such as the model type and materializer. The extracted information will be
the key to loading the model into memory in the inference environment.
datatype: the model type. This is the path to the model class.
materializer: the materializer class. This is the path to the materializer class.
Args:
model_artifact: the artifact to extract the metadata from.
Returns:
The path to the temporary file where the model metadata is saved
"""
metadata = dict()
metadata[METADATA_DATATYPE] = model_artifact.data_type
metadata[METADATA_MATERIALIZER] = model_artifact.materializer
with tempfile.NamedTemporaryFile(
mode="w", suffix=".yaml", delete=False
) as f:
write_yaml(f.name, metadata)
return f.name
upload_artifact(name, data, materializer, artifact_store_id, extract_metadata, include_visualizations)
Upload and publish an artifact.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
name |
str |
The name of the artifact. |
required |
data |
Any |
The artifact data. |
required |
materializer |
BaseMaterializer |
The materializer to store the artifact. |
required |
artifact_store_id |
UUID |
ID of the artifact store in which the artifact should be stored. |
required |
extract_metadata |
bool |
If artifact metadata should be extracted and returned. |
required |
include_visualizations |
bool |
If artifact visualizations should be generated. |
required |
Returns:
| Type | Description |
|---|---|
UUID |
The ID of the published artifact. |
Source code in zenml/utils/artifact_utils.py
def upload_artifact(
name: str,
data: Any,
materializer: "BaseMaterializer",
artifact_store_id: "UUID",
extract_metadata: bool,
include_visualizations: bool,
) -> "UUID":
"""Upload and publish an artifact.
Args:
name: The name of the artifact.
data: The artifact data.
materializer: The materializer to store the artifact.
artifact_store_id: ID of the artifact store in which the artifact should
be stored.
extract_metadata: If artifact metadata should be extracted and returned.
include_visualizations: If artifact visualizations should be generated.
Returns:
The ID of the published artifact.
"""
data_type = type(data)
materializer.validate_type_compatibility(data_type)
materializer.save(data)
visualizations: List[VisualizationModel] = []
if include_visualizations:
try:
vis_data = materializer.save_visualizations(data)
for vis_uri, vis_type in vis_data.items():
vis_model = VisualizationModel(
type=vis_type,
uri=vis_uri,
)
visualizations.append(vis_model)
except Exception as e:
logger.warning(
f"Failed to save visualization for output artifact '{name}': "
f"{e}"
)
artifact_metadata: Dict[str, "MetadataType"] = {}
if extract_metadata:
try:
artifact_metadata = materializer.extract_full_metadata(data)
step_context = get_step_context()
user_provided_metadata = step_context.get_output_metadata(name)
artifact_metadata.update(user_provided_metadata)
except Exception as e:
logger.warning(
f"Failed to extract metadata for output artifact '{name}': {e}"
)
artifact = ArtifactRequestModel(
name=name,
type=materializer.ASSOCIATED_ARTIFACT_TYPE,
uri=materializer.uri,
materializer=source_utils.resolve(materializer.__class__),
data_type=source_utils.resolve(data_type),
user=Client().active_user.id,
workspace=Client().active_workspace.id,
artifact_store_id=artifact_store_id,
visualizations=visualizations,
)
response = Client().zen_store.create_artifact(artifact=artifact)
if artifact_metadata:
Client().create_run_metadata(
metadata=artifact_metadata, artifact_id=response.id
)
return response.id
code_repository_utils
Utilities for code repositories.
find_active_code_repository(path=None)
Find the active code repository for a given path.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path |
Optional[str] |
Path at which to look for the code repository. If not given, the source root will be used. |
None |
Returns:
| Type | Description |
|---|---|
Optional[LocalRepositoryContext] |
The local repository context active at that path or None. |
Source code in zenml/utils/code_repository_utils.py
def find_active_code_repository(
path: Optional[str] = None,
) -> Optional["LocalRepositoryContext"]:
"""Find the active code repository for a given path.
Args:
path: Path at which to look for the code repository. If not given, the
source root will be used.
Returns:
The local repository context active at that path or None.
"""
global _CODE_REPOSITORY_CACHE
from zenml.client import Client
from zenml.code_repositories import BaseCodeRepository
path = path or source_utils.get_source_root()
path = os.path.abspath(path)
if path in _CODE_REPOSITORY_CACHE:
return _CODE_REPOSITORY_CACHE[path]
for model in depaginate(list_method=Client().list_code_repositories):
try:
repo = BaseCodeRepository.from_model(model)
except Exception:
logger.debug(
"Failed to instantiate code repository class.", exc_info=True
)
continue
local_context = repo.get_local_context(path)
if local_context:
_CODE_REPOSITORY_CACHE[path] = local_context
return local_context
return None
set_custom_local_repository(root, commit, repo)
Manually defines a local repository for a path.
To explain what this function does we need to take a dive into source
resolving and what happens inside the Docker image entrypoint:
* When trying to resolve an object to a source, we first determine whether
the file is a user file or not.
* If the file is a user file, we check if that user file is inside a clean
code repository using the
code_repository_utils.find_active_code_repository(...) function. If that
is the case, the object will be resolved to a CodeRepositorySource which
includes additional information about the current commit and the ID of the
code repository.
* The code_repository_utils.find_active_code_repository(...) uses the
code repository implementation classes to check whether the code repository
"exists" at that local path. For git repositories, this check might look as
follows: The code repository first checks if there is a git repository at
that path or in any parent directory. If there is, the remote URLs of this
git repository will be checked to see if one matches the URL defined for
the code repository.
* When running a step inside a Docker image, ZenML potentially downloads
files from a code repository. This usually does not download the entire
repository (and in the case of git might not download a .git directory which
defines a local git repository) but only specific files. If we now try to
resolve any object while running in this container, it will not get resolved
to a CodeRepositorySource as
code_repository_utils.find_active_code_repository(...) won't find an
active repository. As we downloaded these files, we however know that they
belong to a certain code repository at a specific commit, and that's what we
can define using this function.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
root |
str |
The repository root. |
required |
commit |
str |
The commit of the repository. |
required |
repo |
BaseCodeRepository |
The code repository associated with the local repository. |
required |
Source code in zenml/utils/code_repository_utils.py
def set_custom_local_repository(
root: str, commit: str, repo: "BaseCodeRepository"
) -> None:
"""Manually defines a local repository for a path.
To explain what this function does we need to take a dive into source
resolving and what happens inside the Docker image entrypoint:
* When trying to resolve an object to a source, we first determine whether
the file is a user file or not.
* If the file is a user file, we check if that user file is inside a clean
code repository using the
`code_repository_utils.find_active_code_repository(...)` function. If that
is the case, the object will be resolved to a `CodeRepositorySource` which
includes additional information about the current commit and the ID of the
code repository.
* The `code_repository_utils.find_active_code_repository(...)` uses the
code repository implementation classes to check whether the code repository
"exists" at that local path. For git repositories, this check might look as
follows: The code repository first checks if there is a git repository at
that path or in any parent directory. If there is, the remote URLs of this
git repository will be checked to see if one matches the URL defined for
the code repository.
* When running a step inside a Docker image, ZenML potentially downloads
files from a code repository. This usually does not download the entire
repository (and in the case of git might not download a .git directory which
defines a local git repository) but only specific files. If we now try to
resolve any object while running in this container, it will not get resolved
to a `CodeRepositorySource` as
`code_repository_utils.find_active_code_repository(...)` won't find an
active repository. As we downloaded these files, we however know that they
belong to a certain code repository at a specific commit, and that's what we
can define using this function.
Args:
root: The repository root.
commit: The commit of the repository.
repo: The code repository associated with the local repository.
"""
from zenml.utils.downloaded_repository_context import (
_DownloadedRepositoryContext,
)
global _CODE_REPOSITORY_CACHE
path = os.path.abspath(source_utils.get_source_root())
_CODE_REPOSITORY_CACHE[path] = _DownloadedRepositoryContext(
code_repository_id=repo.id, root=root, commit=commit
)
daemon
Utility functions to start/stop daemon processes.
This is only implemented for UNIX systems and therefore doesn't work on Windows. Based on https://www.jejik.com/articles/2007/02/a_simple_unix_linux_daemon_in_python/
check_if_daemon_is_running(pid_file)
Checks whether a daemon process indicated by the PID file is running.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
pid_file |
str |
Path to file containing the PID of the daemon process to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if the daemon process is running, otherwise False. |
Source code in zenml/utils/daemon.py
def check_if_daemon_is_running(pid_file: str) -> bool:
"""Checks whether a daemon process indicated by the PID file is running.
Args:
pid_file: Path to file containing the PID of the daemon
process to check.
Returns:
True if the daemon process is running, otherwise False.
"""
return get_daemon_pid_if_running(pid_file) is not None
daemonize(pid_file, log_file=None, working_directory='/')
Decorator that executes the decorated function as a daemon process.
Use this decorator to easily transform any function into a daemon process.
For example,
import time
from zenml.utils.daemon import daemonize
@daemonize(log_file='/tmp/daemon.log', pid_file='/tmp/daemon.pid')
def sleeping_daemon(period: int) -> None:
print(f"I'm a daemon! I will sleep for {period} seconds.")
time.sleep(period)
print("Done sleeping, flying away.")
sleeping_daemon(period=30)
print("I'm the daemon's parent!.")
time.sleep(10) # just to prove that the daemon is running in parallel
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
pid_file |
str |
a file where the PID of the daemon process will be stored. |
required |
log_file |
Optional[str] |
file where stdout and stderr are redirected for the daemon process. If not supplied, the daemon will be silenced (i.e. have its stdout/stderr redirected to /dev/null). |
None |
working_directory |
str |
working directory for the daemon process, defaults to the root directory. |
'/' |
Returns:
| Type | Description |
|---|---|
Callable[[~F], ~F] |
Decorated function that, when called, will detach from the current process and continue executing in the background, as a daemon process. |
Source code in zenml/utils/daemon.py
def daemonize(
pid_file: str,
log_file: Optional[str] = None,
working_directory: str = "/",
) -> Callable[[F], F]:
"""Decorator that executes the decorated function as a daemon process.
Use this decorator to easily transform any function into a daemon
process.
For example,
```python
import time
from zenml.utils.daemon import daemonize
@daemonize(log_file='/tmp/daemon.log', pid_file='/tmp/daemon.pid')
def sleeping_daemon(period: int) -> None:
print(f"I'm a daemon! I will sleep for {period} seconds.")
time.sleep(period)
print("Done sleeping, flying away.")
sleeping_daemon(period=30)
print("I'm the daemon's parent!.")
time.sleep(10) # just to prove that the daemon is running in parallel
```
Args:
pid_file: a file where the PID of the daemon process will
be stored.
log_file: file where stdout and stderr are redirected for the daemon
process. If not supplied, the daemon will be silenced (i.e. have
its stdout/stderr redirected to /dev/null).
working_directory: working directory for the daemon process,
defaults to the root directory.
Returns:
Decorated function that, when called, will detach from the current
process and continue executing in the background, as a daemon
process.
"""
def inner_decorator(_func: F) -> F:
def daemon(*args: Any, **kwargs: Any) -> None:
"""Standard daemonization of a process.
Args:
*args: Arguments to be passed to the decorated function.
**kwargs: Keyword arguments to be passed to the decorated
function.
"""
if sys.platform == "win32":
logger.error(
"Daemon functionality is currently not supported on Windows."
)
else:
run_as_daemon(
_func,
log_file=log_file,
pid_file=pid_file,
working_directory=working_directory,
*args,
**kwargs,
)
return cast(F, daemon)
return inner_decorator
get_daemon_pid_if_running(pid_file)
Read and return the PID value from a PID file.
It does this if the daemon process tracked by the PID file is running.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
pid_file |
str |
Path to file containing the PID of the daemon process to check. |
required |
Returns:
| Type | Description |
|---|---|
Optional[int] |
The PID of the daemon process if it is running, otherwise None. |
Source code in zenml/utils/daemon.py
def get_daemon_pid_if_running(pid_file: str) -> Optional[int]:
"""Read and return the PID value from a PID file.
It does this if the daemon process tracked by the PID file is running.
Args:
pid_file: Path to file containing the PID of the daemon
process to check.
Returns:
The PID of the daemon process if it is running, otherwise None.
"""
try:
with open(pid_file, "r") as f:
pid = int(f.read().strip())
except (IOError, FileNotFoundError):
logger.debug(
f"Daemon PID file '{pid_file}' does not exist or cannot be read."
)
return None
if not pid or not psutil.pid_exists(pid):
logger.debug(f"Daemon with PID '{pid}' is no longer running.")
return None
logger.debug(f"Daemon with PID '{pid}' is running.")
return pid
run_as_daemon(daemon_function, *args, *, pid_file, log_file=None, working_directory='/', **kwargs)
Runs a function as a daemon process.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
daemon_function |
~F |
The function to run as a daemon. |
required |
pid_file |
str |
Path to file in which to store the PID of the daemon process. |
required |
log_file |
Optional[str] |
Optional file to which the daemons stdout/stderr will be redirected to. |
None |
working_directory |
str |
Working directory for the daemon process, defaults to the root directory. |
'/' |
args |
Any |
Positional arguments to pass to the daemon function. |
() |
kwargs |
Any |
Keyword arguments to pass to the daemon function. |
{} |
Exceptions:
| Type | Description |
|---|---|
FileExistsError |
If the PID file already exists. |
Source code in zenml/utils/daemon.py
def run_as_daemon(
daemon_function: F,
*args: Any,
pid_file: str,
log_file: Optional[str] = None,
working_directory: str = "/",
**kwargs: Any,
) -> None:
"""Runs a function as a daemon process.
Args:
daemon_function: The function to run as a daemon.
pid_file: Path to file in which to store the PID of the daemon
process.
log_file: Optional file to which the daemons stdout/stderr will be
redirected to.
working_directory: Working directory for the daemon process,
defaults to the root directory.
args: Positional arguments to pass to the daemon function.
kwargs: Keyword arguments to pass to the daemon function.
Raises:
FileExistsError: If the PID file already exists.
"""
# convert to absolute path as we will change working directory later
if pid_file:
pid_file = os.path.abspath(pid_file)
if log_file:
log_file = os.path.abspath(log_file)
# create parent directory if necessary
dir_name = os.path.dirname(pid_file)
if not os.path.exists(dir_name):
os.makedirs(dir_name)
# check if PID file exists
if pid_file and os.path.exists(pid_file):
pid = get_daemon_pid_if_running(pid_file)
if pid:
raise FileExistsError(
f"The PID file '{pid_file}' already exists and a daemon "
f"process with the same PID '{pid}' is already running."
f"Please remove the PID file or kill the daemon process "
f"before starting a new daemon."
)
logger.warning(
f"Removing left over PID file '{pid_file}' from a previous "
f"daemon process that didn't shut down correctly."
)
os.remove(pid_file)
# first fork
try:
pid = os.fork()
if pid > 0:
# this is the process that called `run_as_daemon` so we
# wait for the child process to finish to avoid creating
# zombie processes. Then we simply return so the current process
# can continue what it was doing.
os.wait()
return
except OSError as e:
logger.error("Unable to fork (error code: %d)", e.errno)
sys.exit(1)
# decouple from parent environment
os.chdir(working_directory)
os.setsid()
os.umask(0o22)
# second fork
try:
pid = os.fork()
if pid > 0:
# this is the parent of the future daemon process, kill it
# so the daemon gets adopted by the init process.
# we use os._exit here to prevent the inherited code from
# catching the SystemExit exception and doing something else.
os._exit(0)
except OSError as e:
sys.stderr.write(f"Unable to fork (error code: {e.errno})")
# we use os._exit here to prevent the inherited code from
# catching the SystemExit exception and doing something else.
os._exit(1)
# redirect standard file descriptors to devnull (or the given logfile)
devnull = "/dev/null"
if hasattr(os, "devnull"):
devnull = os.devnull
devnull_fd = os.open(devnull, os.O_RDWR)
log_fd = (
os.open(log_file, os.O_CREAT | os.O_RDWR | os.O_APPEND)
if log_file
else None
)
out_fd = log_fd or devnull_fd
try:
os.dup2(devnull_fd, sys.stdin.fileno())
except io.UnsupportedOperation:
# stdin is not a file descriptor
pass
try:
os.dup2(out_fd, sys.stdout.fileno())
except io.UnsupportedOperation:
# stdout is not a file descriptor
pass
try:
os.dup2(out_fd, sys.stderr.fileno())
except io.UnsupportedOperation:
# stderr is not a file descriptor
pass
if pid_file:
# write the PID file
with open(pid_file, "w+") as f:
f.write(f"{os.getpid()}\n")
# register actions in case this process exits/gets killed
def cleanup() -> None:
"""Daemon cleanup."""
sys.stderr.write("Cleanup: terminating children processes...\n")
terminate_children()
if pid_file and os.path.exists(pid_file):
sys.stderr.write(f"Cleanup: removing PID file {pid_file}...\n")
os.remove(pid_file)
sys.stderr.flush()
def sighndl(signum: int, frame: Optional[types.FrameType]) -> None:
"""Daemon signal handler.
Args:
signum: Signal number.
frame: Frame object.
"""
sys.stderr.write(f"Handling signal {signum}...\n")
cleanup()
signal.signal(signal.SIGTERM, sighndl)
signal.signal(signal.SIGINT, sighndl)
atexit.register(cleanup)
# finally run the actual daemon code
daemon_function(*args, **kwargs)
sys.exit(0)
stop_daemon(pid_file)
Stops a daemon process.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
pid_file |
str |
Path to file containing the PID of the daemon process to kill. |
required |
Source code in zenml/utils/daemon.py
def stop_daemon(pid_file: str) -> None:
"""Stops a daemon process.
Args:
pid_file: Path to file containing the PID of the daemon process to
kill.
"""
try:
with open(pid_file, "r") as f:
pid = int(f.read().strip())
except (IOError, FileNotFoundError):
logger.warning("Daemon PID file '%s' does not exist.", pid_file)
return
if psutil.pid_exists(pid):
process = psutil.Process(pid)
process.terminate()
else:
logger.warning("PID from '%s' does not exist.", pid_file)
terminate_children()
Terminate all processes that are children of the currently running process.
Source code in zenml/utils/daemon.py
def terminate_children() -> None:
"""Terminate all processes that are children of the currently running process."""
pid = os.getpid()
try:
parent = psutil.Process(pid)
except psutil.Error:
# could not find parent process id
return
children = parent.children(recursive=False)
for p in children:
sys.stderr.write(
f"Terminating child process with PID {p.pid}...\n"
)
p.terminate()
_, alive = psutil.wait_procs(
children, timeout=CHILD_PROCESS_WAIT_TIMEOUT
)
for p in alive:
sys.stderr.write(f"Killing child process with PID {p.pid}...\n")
p.kill()
_, alive = psutil.wait_procs(
children, timeout=CHILD_PROCESS_WAIT_TIMEOUT
)
dashboard_utils
Utility class to help with interacting with the dashboard.
get_base_url()
Function to get the base workspace-scoped url.
Returns:
| Type | Description |
|---|---|
Optional[str] |
the base url if the client is using a rest zen store, else None |
Source code in zenml/utils/dashboard_utils.py
def get_base_url() -> Optional[str]:
"""Function to get the base workspace-scoped url.
Returns:
the base url if the client is using a rest zen store, else None
"""
client = Client()
if client.zen_store.type == StoreType.REST:
url = (
client.zen_store.url
+ f"{constants.WORKSPACES}/{client.active_workspace.name}"
)
return url
return None
get_component_url(component)
Function to get the dashboard URL of a given component model.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
component |
ComponentResponseModel |
the response model of the given component. |
required |
Returns:
| Type | Description |
|---|---|
Optional[str] |
the URL to the component if the dashboard is available, else None. |
Source code in zenml/utils/dashboard_utils.py
def get_component_url(component: ComponentResponseModel) -> Optional[str]:
"""Function to get the dashboard URL of a given component model.
Args:
component: the response model of the given component.
Returns:
the URL to the component if the dashboard is available, else None.
"""
base_url = get_base_url()
if base_url:
return (
base_url
+ f"{constants.STACK_COMPONENTS}/{component.type.value}/{component.id}/configuration"
)
return None
get_run_url(run)
Function to get the dashboard URL of a given pipeline run.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
run |
PipelineRunResponseModel |
the response model of the given pipeline run. |
required |
Returns:
| Type | Description |
|---|---|
Optional[str] |
the URL to the pipeline run if the dashboard is available, else None. |
Source code in zenml/utils/dashboard_utils.py
def get_run_url(run: PipelineRunResponseModel) -> Optional[str]:
"""Function to get the dashboard URL of a given pipeline run.
Args:
run: the response model of the given pipeline run.
Returns:
the URL to the pipeline run if the dashboard is available, else None.
"""
base_url = get_base_url()
if base_url:
if run.pipeline:
return f"{base_url}{constants.PIPELINES}/{run.pipeline.id}{constants.RUNS}/{run.id}/dag"
else:
return f"{base_url}/all-runs/{run.id}/dag"
return None
get_stack_url(stack)
Function to get the dashboard URL of a given stack model.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack |
StackResponseModel |
the response model of the given stack. |
required |
Returns:
| Type | Description |
|---|---|
Optional[str] |
the URL to the stack if the dashboard is available, else None. |
Source code in zenml/utils/dashboard_utils.py
def get_stack_url(stack: StackResponseModel) -> Optional[str]:
"""Function to get the dashboard URL of a given stack model.
Args:
stack: the response model of the given stack.
Returns:
the URL to the stack if the dashboard is available, else None.
"""
base_url = get_base_url()
if base_url:
return base_url + f"{constants.STACKS}/{stack.id}/configuration"
return None
show_dashboard(url)
Show the ZenML dashboard at the given URL.
In native environments, the dashboard is opened in the default browser. In notebook environments, the dashboard is embedded in an iframe.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
url |
str |
URL of the ZenML dashboard. |
required |
Source code in zenml/utils/dashboard_utils.py
def show_dashboard(url: str) -> None:
"""Show the ZenML dashboard at the given URL.
In native environments, the dashboard is opened in the default browser.
In notebook environments, the dashboard is embedded in an iframe.
Args:
url: URL of the ZenML dashboard.
"""
environment = get_environment()
if environment in (EnvironmentType.NOTEBOOK, EnvironmentType.COLAB):
from IPython.core.display import display
from IPython.display import IFrame
display(IFrame(src=url, width="100%", height=720))
elif environment in (EnvironmentType.NATIVE, EnvironmentType.WSL):
if constants.handle_bool_env_var(
constants.ENV_AUTO_OPEN_DASHBOARD, default=True
):
try:
import webbrowser
if environment == EnvironmentType.WSL:
webbrowser.get("wslview %s").open(url)
else:
webbrowser.open(url)
logger.info(
"Automatically opening the dashboard in your "
"browser. To disable this, set the env variable "
"AUTO_OPEN_DASHBOARD=false."
)
except Exception as e:
logger.error(e)
else:
logger.info(
"To open the dashboard in a browser automatically, "
"set the env variable AUTO_OPEN_DASHBOARD=true."
)
else:
logger.info(f"The ZenML dashboard is available at {url}.")
deprecation_utils
Deprecation utilities.
deprecate_pydantic_attributes(*attributes)
Utility function for deprecating and migrating pydantic attributes.
Usage: To use this, you can specify it on any pydantic BaseModel subclass like this (all the deprecated attributes need to be non-required):
from pydantic import BaseModel
from typing import Optional
class MyModel(BaseModel):
deprecated: Optional[int] = None
old_name: Optional[str] = None
new_name: str
_deprecation_validator = deprecate_pydantic_attributes(
"deprecated", ("old_name", "new_name")
)
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
*attributes |
Union[str, Tuple[str, str]] |
List of attributes to deprecate. This is either the name of the attribute to deprecate, or a tuple containing the name of the deprecated attribute and it's replacement. |
() |
Returns:
| Type | Description |
|---|---|
AnyClassMethod |
Pydantic validator class method to be used on BaseModel subclasses to deprecate or migrate attributes. |
Source code in zenml/utils/deprecation_utils.py
def deprecate_pydantic_attributes(
*attributes: Union[str, Tuple[str, str]]
) -> "AnyClassMethod":
"""Utility function for deprecating and migrating pydantic attributes.
**Usage**:
To use this, you can specify it on any pydantic BaseModel subclass like
this (all the deprecated attributes need to be non-required):
```python
from pydantic import BaseModel
from typing import Optional
class MyModel(BaseModel):
deprecated: Optional[int] = None
old_name: Optional[str] = None
new_name: str
_deprecation_validator = deprecate_pydantic_attributes(
"deprecated", ("old_name", "new_name")
)
```
Args:
*attributes: List of attributes to deprecate. This is either the name
of the attribute to deprecate, or a tuple containing the name of
the deprecated attribute and it's replacement.
Returns:
Pydantic validator class method to be used on BaseModel subclasses
to deprecate or migrate attributes.
"""
@root_validator(pre=True, allow_reuse=True)
def _deprecation_validator(
cls: Type[BaseModel], values: Dict[str, Any]
) -> Dict[str, Any]:
"""Pydantic validator function for deprecating pydantic attributes.
Args:
cls: The class on which the attributes are defined.
values: All values passed at model initialization.
Raises:
AssertionError: If either the deprecated or replacement attribute
don't exist.
TypeError: If the deprecated attribute is a required attribute.
ValueError: If the deprecated attribute and replacement attribute
contain different values.
Returns:
Input values with potentially migrated values.
"""
previous_deprecation_warnings: Set[str] = getattr(
cls, PREVIOUS_DEPRECATION_WARNINGS_ATTRIBUTE, set()
)
def _warn(message: str, attribute: str) -> None:
"""Logs and raises a warning for a deprecated attribute.
Args:
message: The warning message.
attribute: The name of the attribute.
"""
if attribute not in previous_deprecation_warnings:
logger.warning(message)
previous_deprecation_warnings.add(attribute)
warnings.warn(
message,
DeprecationWarning,
)
for attribute in attributes:
if isinstance(attribute, str):
deprecated_attribute = attribute
replacement_attribute = None
else:
deprecated_attribute, replacement_attribute = attribute
assert (
replacement_attribute in cls.__fields__
), f"Unable to find attribute {replacement_attribute}."
assert (
deprecated_attribute in cls.__fields__
), f"Unable to find attribute {deprecated_attribute}."
if cls.__fields__[deprecated_attribute].required:
raise TypeError(
f"Unable to deprecate attribute '{deprecated_attribute}' "
f"of class {cls.__name__}. In order to deprecate an "
"attribute, it needs to be a non-required attribute. "
"To do so, mark the attribute with an `Optional[...] type "
"annotation."
)
if values.get(deprecated_attribute, None) is None:
continue
if replacement_attribute is None:
_warn(
message=f"The attribute `{deprecated_attribute}` of class "
f"`{cls.__name__}` will be deprecated soon.",
attribute=deprecated_attribute,
)
continue
_warn(
message=f"The attribute `{deprecated_attribute}` of class "
f"`{cls.__name__}` will be deprecated soon. Use the "
f"attribute `{replacement_attribute}` instead.",
attribute=deprecated_attribute,
)
if values.get(replacement_attribute, None) is None:
logger.debug(
"Migrating value of deprecated attribute %s to "
"replacement attribute %s.",
deprecated_attribute,
replacement_attribute,
)
values[replacement_attribute] = values.pop(
deprecated_attribute
)
elif values[deprecated_attribute] != values[replacement_attribute]:
raise ValueError(
"Got different values for deprecated attribute "
f"{deprecated_attribute} and replacement "
f"attribute {replacement_attribute}."
)
else:
# Both values are identical, no need to do anything
pass
setattr(
cls,
PREVIOUS_DEPRECATION_WARNINGS_ATTRIBUTE,
previous_deprecation_warnings,
)
return values
return _deprecation_validator
dict_utils
Util functions for dictionaries.
recursive_update(original, update)
Recursively updates a dictionary.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
original |
Dict[str, Any] |
The dictionary to update. |
required |
update |
Dict[str, Any] |
The dictionary containing the updated values. |
required |
Returns:
| Type | Description |
|---|---|
Dict[str, Any] |
The updated dictionary. |
Source code in zenml/utils/dict_utils.py
def recursive_update(
original: Dict[str, Any], update: Dict[str, Any]
) -> Dict[str, Any]:
"""Recursively updates a dictionary.
Args:
original: The dictionary to update.
update: The dictionary containing the updated values.
Returns:
The updated dictionary.
"""
for key, value in update.items():
if isinstance(value, Dict):
original_value = original.get(key, None) or {}
if isinstance(original_value, Dict):
original[key] = recursive_update(original_value, value)
else:
original[key] = value
else:
original[key] = value
return original
remove_none_values(dict_, recursive=False)
Removes all key-value pairs with None value.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dict_ |
Dict[str, Any] |
The dict from which the key-value pairs should be removed. |
required |
recursive |
bool |
If |
False |
Returns:
| Type | Description |
|---|---|
Dict[str, Any] |
The updated dictionary. |
Source code in zenml/utils/dict_utils.py
def remove_none_values(
dict_: Dict[str, Any], recursive: bool = False
) -> Dict[str, Any]:
"""Removes all key-value pairs with `None` value.
Args:
dict_: The dict from which the key-value pairs should be removed.
recursive: If `True`, will recursively remove `None` values in all
child dicts.
Returns:
The updated dictionary.
"""
def _maybe_recurse(value: Any) -> Any:
"""Calls `remove_none_values` recursively if required.
Args:
value: A dictionary value.
Returns:
The updated dictionary value.
"""
if recursive and isinstance(value, Dict):
return remove_none_values(value, recursive=True)
else:
return value
return {k: _maybe_recurse(v) for k, v in dict_.items() if v is not None}
docker_utils
Utility functions relating to Docker.
build_image(image_name, dockerfile, build_context_root=None, dockerignore=None, extra_files=(), **custom_build_options)
Builds a docker image.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
image_name |
str |
The name to use for the built docker image. |
required |
dockerfile |
Union[str, List[str]] |
Path to a dockerfile or a list of strings representing the Dockerfile lines/commands. |
required |
build_context_root |
Optional[str] |
Optional path to a directory that will be sent to the Docker daemon as build context. If left empty, the Docker build context will be empty. |
None |
dockerignore |
Optional[str] |
Optional path to a dockerignore file. If no value is
given, the .dockerignore in the root of the build context will be
used if it exists. Otherwise, all files inside |
None |
extra_files |
Sequence[Tuple[str, str]] |
Additional files to include in the build context. The files should be passed as a tuple (filepath_inside_build_context, file_content) and will overwrite existing files in the build context if they share the same path. |
() |
**custom_build_options |
Any |
Additional options that will be passed unmodified to the Docker build call when building the image. You can use this to for example specify build args or a target stage. See https://docker-py.readthedocs.io/en/stable/images.html#docker.models.images.ImageCollection.build for a full list of available options. |
{} |
Source code in zenml/utils/docker_utils.py
def build_image(
image_name: str,
dockerfile: Union[str, List[str]],
build_context_root: Optional[str] = None,
dockerignore: Optional[str] = None,
extra_files: Sequence[Tuple[str, str]] = (),
**custom_build_options: Any,
) -> None:
"""Builds a docker image.
Args:
image_name: The name to use for the built docker image.
dockerfile: Path to a dockerfile or a list of strings representing the
Dockerfile lines/commands.
build_context_root: Optional path to a directory that will be sent to
the Docker daemon as build context. If left empty, the Docker build
context will be empty.
dockerignore: Optional path to a dockerignore file. If no value is
given, the .dockerignore in the root of the build context will be
used if it exists. Otherwise, all files inside `build_context_root`
are included in the build context.
extra_files: Additional files to include in the build context. The
files should be passed as a tuple
(filepath_inside_build_context, file_content) and will overwrite
existing files in the build context if they share the same path.
**custom_build_options: Additional options that will be passed
unmodified to the Docker build call when building the image. You
can use this to for example specify build args or a target stage.
See https://docker-py.readthedocs.io/en/stable/images.html#docker.models.images.ImageCollection.build
for a full list of available options.
"""
if isinstance(dockerfile, str):
dockerfile_contents = io_utils.read_file_contents_as_string(dockerfile)
logger.info("Using Dockerfile `%s`.", os.path.abspath(dockerfile))
else:
dockerfile_contents = "\n".join(dockerfile)
build_context = _create_custom_build_context(
dockerfile_contents=dockerfile_contents,
build_context_root=build_context_root,
dockerignore=dockerignore,
extra_files=extra_files,
)
build_options = {
"rm": False, # don't remove intermediate containers to improve caching
"pull": True, # always pull parent images
**custom_build_options,
}
logger.info("Building Docker image `%s`.", image_name)
logger.debug("Docker build options: %s", build_options)
logger.info("Building the image might take a while...")
docker_client = DockerClient.from_env()
# We use the client api directly here, so we can stream the logs
output_stream = docker_client.images.client.api.build(
fileobj=build_context,
custom_context=True,
tag=image_name,
**build_options,
)
_process_stream(output_stream)
logger.info("Finished building Docker image `%s`.", image_name)
check_docker()
Checks if Docker is installed and running.
Returns:
| Type | Description |
|---|---|
bool |
|
Source code in zenml/utils/docker_utils.py
def check_docker() -> bool:
"""Checks if Docker is installed and running.
Returns:
`True` if Docker is installed, `False` otherwise.
"""
# Try to ping Docker, to see if it's running
try:
docker_client = DockerClient.from_env()
docker_client.ping()
return True
except Exception:
logger.debug("Docker is not running.", exc_info=True)
return False
get_image_digest(image_name)
Gets the digest of an image.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
image_name |
str |
Name of the image to get the digest for. |
required |
Returns:
| Type | Description |
|---|---|
Optional[str] |
Returns the repo digest for the given image if there exists exactly one.
If there are zero or multiple repo digests, returns |
Source code in zenml/utils/docker_utils.py
def get_image_digest(image_name: str) -> Optional[str]:
"""Gets the digest of an image.
Args:
image_name: Name of the image to get the digest for.
Returns:
Returns the repo digest for the given image if there exists exactly one.
If there are zero or multiple repo digests, returns `None`.
"""
docker_client = DockerClient.from_env()
image = docker_client.images.get(image_name)
repo_digests = image.attrs["RepoDigests"]
if len(repo_digests) == 1:
return cast(str, repo_digests[0])
else:
logger.debug(
"Found zero or more repo digests for docker image '%s': %s",
image_name,
repo_digests,
)
return None
is_local_image(image_name)
Returns whether an image was pulled from a registry or not.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
image_name |
str |
Name of the image to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
|
Source code in zenml/utils/docker_utils.py
def is_local_image(image_name: str) -> bool:
"""Returns whether an image was pulled from a registry or not.
Args:
image_name: Name of the image to check.
Returns:
`True` if the image was pulled from a registry, `False` otherwise.
"""
docker_client = DockerClient.from_env()
images = docker_client.images.list(name=image_name)
if images:
# An image with this name is available locally -> now check whether it
# was pulled from a repo or built locally (in which case the repo
# digest is empty)
return get_image_digest(image_name) is None
else:
# no image with this name found locally
return False
push_image(image_name, docker_client=None)
Pushes an image to a container registry.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
image_name |
str |
The full name (including a tag) of the image to push. |
required |
docker_client |
Optional[docker.client.DockerClient] |
Optional Docker client to use for pushing the image. If no client is given, a new client will be created using the default Docker environment. |
None |
Returns:
| Type | Description |
|---|---|
str |
The Docker repository digest of the pushed image. |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If fetching the repository digest of the image failed. |
Source code in zenml/utils/docker_utils.py
def push_image(
image_name: str, docker_client: Optional[DockerClient] = None
) -> str:
"""Pushes an image to a container registry.
Args:
image_name: The full name (including a tag) of the image to push.
docker_client: Optional Docker client to use for pushing the image. If
no client is given, a new client will be created using the default
Docker environment.
Returns:
The Docker repository digest of the pushed image.
Raises:
RuntimeError: If fetching the repository digest of the image failed.
"""
logger.info("Pushing Docker image `%s`.", image_name)
docker_client = docker_client or DockerClient.from_env()
output_stream = docker_client.images.push(image_name, stream=True)
aux_info = _process_stream(output_stream)
logger.info("Finished pushing Docker image.")
image_name_without_tag, _ = image_name.rsplit(":", maxsplit=1)
for info in reversed(aux_info):
try:
repo_digest = info["Digest"]
return f"{image_name_without_tag}@{repo_digest}"
except KeyError:
pass
else:
raise RuntimeError(
f"Unable to find repo digest after pushing image {image_name}."
)
tag_image(image_name, target)
Tags an image.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
image_name |
str |
The name of the image to tag. |
required |
target |
str |
The full target name including a tag. |
required |
Source code in zenml/utils/docker_utils.py
def tag_image(image_name: str, target: str) -> None:
"""Tags an image.
Args:
image_name: The name of the image to tag.
target: The full target name including a tag.
"""
docker_client = DockerClient.from_env()
image = docker_client.images.get(image_name)
image.tag(target)
downloaded_repository_context
Downloaded code repository.
enum_utils
Util functions for enums.
StrEnum (str, Enum)
Base enum type for string enum values.
Source code in zenml/utils/enum_utils.py
class StrEnum(str, Enum):
"""Base enum type for string enum values."""
def __str__(self) -> str:
"""Returns the enum string value.
Returns:
The enum string value.
"""
return self.value # type: ignore
@classmethod
def names(cls) -> List[str]:
"""Get all enum names as a list of strings.
Returns:
A list of all enum names.
"""
return [c.name for c in cls]
@classmethod
def values(cls) -> List[str]:
"""Get all enum values as a list of strings.
Returns:
A list of all enum values.
"""
return [c.value for c in cls]
filesync_model
Filesync utils for ZenML.
FileSyncModel (BaseModel)
pydantic-model
Pydantic model synchronized with a configuration file.
Use this class as a base Pydantic model that is automatically synchronized with a configuration file on disk.
This class overrides the setattr and getattr magic methods to ensure that the FileSyncModel instance acts as an in-memory cache of the information stored in the associated configuration file.
Source code in zenml/utils/filesync_model.py
class FileSyncModel(BaseModel):
"""Pydantic model synchronized with a configuration file.
Use this class as a base Pydantic model that is automatically synchronized
with a configuration file on disk.
This class overrides the __setattr__ and __getattr__ magic methods to
ensure that the FileSyncModel instance acts as an in-memory cache of the
information stored in the associated configuration file.
"""
_config_file: str
_config_file_timestamp: Optional[float] = None
def __init__(self, config_file: str, **kwargs: Any) -> None:
"""Create a FileSyncModel instance synchronized with a configuration file on disk.
Args:
config_file: configuration file path. If the file exists, the model
will be initialized with the values from the file.
**kwargs: additional keyword arguments to pass to the Pydantic model
constructor. If supplied, these values will override those
loaded from the configuration file.
"""
config_dict = {}
if fileio.exists(config_file):
config_dict = yaml_utils.read_yaml(config_file)
self._config_file = config_file
self._config_file_timestamp = None
config_dict.update(kwargs)
super(FileSyncModel, self).__init__(**config_dict)
# write the configuration file to disk, to reflect new attributes
# and schema changes
self.write_config()
def __setattr__(self, key: str, value: Any) -> None:
"""Sets an attribute on the model and persists it in the configuration file.
Args:
key: attribute name.
value: attribute value.
"""
super(FileSyncModel, self).__setattr__(key, value)
if key.startswith("_"):
return
self.write_config()
def __getattribute__(self, key: str) -> Any:
"""Gets an attribute value for a specific key.
Args:
key: attribute name.
Returns:
attribute value.
"""
if not key.startswith("_") and key in self.__dict__:
self.load_config()
return super(FileSyncModel, self).__getattribute__(key)
def write_config(self) -> None:
"""Writes the model to the configuration file."""
config_dict = json.loads(self.json())
yaml_utils.write_yaml(self._config_file, config_dict)
self._config_file_timestamp = os.path.getmtime(self._config_file)
def load_config(self) -> None:
"""Loads the model from the configuration file on disk."""
if not fileio.exists(self._config_file):
return
# don't reload the configuration if the file hasn't
# been updated since the last load
file_timestamp = os.path.getmtime(self._config_file)
if file_timestamp == self._config_file_timestamp:
return
if self._config_file_timestamp is not None:
logger.info(f"Reloading configuration file {self._config_file}")
# refresh the model from the configuration file values
config_dict = yaml_utils.read_yaml(self._config_file)
for key, value in config_dict.items():
super(FileSyncModel, self).__setattr__(key, value)
self._config_file_timestamp = file_timestamp
class Config:
"""Pydantic configuration class."""
# all attributes with leading underscore are private and therefore
# are mutable and not included in serialization
underscore_attrs_are_private = True
Config
Pydantic configuration class.
Source code in zenml/utils/filesync_model.py
class Config:
"""Pydantic configuration class."""
# all attributes with leading underscore are private and therefore
# are mutable and not included in serialization
underscore_attrs_are_private = True
__getattribute__(self, key)
special
Gets an attribute value for a specific key.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key |
str |
attribute name. |
required |
Returns:
| Type | Description |
|---|---|
Any |
attribute value. |
Source code in zenml/utils/filesync_model.py
def __getattribute__(self, key: str) -> Any:
"""Gets an attribute value for a specific key.
Args:
key: attribute name.
Returns:
attribute value.
"""
if not key.startswith("_") and key in self.__dict__:
self.load_config()
return super(FileSyncModel, self).__getattribute__(key)
__init__(self, config_file, **kwargs)
special
Create a FileSyncModel instance synchronized with a configuration file on disk.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
config_file |
str |
configuration file path. If the file exists, the model will be initialized with the values from the file. |
required |
**kwargs |
Any |
additional keyword arguments to pass to the Pydantic model constructor. If supplied, these values will override those loaded from the configuration file. |
{} |
Source code in zenml/utils/filesync_model.py
def __init__(self, config_file: str, **kwargs: Any) -> None:
"""Create a FileSyncModel instance synchronized with a configuration file on disk.
Args:
config_file: configuration file path. If the file exists, the model
will be initialized with the values from the file.
**kwargs: additional keyword arguments to pass to the Pydantic model
constructor. If supplied, these values will override those
loaded from the configuration file.
"""
config_dict = {}
if fileio.exists(config_file):
config_dict = yaml_utils.read_yaml(config_file)
self._config_file = config_file
self._config_file_timestamp = None
config_dict.update(kwargs)
super(FileSyncModel, self).__init__(**config_dict)
# write the configuration file to disk, to reflect new attributes
# and schema changes
self.write_config()
__setattr__(self, key, value)
special
Sets an attribute on the model and persists it in the configuration file.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key |
str |
attribute name. |
required |
value |
Any |
attribute value. |
required |
Source code in zenml/utils/filesync_model.py
def __setattr__(self, key: str, value: Any) -> None:
"""Sets an attribute on the model and persists it in the configuration file.
Args:
key: attribute name.
value: attribute value.
"""
super(FileSyncModel, self).__setattr__(key, value)
if key.startswith("_"):
return
self.write_config()
load_config(self)
Loads the model from the configuration file on disk.
Source code in zenml/utils/filesync_model.py
def load_config(self) -> None:
"""Loads the model from the configuration file on disk."""
if not fileio.exists(self._config_file):
return
# don't reload the configuration if the file hasn't
# been updated since the last load
file_timestamp = os.path.getmtime(self._config_file)
if file_timestamp == self._config_file_timestamp:
return
if self._config_file_timestamp is not None:
logger.info(f"Reloading configuration file {self._config_file}")
# refresh the model from the configuration file values
config_dict = yaml_utils.read_yaml(self._config_file)
for key, value in config_dict.items():
super(FileSyncModel, self).__setattr__(key, value)
self._config_file_timestamp = file_timestamp
write_config(self)
Writes the model to the configuration file.
Source code in zenml/utils/filesync_model.py
def write_config(self) -> None:
"""Writes the model to the configuration file."""
config_dict = json.loads(self.json())
yaml_utils.write_yaml(self._config_file, config_dict)
self._config_file_timestamp = os.path.getmtime(self._config_file)
git_utils
Utility function to clone a Git repository.
clone_git_repository(url, to_path, branch=None, commit=None)
Clone a Git repository.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
url |
str |
URL of the repository to clone. |
required |
to_path |
str |
Path to clone the repository to. |
required |
branch |
Optional[str] |
Branch to clone. Defaults to "main". |
None |
commit |
Optional[str] |
Commit to checkout. If specified, the branch argument is ignored. |
None |
Returns:
| Type | Description |
|---|---|
Repo |
The cloned repository. |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If the repository could not be cloned. |
Source code in zenml/utils/git_utils.py
def clone_git_repository(
url: str,
to_path: str,
branch: Optional[str] = None,
commit: Optional[str] = None,
) -> Repo:
"""Clone a Git repository.
Args:
url: URL of the repository to clone.
to_path: Path to clone the repository to.
branch: Branch to clone. Defaults to "main".
commit: Commit to checkout. If specified, the branch argument is
ignored.
Returns:
The cloned repository.
Raises:
RuntimeError: If the repository could not be cloned.
"""
os.makedirs(os.path.basename(to_path), exist_ok=True)
try:
if commit:
repo = Repo.clone_from(
url=url,
to_path=to_path,
no_checkout=True,
)
repo.git.checkout(commit)
else:
repo = Repo.clone_from(
url=url,
to_path=to_path,
branch=branch or "main",
)
return repo
except GitCommandError as e:
raise RuntimeError from e
integration_utils
Util functions for integration.
parse_requirement(requirement)
Parse a requirement string into name and extras.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
requirement |
str |
A requirement string. |
required |
Returns:
| Type | Description |
|---|---|
Tuple[Optional[str], Optional[str]] |
A tuple of name and extras. |
Source code in zenml/utils/integration_utils.py
def parse_requirement(requirement: str) -> Tuple[Optional[str], Optional[str]]:
"""Parse a requirement string into name and extras.
Args:
requirement: A requirement string.
Returns:
A tuple of name and extras.
"""
if match := re.match(
r"([a-zA-Z0-9\-_]+)(?![^\[]*\s)(\[[a-zA-Z0-9\-_,]+\])?", requirement
):
name, extras = match.groups()
return (None, None) if " " in requirement else (name, extras)
return None, None
io_utils
Various utility functions for the io module.
copy_dir(source_dir, destination_dir, overwrite=False)
Copies dir from source to destination.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
source_dir |
str |
Path to copy from. |
required |
destination_dir |
str |
Path to copy to. |
required |
overwrite |
bool |
Boolean. If false, function throws an error before overwrite. |
False |
Source code in zenml/utils/io_utils.py
def copy_dir(
source_dir: str, destination_dir: str, overwrite: bool = False
) -> None:
"""Copies dir from source to destination.
Args:
source_dir: Path to copy from.
destination_dir: Path to copy to.
overwrite: Boolean. If false, function throws an error before overwrite.
"""
for source_file in listdir(source_dir):
source_path = os.path.join(source_dir, convert_to_str(source_file))
destination_path = os.path.join(
destination_dir, convert_to_str(source_file)
)
if isdir(source_path):
if source_path == destination_dir:
# if the destination is a subdirectory of the source, we skip
# copying it to avoid an infinite loop.
continue
copy_dir(source_path, destination_path, overwrite)
else:
create_dir_recursive_if_not_exists(
os.path.dirname(destination_path)
)
copy(str(source_path), str(destination_path), overwrite)
create_dir_if_not_exists(dir_path)
Creates directory if it does not exist.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dir_path |
str |
Local path in filesystem. |
required |
Source code in zenml/utils/io_utils.py
def create_dir_if_not_exists(dir_path: str) -> None:
"""Creates directory if it does not exist.
Args:
dir_path: Local path in filesystem.
"""
if not isdir(dir_path):
mkdir(dir_path)
create_dir_recursive_if_not_exists(dir_path)
Creates directory recursively if it does not exist.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dir_path |
str |
Local path in filesystem. |
required |
Source code in zenml/utils/io_utils.py
def create_dir_recursive_if_not_exists(dir_path: str) -> None:
"""Creates directory recursively if it does not exist.
Args:
dir_path: Local path in filesystem.
"""
if not isdir(dir_path):
makedirs(dir_path)
create_file_if_not_exists(file_path, file_contents='{}')
Creates file if it does not exist.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Local path in filesystem. |
required |
file_contents |
str |
Contents of file. |
'{}' |
Source code in zenml/utils/io_utils.py
def create_file_if_not_exists(
file_path: str, file_contents: str = "{}"
) -> None:
"""Creates file if it does not exist.
Args:
file_path: Local path in filesystem.
file_contents: Contents of file.
"""
full_path = Path(file_path)
if not exists(file_path):
create_dir_recursive_if_not_exists(str(full_path.parent))
with open(str(full_path), "w") as f:
f.write(file_contents)
find_files(dir_path, pattern)
Find files in a directory that match pattern.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dir_path |
PathType |
The path to directory. |
required |
pattern |
str |
pattern like *.png. |
required |
Yields:
| Type | Description |
|---|---|
Iterable[str] |
All matching filenames in the directory. |
Source code in zenml/utils/io_utils.py
def find_files(dir_path: "PathType", pattern: str) -> Iterable[str]:
"""Find files in a directory that match pattern.
Args:
dir_path: The path to directory.
pattern: pattern like *.png.
Yields:
All matching filenames in the directory.
"""
for root, _, files in walk(dir_path):
for basename in files:
if fnmatch.fnmatch(convert_to_str(basename), pattern):
filename = os.path.join(
convert_to_str(root), convert_to_str(basename)
)
yield filename
get_global_config_directory()
Gets the global config directory for ZenML.
Returns:
| Type | Description |
|---|---|
str |
The global config directory for ZenML. |
Source code in zenml/utils/io_utils.py
def get_global_config_directory() -> str:
"""Gets the global config directory for ZenML.
Returns:
The global config directory for ZenML.
"""
env_var_path = os.getenv(ENV_ZENML_CONFIG_PATH)
if env_var_path:
return str(Path(env_var_path).resolve())
return click.get_app_dir(APP_NAME)
get_grandparent(dir_path)
Get grandparent of dir.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dir_path |
str |
The path to directory. |
required |
Returns:
| Type | Description |
|---|---|
str |
The input paths parents parent. |
Exceptions:
| Type | Description |
|---|---|
ValueError |
If dir_path does not exist. |
Source code in zenml/utils/io_utils.py
def get_grandparent(dir_path: str) -> str:
"""Get grandparent of dir.
Args:
dir_path: The path to directory.
Returns:
The input paths parents parent.
Raises:
ValueError: If dir_path does not exist.
"""
if not os.path.exists(dir_path):
raise ValueError(f"Path '{dir_path}' does not exist.")
return Path(dir_path).parent.parent.stem
get_parent(dir_path)
Get parent of dir.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dir_path |
str |
The path to directory. |
required |
Returns:
| Type | Description |
|---|---|
str |
Parent (stem) of the dir as a string. |
Exceptions:
| Type | Description |
|---|---|
ValueError |
If dir_path does not exist. |
Source code in zenml/utils/io_utils.py
def get_parent(dir_path: str) -> str:
"""Get parent of dir.
Args:
dir_path: The path to directory.
Returns:
Parent (stem) of the dir as a string.
Raises:
ValueError: If dir_path does not exist.
"""
if not os.path.exists(dir_path):
raise ValueError(f"Path '{dir_path}' does not exist.")
return Path(dir_path).parent.stem
is_remote(path)
Returns True if path exists remotely.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path |
str |
Any path as a string. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if remote path, else False. |
Source code in zenml/utils/io_utils.py
def is_remote(path: str) -> bool:
"""Returns True if path exists remotely.
Args:
path: Any path as a string.
Returns:
True if remote path, else False.
"""
return any(path.startswith(prefix) for prefix in REMOTE_FS_PREFIX)
is_root(path)
Returns true if path has no parent in local filesystem.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path |
str |
Local path in filesystem. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if root, else False. |
Source code in zenml/utils/io_utils.py
def is_root(path: str) -> bool:
"""Returns true if path has no parent in local filesystem.
Args:
path: Local path in filesystem.
Returns:
True if root, else False.
"""
return Path(path).parent == Path(path)
move(source, destination, overwrite=False)
Moves dir or file from source to destination. Can be used to rename.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
source |
str |
Local path to copy from. |
required |
destination |
str |
Local path to copy to. |
required |
overwrite |
bool |
boolean, if false, then throws an error before overwrite. |
False |
Source code in zenml/utils/io_utils.py
def move(source: str, destination: str, overwrite: bool = False) -> None:
"""Moves dir or file from source to destination. Can be used to rename.
Args:
source: Local path to copy from.
destination: Local path to copy to.
overwrite: boolean, if false, then throws an error before overwrite.
"""
rename(source, destination, overwrite)
read_file_contents_as_string(file_path)
Reads contents of file.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to file. |
required |
Returns:
| Type | Description |
|---|---|
str |
Contents of file. |
Exceptions:
| Type | Description |
|---|---|
FileNotFoundError |
If file does not exist. |
Source code in zenml/utils/io_utils.py
def read_file_contents_as_string(file_path: str) -> str:
"""Reads contents of file.
Args:
file_path: Path to file.
Returns:
Contents of file.
Raises:
FileNotFoundError: If file does not exist.
"""
if not exists(file_path):
raise FileNotFoundError(f"{file_path} does not exist!")
with open(file_path) as f:
return f.read() # type: ignore[no-any-return]
resolve_relative_path(path)
Takes relative path and resolves it absolutely.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path |
str |
Local path in filesystem. |
required |
Returns:
| Type | Description |
|---|---|
str |
Resolved path. |
Source code in zenml/utils/io_utils.py
def resolve_relative_path(path: str) -> str:
"""Takes relative path and resolves it absolutely.
Args:
path: Local path in filesystem.
Returns:
Resolved path.
"""
if is_remote(path):
return path
return str(Path(path).resolve())
write_file_contents_as_string(file_path, content)
Writes contents of file.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to file. |
required |
content |
str |
Contents of file. |
required |
Exceptions:
| Type | Description |
|---|---|
ValueError |
If content is not of type str. |
Source code in zenml/utils/io_utils.py
def write_file_contents_as_string(file_path: str, content: str) -> None:
"""Writes contents of file.
Args:
file_path: Path to file.
content: Contents of file.
Raises:
ValueError: If content is not of type str.
"""
if not isinstance(content, str):
raise ValueError(f"Content must be of type str, got {type(content)}")
with open(file_path, "w") as f:
f.write(content)
materializer_utils
Util functions for materializers.
select_materializer(data_type, materializer_classes)
Select a materializer for a given data type.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
data_type |
Type[Any] |
The data type for which to select the materializer. |
required |
materializer_classes |
Sequence[Type[BaseMaterializer]] |
Available materializer classes. |
required |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If no materializer can handle the given data type. |
Returns:
| Type | Description |
|---|---|
Type[BaseMaterializer] |
The first materializer that can handle the given data type. |
Source code in zenml/utils/materializer_utils.py
def select_materializer(
data_type: Type[Any],
materializer_classes: Sequence[Type["BaseMaterializer"]],
) -> Type["BaseMaterializer"]:
"""Select a materializer for a given data type.
Args:
data_type: The data type for which to select the materializer.
materializer_classes: Available materializer classes.
Raises:
RuntimeError: If no materializer can handle the given data type.
Returns:
The first materializer that can handle the given data type.
"""
fallback: Optional[Type["BaseMaterializer"]] = None
for class_ in data_type.__mro__:
for materializer_class in materializer_classes:
if class_ in materializer_class.ASSOCIATED_TYPES:
return materializer_class
elif not fallback and materializer_class.can_handle_type(class_):
fallback = materializer_class
if fallback:
return fallback
raise RuntimeError(f"No materializer found for type {data_type}.")
mlstacks_utils
Functionality to handle interaction with the mlstacks package.
convert_click_params_to_mlstacks_primitives(params, zenml_component_deploy=False)
Converts raw Click CLI params to mlstacks primitives.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
params |
Dict[str, Any] |
Raw Click CLI params. |
required |
zenml_component_deploy |
bool |
A boolean indicating whether the stack
contains a single component and is being deployed through |
False |
Returns:
| Type | Description |
|---|---|
Tuple[Stack, List[Component]] |
A tuple of Stack and List[Component] objects. |
Source code in zenml/utils/mlstacks_utils.py
def convert_click_params_to_mlstacks_primitives(
params: Dict[str, Any],
zenml_component_deploy: bool = False,
) -> Tuple["Stack", List["Component"]]:
"""Converts raw Click CLI params to mlstacks primitives.
Args:
params: Raw Click CLI params.
zenml_component_deploy: A boolean indicating whether the stack
contains a single component and is being deployed through `zenml
component deploy`
Returns:
A tuple of Stack and List[Component] objects.
"""
from zenml.cli.utils import verify_mlstacks_prerequisites_installation
verify_mlstacks_prerequisites_installation()
from mlstacks.constants import MLSTACKS_PACKAGE_NAME
from mlstacks.models import Component, Stack
stack: Stack = _construct_base_stack(params)
components: List[Component] = _construct_components(
params, zenml_component_deploy
)
# writes the file names to the stack spec
# using format '<provider>-<component_name>.yaml'
for component in components:
stack.components.append(
os.path.join(
click.get_app_dir(MLSTACKS_PACKAGE_NAME),
"stack_specs",
stack.name,
f"{component.name}.yaml",
)
)
return stack, components
convert_mlstacks_primitives_to_dicts(stack, components)
Converts mlstacks Stack and Components to dicts.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack |
Stack |
A mlstacks Stack object. |
required |
components |
List[Component] |
A list of mlstacks Component objects. |
required |
Returns:
| Type | Description |
|---|---|
Tuple[Dict[str, Any], List[Dict[str, Any]]] |
A tuple of Stack and List[Component] dicts. |
Source code in zenml/utils/mlstacks_utils.py
def convert_mlstacks_primitives_to_dicts(
stack: "Stack", components: List["Component"]
) -> Tuple[Dict[str, Any], List[Dict[str, Any]]]:
"""Converts mlstacks Stack and Components to dicts.
Args:
stack: A mlstacks Stack object.
components: A list of mlstacks Component objects.
Returns:
A tuple of Stack and List[Component] dicts.
"""
from zenml.cli.utils import verify_mlstacks_prerequisites_installation
verify_mlstacks_prerequisites_installation()
# convert to json first to strip out Enums objects
stack_dict = json.loads(stack.json())
components_dicts = [
json.loads(component.json()) for component in components
]
return stack_dict, components_dicts
deploy_mlstacks_stack(spec_file_path, stack_name, stack_provider, debug_mode=False, no_import_stack_flag=False, user_created_spec=False)
Deploys an MLStacks stack given a spec file path.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
spec_file_path |
str |
The path to the spec file. |
required |
stack_name |
str |
The name of the stack. |
required |
stack_provider |
str |
The cloud provider for which the stack is deployed. |
required |
debug_mode |
bool |
A boolean indicating whether to run in debug mode. |
False |
no_import_stack_flag |
bool |
A boolean indicating whether to import the stack into ZenML. |
False |
user_created_spec |
bool |
A boolean indicating whether the user created the spec file. |
False |
Source code in zenml/utils/mlstacks_utils.py
def deploy_mlstacks_stack(
spec_file_path: str,
stack_name: str,
stack_provider: str,
debug_mode: bool = False,
no_import_stack_flag: bool = False,
user_created_spec: bool = False,
) -> None:
"""Deploys an MLStacks stack given a spec file path.
Args:
spec_file_path: The path to the spec file.
stack_name: The name of the stack.
stack_provider: The cloud provider for which the stack is deployed.
debug_mode: A boolean indicating whether to run in debug mode.
no_import_stack_flag: A boolean indicating whether to import the stack
into ZenML.
user_created_spec: A boolean indicating whether the user created the
spec file.
"""
from zenml.cli.utils import (
declare,
verify_mlstacks_prerequisites_installation,
)
verify_mlstacks_prerequisites_installation()
from mlstacks.constants import MLSTACKS_PACKAGE_NAME
from mlstacks.utils import terraform_utils
spec_dir = os.path.join(
click.get_app_dir(MLSTACKS_PACKAGE_NAME), "stack_specs", stack_name
)
declare("Deploying stack using Terraform...")
terraform_utils.deploy_stack(spec_file_path, debug_mode=debug_mode)
declare("Stack successfully deployed.")
if not no_import_stack_flag:
declare(f"Importing stack '{stack_name}' into ZenML...")
import_new_mlstacks_stack(
stack_name=stack_name,
provider=stack_provider,
stack_spec_dir=spec_dir,
user_stack_spec_file=spec_file_path if user_created_spec else None,
)
declare("Stack successfully imported into ZenML.")
get_stack_spec_file_path(stack_name)
Gets the path to the stack spec file for the given stack name.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack_name |
str |
The name of the stack spec to get the path for. |
required |
Returns:
| Type | Description |
|---|---|
str |
The path to the stack spec file for the given stack name. |
Exceptions:
| Type | Description |
|---|---|
KeyError |
If the stack does not exist. |
Source code in zenml/utils/mlstacks_utils.py
def get_stack_spec_file_path(stack_name: str) -> str:
"""Gets the path to the stack spec file for the given stack name.
Args:
stack_name: The name of the stack spec to get the path for.
Returns:
The path to the stack spec file for the given stack name.
Raises:
KeyError: If the stack does not exist.
"""
try:
stack = Client().get_stack(
name_id_or_prefix=stack_name, allow_name_prefix_match=False
)
except KeyError as e:
raise e
return stack.stack_spec_path or ""
import_new_mlstacks_component(stack_name, component_name, provider, stack_spec_dir)
Import a new component deployed for a particular cloud provider.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack_name |
str |
The name of the stack to import. |
required |
component_name |
str |
The name of the component to import. |
required |
provider |
str |
The cloud provider for which the stack is deployed. |
required |
stack_spec_dir |
str |
The path to the directory containing the stack spec. |
required |
Source code in zenml/utils/mlstacks_utils.py
def import_new_mlstacks_component(
stack_name: str, component_name: str, provider: str, stack_spec_dir: str
) -> None:
"""Import a new component deployed for a particular cloud provider.
Args:
stack_name: The name of the stack to import.
component_name: The name of the component to import.
provider: The cloud provider for which the stack is deployed.
stack_spec_dir: The path to the directory containing the stack spec.
"""
data, _ = _setup_import(provider, stack_name, stack_spec_dir)
component_ids = _import_components(
data, stack_spec_dir, component_name=component_name
)
component_type = list(component_ids.keys())[
0
] # Assuming only one component is imported
component = Client().get_stack_component(
component_type, component_ids[component_type]
)
from zenml.cli.utils import print_model_url
print_model_url(get_component_url(component))
import_new_mlstacks_stack(stack_name, provider, stack_spec_dir, user_stack_spec_file=None)
Import a new stack deployed for a particular cloud provider.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack_name |
str |
The name of the stack to import. |
required |
provider |
str |
The cloud provider for which the stack is deployed. |
required |
stack_spec_dir |
str |
The path to the directory containing the stack spec. |
required |
user_stack_spec_file |
Optional[str] |
The path to the user-created stack spec file. |
None |
Source code in zenml/utils/mlstacks_utils.py
def import_new_mlstacks_stack(
stack_name: str,
provider: str,
stack_spec_dir: str,
user_stack_spec_file: Optional[str] = None,
) -> None:
"""Import a new stack deployed for a particular cloud provider.
Args:
stack_name: The name of the stack to import.
provider: The cloud provider for which the stack is deployed.
stack_spec_dir: The path to the directory containing the stack spec.
user_stack_spec_file: The path to the user-created stack spec file.
"""
data, stack_spec_file = _setup_import(
provider, stack_name, stack_spec_dir, user_stack_spec_file
)
component_ids = _import_components(data, stack_spec_dir)
imported_stack = Client().create_stack(
name=stack_name,
components=component_ids,
is_shared=False,
stack_spec_file=stack_spec_file,
)
from zenml.cli.utils import print_model_url
print_model_url(get_stack_url(imported_stack))
stack_exists(stack_name)
Checks whether a stack with that name exists or not.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack_name |
str |
The name of the stack to check for. |
required |
Returns:
| Type | Description |
|---|---|
bool |
A boolean indicating whether the stack exists or not. |
Source code in zenml/utils/mlstacks_utils.py
def stack_exists(stack_name: str) -> bool:
"""Checks whether a stack with that name exists or not.
Args:
stack_name: The name of the stack to check for.
Returns:
A boolean indicating whether the stack exists or not.
"""
try:
Client().get_stack(
name_id_or_prefix=stack_name, allow_name_prefix_match=False
)
except KeyError:
return False
return True
stack_spec_exists(stack_name)
Checks whether a stack spec with that name exists or not.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack_name |
str |
The name of the stack spec to check for. |
required |
Returns:
| Type | Description |
|---|---|
bool |
A boolean indicating whether the stack spec exists or not. |
Source code in zenml/utils/mlstacks_utils.py
def stack_spec_exists(stack_name: str) -> bool:
"""Checks whether a stack spec with that name exists or not.
Args:
stack_name: The name of the stack spec to check for.
Returns:
A boolean indicating whether the stack spec exists or not.
"""
from zenml.cli.utils import verify_mlstacks_prerequisites_installation
verify_mlstacks_prerequisites_installation()
from mlstacks.constants import MLSTACKS_PACKAGE_NAME
spec_dir = os.path.join(
click.get_app_dir(MLSTACKS_PACKAGE_NAME), "stack_specs", stack_name
)
return Path(spec_dir).exists()
verify_spec_and_tf_files_exist(spec_file_path, tf_file_path)
Checks whether both the spec and tf files exist.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
spec_file_path |
str |
The path to the spec file. |
required |
tf_file_path |
str |
The path to the tf file. |
required |
Source code in zenml/utils/mlstacks_utils.py
def verify_spec_and_tf_files_exist(
spec_file_path: str, tf_file_path: str
) -> None:
"""Checks whether both the spec and tf files exist.
Args:
spec_file_path: The path to the spec file.
tf_file_path: The path to the tf file.
"""
from zenml.cli.utils import error
if not Path(spec_file_path).exists():
error(f"Could not find the Stack spec file at {spec_file_path}.")
elif not Path(tf_file_path).exists():
error(
f"Could not find the Terraform files for the stack "
f"at {tf_file_path}."
)
networking_utils
Utility functions for networking.
find_available_port()
Finds a local random unoccupied TCP port.
Returns:
| Type | Description |
|---|---|
int |
A random unoccupied TCP port. |
Source code in zenml/utils/networking_utils.py
def find_available_port() -> int:
"""Finds a local random unoccupied TCP port.
Returns:
A random unoccupied TCP port.
"""
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
s.bind(("127.0.0.1", 0))
_, port = s.getsockname()
return cast(int, port)
get_or_create_ngrok_tunnel(ngrok_token, port)
Get or create an ngrok tunnel at the given port.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
ngrok_token |
str |
The ngrok auth token. |
required |
port |
int |
The port to tunnel. |
required |
Returns:
| Type | Description |
|---|---|
str |
The public URL of the ngrok tunnel. |
Exceptions:
| Type | Description |
|---|---|
ImportError |
If the |
Source code in zenml/utils/networking_utils.py
def get_or_create_ngrok_tunnel(ngrok_token: str, port: int) -> str:
"""Get or create an ngrok tunnel at the given port.
Args:
ngrok_token: The ngrok auth token.
port: The port to tunnel.
Returns:
The public URL of the ngrok tunnel.
Raises:
ImportError: If the `pyngrok` package is not installed.
"""
try:
from pyngrok import ngrok as ngrok_client
except ImportError:
raise ImportError(
"The `pyngrok` package is required to create ngrok tunnels. "
"Please install it by running `pip install pyngrok`."
)
# Check if ngrok is already tunneling the port
tunnels = ngrok_client.get_tunnels()
for tunnel in tunnels:
if tunnel.config and isinstance(tunnel.config, dict):
tunnel_protocol = tunnel.config.get("proto")
tunnel_port = tunnel.config.get("addr")
if tunnel_protocol == "http" and tunnel_port == port:
return str(tunnel.public_url)
# Create new tunnel
ngrok_client.set_auth_token(ngrok_token)
return str(ngrok_client.connect(port).public_url)
port_available(port, address='127.0.0.1')
Checks if a local port is available.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
port |
int |
TCP port number |
required |
address |
str |
IP address on the local machine |
'127.0.0.1' |
Returns:
| Type | Description |
|---|---|
bool |
True if the port is available, otherwise False |
Source code in zenml/utils/networking_utils.py
def port_available(port: int, address: str = "127.0.0.1") -> bool:
"""Checks if a local port is available.
Args:
port: TCP port number
address: IP address on the local machine
Returns:
True if the port is available, otherwise False
"""
try:
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
if hasattr(socket, "SO_REUSEPORT"):
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEPORT, 1)
else:
# The SO_REUSEPORT socket option is not supported on Windows.
# This if clause exists just for mypy to not complain about
# missing code paths.
pass
s.bind((address, port))
except socket.error as e:
logger.debug("Port %d unavailable on %s: %s", port, address, e)
return False
return True
port_is_open(hostname, port)
Check if a TCP port is open on a remote host.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
hostname |
str |
hostname of the remote machine |
required |
port |
int |
TCP port number |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if the port is open, False otherwise |
Source code in zenml/utils/networking_utils.py
def port_is_open(hostname: str, port: int) -> bool:
"""Check if a TCP port is open on a remote host.
Args:
hostname: hostname of the remote machine
port: TCP port number
Returns:
True if the port is open, False otherwise
"""
try:
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as sock:
result = sock.connect_ex((hostname, port))
return result == 0
except socket.error as e:
logger.debug(
f"Error checking TCP port {port} on host {hostname}: {str(e)}"
)
return False
replace_internal_hostname_with_localhost(hostname)
Replaces an internal Docker or K3D hostname with localhost.
Localhost URLs that are directly accessible on the host machine are not
accessible from within a Docker or K3D container running on that same
machine, but there are special hostnames featured by both Docker
(host.docker.internal) and K3D (host.k3d.internal) that can be used to
access host services from within the containers.
Use this method to replace one of these special hostnames with localhost if used outside a container or in a container where special hostnames are not available.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
hostname |
str |
The hostname to replace. |
required |
Returns:
| Type | Description |
|---|---|
str |
The original or replaced hostname. |
Source code in zenml/utils/networking_utils.py
def replace_internal_hostname_with_localhost(hostname: str) -> str:
"""Replaces an internal Docker or K3D hostname with localhost.
Localhost URLs that are directly accessible on the host machine are not
accessible from within a Docker or K3D container running on that same
machine, but there are special hostnames featured by both Docker
(`host.docker.internal`) and K3D (`host.k3d.internal`) that can be used to
access host services from within the containers.
Use this method to replace one of these special hostnames with localhost
if used outside a container or in a container where special hostnames are
not available.
Args:
hostname: The hostname to replace.
Returns:
The original or replaced hostname.
"""
if hostname not in ("host.docker.internal", "host.k3d.internal"):
return hostname
if Environment.in_container():
# Try to resolve one of the special hostnames to see if it is available
# inside the container and use that if it is.
for internal_hostname in (
"host.docker.internal",
"host.k3d.internal",
):
try:
socket.gethostbyname(internal_hostname)
if internal_hostname != hostname:
logger.debug(
f"Replacing internal hostname {hostname} with "
f"{internal_hostname}"
)
return internal_hostname
except socket.gaierror:
continue
logger.debug(f"Replacing internal hostname {hostname} with localhost.")
return "127.0.0.1"
replace_localhost_with_internal_hostname(url)
Replaces the localhost with an internal Docker or K3D hostname in a given URL.
Localhost URLs that are directly accessible on the host machine are not
accessible from within a Docker or K3D container running on that same
machine, but there are special hostnames featured by both Docker
(host.docker.internal) and K3D (host.k3d.internal) that can be used to
access host services from within the containers.
Use this method to attempt to replace localhost in a URL with one of these
special hostnames, if they are available inside a container.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
url |
str |
The URL to update. |
required |
Returns:
| Type | Description |
|---|---|
str |
The updated URL. |
Source code in zenml/utils/networking_utils.py
def replace_localhost_with_internal_hostname(url: str) -> str:
"""Replaces the localhost with an internal Docker or K3D hostname in a given URL.
Localhost URLs that are directly accessible on the host machine are not
accessible from within a Docker or K3D container running on that same
machine, but there are special hostnames featured by both Docker
(`host.docker.internal`) and K3D (`host.k3d.internal`) that can be used to
access host services from within the containers.
Use this method to attempt to replace `localhost` in a URL with one of these
special hostnames, if they are available inside a container.
Args:
url: The URL to update.
Returns:
The updated URL.
"""
if not Environment.in_container():
return url
parsed_url = urlparse(url)
if parsed_url.hostname in ("localhost", "127.0.0.1"):
for internal_hostname in (
"host.docker.internal",
"host.k3d.internal",
):
try:
socket.gethostbyname(internal_hostname)
parsed_url = parsed_url._replace(
netloc=parsed_url.netloc.replace(
parsed_url.hostname,
internal_hostname,
)
)
logger.debug(
f"Replacing localhost with {internal_hostname} in URL: "
f"{url}"
)
return parsed_url.geturl()
except socket.gaierror:
continue
return url
scan_for_available_port(start=8000, stop=65535)
Scan the local network for an available port in the given range.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
start |
int |
the beginning of the port range value to scan |
8000 |
stop |
int |
the (inclusive) end of the port range value to scan |
65535 |
Returns:
| Type | Description |
|---|---|
Optional[int] |
The first available port in the given range, or None if no available port is found. |
Source code in zenml/utils/networking_utils.py
def scan_for_available_port(
start: int = SCAN_PORT_RANGE[0], stop: int = SCAN_PORT_RANGE[1]
) -> Optional[int]:
"""Scan the local network for an available port in the given range.
Args:
start: the beginning of the port range value to scan
stop: the (inclusive) end of the port range value to scan
Returns:
The first available port in the given range, or None if no available
port is found.
"""
for port in range(start, stop + 1):
if port_available(port):
return port
logger.debug(
"No free TCP ports found in the range %d - %d",
start,
stop,
)
return None
pagination_utils
Pagination utilities.
Page[AnyResponseModel] (Page)
pydantic-model
Config
Pydantic configuration class.
Source code in zenml/utils/pagination_utils.py
class Config:
"""Pydantic configuration class."""
# This is needed to allow the REST API server to unpack SecretStr
# values correctly before sending them to the client.
json_encoders = {
SecretStr: lambda v: v.get_secret_value() if v else None
}
__json_encoder__(obj)
special
staticmethod
partial(func, args, *keywords) - new function with partial application of the given arguments and keywords.
depaginate(list_method)
Depaginate the results from a client or store method that returns pages.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
list_method |
Callable[..., zenml.utils.pagination_utils.Page[AnyResponseModel]] |
The list method to wrap around. |
required |
Returns:
| Type | Description |
|---|---|
List[~AnyResponseModel] |
A list of the corresponding Response Models. |
Source code in zenml/utils/pagination_utils.py
def depaginate(
list_method: Callable[..., Page[AnyResponseModel]],
) -> List[AnyResponseModel]:
"""Depaginate the results from a client or store method that returns pages.
Args:
list_method: The list method to wrap around.
Returns:
A list of the corresponding Response Models.
"""
page = list_method()
items = list(page.items)
while page.index < page.total_pages:
page = list_method(page=page.index + 1)
items += list(page.items)
return items
pipeline_docker_image_builder
Implementation of Docker image builds to run ZenML pipelines.
PipelineDockerImageBuilder
Builds Docker images to run a ZenML pipeline.
Source code in zenml/utils/pipeline_docker_image_builder.py
class PipelineDockerImageBuilder:
"""Builds Docker images to run a ZenML pipeline."""
def build_docker_image(
self,
docker_settings: "DockerSettings",
tag: str,
stack: "Stack",
include_files: bool,
download_files: bool,
entrypoint: Optional[str] = None,
extra_files: Optional[Dict[str, str]] = None,
code_repository: Optional["BaseCodeRepository"] = None,
) -> Tuple[str, Optional[str], Optional[str]]:
"""Builds (and optionally pushes) a Docker image to run a pipeline.
Use the image name returned by this method whenever you need to uniquely
reference the pushed image in order to pull or run it.
Args:
docker_settings: The settings for the image build.
tag: The tag to use for the image.
stack: The stack on which the pipeline will be deployed.
include_files: Whether to include files in the build context.
download_files: Whether to download files in the build context.
entrypoint: Entrypoint to use for the final image. If left empty,
no entrypoint will be included in the image.
extra_files: Extra files to add to the build context. Keys are the
path inside the build context, values are either the file
content or a file path.
code_repository: The code repository from which files will be
downloaded.
Returns:
A tuple (image_digest, dockerfile, requirements):
- The Docker image repo digest or local name, depending on whether
the image was pushed or is just stored locally.
- Dockerfile will contain the contents of the Dockerfile used to
build the image.
- Requirements is a string with a single pip requirement per line.
Raises:
RuntimeError: If the stack does not contain an image builder.
ValueError: If no Dockerfile and/or custom parent image is
specified and the Docker configuration doesn't require an
image build.
"""
requirements: Optional[str] = None
dockerfile: Optional[str] = None
if docker_settings.skip_build:
assert (
docker_settings.parent_image
) # checked via validator already
# Should we tag this here and push it to the container registry of
# the stack to make sure it's always accessible when running the
# pipeline?
return docker_settings.parent_image, dockerfile, requirements
stack.validate_image_builder()
image_builder = stack.image_builder
if not image_builder:
raise RuntimeError(
"Unable to build Docker images without an image builder in the "
f"stack `{stack.name}`."
)
container_registry = stack.container_registry
build_context_class = image_builder.build_context_class
target_image_name = self._get_target_image_name(
docker_settings=docker_settings,
tag=tag,
container_registry=container_registry,
)
requires_zenml_build = any(
[
docker_settings.requirements,
docker_settings.required_integrations,
docker_settings.required_hub_plugins,
docker_settings.replicate_local_python_environment,
docker_settings.install_stack_requirements,
docker_settings.apt_packages,
docker_settings.environment,
include_files,
download_files,
entrypoint,
extra_files,
]
)
# Fallback to the value defined on the stack component if the
# pipeline configuration doesn't have a configured value
parent_image = (
docker_settings.parent_image or DEFAULT_DOCKER_PARENT_IMAGE
)
if docker_settings.dockerfile:
if parent_image != DEFAULT_DOCKER_PARENT_IMAGE:
logger.warning(
"You've specified both a Dockerfile and a custom parent "
"image, ignoring the parent image."
)
push = (
not image_builder.is_building_locally
or not requires_zenml_build
)
if requires_zenml_build:
# We will build an additional image on top of this one later
# to include user files and/or install requirements. The image
# we build now will be used as the parent for the next build.
user_image_name = (
f"{docker_settings.target_repository}:"
f"{tag}-intermediate-build"
)
if push and container_registry:
user_image_name = (
f"{container_registry.config.uri}/{user_image_name}"
)
parent_image = user_image_name
else:
# The image we'll build from the custom Dockerfile will be
# used directly, so we tag it with the requested target name.
user_image_name = target_image_name
build_context = build_context_class(
root=docker_settings.build_context_root
)
build_context.add_file(
source=docker_settings.dockerfile, destination="Dockerfile"
)
logger.info("Building Docker image `%s`.", user_image_name)
image_name_or_digest = image_builder.build(
image_name=user_image_name,
build_context=build_context,
docker_build_options=docker_settings.build_options,
container_registry=container_registry if push else None,
)
elif not requires_zenml_build:
if parent_image == DEFAULT_DOCKER_PARENT_IMAGE:
raise ValueError(
"Unable to run a ZenML pipeline with the given Docker "
"settings: No Dockerfile or custom parent image "
"specified and no files will be copied or requirements "
"installed."
)
else:
# The parent image will be used directly to run the pipeline and
# needs to be tagged/pushed
docker_utils.tag_image(parent_image, target=target_image_name)
if container_registry:
image_name_or_digest = container_registry.push_image(
target_image_name
)
else:
image_name_or_digest = target_image_name
if requires_zenml_build:
logger.info("Building Docker image `%s`.", target_image_name)
# Leave the build context empty if we don't want to include any files
build_context_root = (
source_utils.get_source_root() if include_files else None
)
build_context = build_context_class(
root=build_context_root,
dockerignore_file=docker_settings.dockerignore,
)
requirements_files = self.gather_requirements_files(
docker_settings=docker_settings,
stack=stack,
# Only pass code repo to include its dependencies if we actually
# need to download code
code_repository=code_repository if download_files else None,
)
self._add_requirements_files(
requirements_files=requirements_files,
build_context=build_context,
)
requirements = (
"\n".join(
file_content for _, file_content, _ in requirements_files
)
or None
)
apt_packages = docker_settings.apt_packages.copy()
if docker_settings.install_stack_requirements:
apt_packages += stack.apt_packages
if apt_packages:
logger.info(
"Including apt packages: %s",
", ".join(f"`{p}`" for p in apt_packages),
)
if parent_image == DEFAULT_DOCKER_PARENT_IMAGE:
# The default parent image is static and doesn't require a pull
# each time
pull_parent_image = False
elif docker_settings.dockerfile and not container_registry:
# We built a custom parent image and there was no container
# registry in the stack to push to, this is a local image
pull_parent_image = False
elif not image_builder.is_building_locally:
# Remote image builders always need to pull the image
pull_parent_image = True
else:
# If the image is local, we don't need to pull it. Otherwise
# we play it safe and always pull in case the user pushed a new
# image for the given name and tag
pull_parent_image = not docker_utils.is_local_image(
parent_image
)
build_options = {"pull": pull_parent_image, "rm": False}
dockerfile = self._generate_zenml_pipeline_dockerfile(
parent_image=parent_image,
docker_settings=docker_settings,
download_files=download_files,
requirements_files=requirements_files,
apt_packages=apt_packages,
entrypoint=entrypoint,
)
build_context.add_file(destination="Dockerfile", source=dockerfile)
if extra_files:
for destination, source in extra_files.items():
build_context.add_file(
destination=destination, source=source
)
image_name_or_digest = image_builder.build(
image_name=target_image_name,
build_context=build_context,
docker_build_options=build_options,
container_registry=container_registry,
)
return image_name_or_digest, dockerfile, requirements
@staticmethod
def _get_target_image_name(
docker_settings: "DockerSettings",
tag: str,
container_registry: Optional["BaseContainerRegistry"] = None,
) -> str:
"""Returns the target image name.
If a container registry is given, the image name will include the
registry URI
Args:
docker_settings: The settings for the image build.
tag: The tag to use for the image.
container_registry: Optional container registry to which this
image will be pushed.
Returns:
The docker image name.
"""
target_image_name = f"{docker_settings.target_repository}:{tag}"
if container_registry:
target_image_name = (
f"{container_registry.config.uri}/{target_image_name}"
)
return target_image_name
@classmethod
def _add_requirements_files(
cls,
requirements_files: List[Tuple[str, str, List[str]]],
build_context: "BuildContext",
) -> None:
"""Adds requirements files to the build context.
Args:
requirements_files: List of tuples
(filename, file_content, pip_options).
build_context: Build context to add the requirements files to.
"""
for filename, file_content, _ in requirements_files:
build_context.add_file(source=file_content, destination=filename)
@staticmethod
def gather_requirements_files(
docker_settings: DockerSettings,
stack: "Stack",
code_repository: Optional["BaseCodeRepository"] = None,
log: bool = True,
) -> List[Tuple[str, str, List[str]]]:
"""Gathers and/or generates pip requirements files.
This method is called in `PipelineDockerImageBuilder.build_docker_image`
but it is also called by other parts of the codebase, e.g. the
`AzureMLStepOperator`, which needs to upload the requirements files to
AzureML where the step image is then built.
Args:
docker_settings: Docker settings that specifies which
requirements to install.
stack: The stack on which the pipeline will run.
code_repository: The code repository from which files will be
downloaded.
log: If True, will log the requirements.
Raises:
RuntimeError: If the command to export the local python packages
failed.
FileNotFoundError: If the specified requirements file does not
exist.
Returns:
List of tuples (filename, file_content, pip_options) of all
requirements files.
The files will be in the following order:
- Packages installed in the local Python environment
- User-defined requirements
- Requirements defined by user-defined and/or stack integrations
"""
requirements_files: List[Tuple[str, str, List[str]]] = []
# Generate requirements file for the local environment if configured
if docker_settings.replicate_local_python_environment:
if isinstance(
docker_settings.replicate_local_python_environment,
PythonEnvironmentExportMethod,
):
command = (
docker_settings.replicate_local_python_environment.command
)
else:
command = " ".join(
docker_settings.replicate_local_python_environment
)
try:
local_requirements = subprocess.check_output(
command, shell=True # nosec
).decode()
except subprocess.CalledProcessError as e:
raise RuntimeError(
"Unable to export local python packages."
) from e
requirements_files.append(
(".zenml_local_requirements", local_requirements, [])
)
if log:
logger.info(
"- Including python packages from local environment"
)
# Generate/Read requirements file for user-defined requirements
if isinstance(docker_settings.requirements, str):
path = os.path.abspath(docker_settings.requirements)
try:
user_requirements = io_utils.read_file_contents_as_string(path)
except FileNotFoundError as e:
raise FileNotFoundError(
f"Requirements file {path} does not exist."
) from e
if log:
logger.info(
"- Including user-defined requirements from file `%s`",
path,
)
elif isinstance(docker_settings.requirements, List):
user_requirements = "\n".join(docker_settings.requirements)
if log:
logger.info(
"- Including user-defined requirements: %s",
", ".join(f"`{r}`" for r in docker_settings.requirements),
)
else:
user_requirements = None
if user_requirements:
requirements_files.append(
(".zenml_user_requirements", user_requirements, [])
)
# Generate requirements file for all required integrations
integration_requirements = set(
itertools.chain.from_iterable(
integration_registry.select_integration_requirements(
integration_name=integration,
target_os=OperatingSystemType.LINUX,
)
for integration in docker_settings.required_integrations
)
)
if docker_settings.install_stack_requirements:
integration_requirements.update(stack.requirements())
if code_repository:
integration_requirements.update(code_repository.requirements)
if integration_requirements:
integration_requirements_list = sorted(integration_requirements)
integration_requirements_file = "\n".join(
integration_requirements_list
)
requirements_files.append(
(
".zenml_integration_requirements",
integration_requirements_file,
[],
)
)
if log:
logger.info(
"- Including integration requirements: %s",
", ".join(f"`{r}`" for r in integration_requirements_list),
)
# Generate requirements files for all ZenML Hub plugins
if docker_settings.required_hub_plugins:
(
hub_internal_requirements,
hub_pypi_requirements,
) = PipelineDockerImageBuilder._get_hub_requirements(
docker_settings.required_hub_plugins
)
# Plugin packages themselves
for i, (index, packages) in enumerate(
hub_internal_requirements.items()
):
file_name = f".zenml_hub_internal_requirements_{i}"
file_lines = [f"-i {index}", *packages]
file_contents = "\n".join(file_lines)
requirements_files.append(
(file_name, file_contents, ["--no-deps"])
)
if log:
logger.info(
"- Including internal hub packages from index `%s`: %s",
index,
", ".join(f"`{r}`" for r in packages),
)
# PyPI requirements of plugin packages
if hub_pypi_requirements:
file_name = ".zenml_hub_pypi_requirements"
file_contents = "\n".join(hub_pypi_requirements)
requirements_files.append((file_name, file_contents, []))
if log:
logger.info(
"- Including hub requirements from PyPI: %s",
", ".join(f"`{r}`" for r in hub_pypi_requirements),
)
return requirements_files
@staticmethod
def _get_hub_requirements(
required_hub_plugins: List[str],
) -> Tuple[Dict[str, List[str]], List[str]]:
"""Get package requirements for ZenML Hub plugins.
Args:
required_hub_plugins: List of hub plugin names in the format
`(<author_username>/)<plugin_name>(==<version>)`.
Returns:
- A dict of the hub plugin packages themselves (which need to be
installed from a custom index, mapping index URLs to lists of
package names.
- A list of all unique dependencies of the required hub plugins
(which can be installed from PyPI).
"""
from zenml._hub.client import HubClient
from zenml._hub.utils import parse_plugin_name, plugin_display_name
client = HubClient()
internal_requirements: DefaultDict[str, List[str]] = defaultdict(list)
pypi_requirements: List[str] = []
for plugin_str in required_hub_plugins:
author, name, version = parse_plugin_name(
plugin_str, version_separator="=="
)
plugin = client.get_plugin(
name=name,
version=version,
author=author,
)
if plugin and plugin.index_url and plugin.package_name:
internal_requirements[plugin.index_url].append(
plugin.package_name
)
if plugin.requirements:
pypi_requirements.extend(plugin.requirements)
else:
display_name = plugin_display_name(name, version, author)
logger.warning(
"Hub plugin `%s` does not exist or cannot be installed."
"Skipping installation of this plugin.",
display_name,
)
pypi_requirements = sorted(set(pypi_requirements))
return dict(internal_requirements), pypi_requirements
@staticmethod
def _generate_zenml_pipeline_dockerfile(
parent_image: str,
docker_settings: DockerSettings,
download_files: bool,
requirements_files: Sequence[Tuple[str, str, List[str]]] = (),
apt_packages: Sequence[str] = (),
entrypoint: Optional[str] = None,
) -> str:
"""Generates a Dockerfile.
Args:
parent_image: The image to use as parent for the Dockerfile.
docker_settings: Docker settings for this image build.
download_files: Whether to download files in the build context.
requirements_files: List of tuples that contain three items:
- the name of a requirements file,
- the content of that file,
- options that should be passed to pip when installing the
requirements file.
apt_packages: APT packages to install.
entrypoint: The default entrypoint command that gets executed when
running a container of an image created by this Dockerfile.
Returns:
The generated Dockerfile.
"""
lines = [f"FROM {parent_image}", f"WORKDIR {DOCKER_IMAGE_WORKDIR}"]
for key, value in docker_settings.environment.items():
lines.append(f"ENV {key.upper()}={value}")
if apt_packages:
apt_packages = " ".join(f"'{p}'" for p in apt_packages)
lines.append(
"RUN apt-get update && apt-get install -y "
f"--no-install-recommends {apt_packages}"
)
for file, _, options in requirements_files:
lines.append(f"COPY {file} .")
option_string = " ".join(options)
lines.append(
f"RUN pip install --default-timeout=60 --no-cache-dir "
f"{option_string} -r {file}"
)
lines.append(f"ENV {ENV_ZENML_ENABLE_REPO_INIT_WARNINGS}=False")
if download_files:
lines.append(f"ENV {ENV_ZENML_REQUIRES_CODE_DOWNLOAD}=True")
lines.append(
f"ENV {ENV_ZENML_CONFIG_PATH}={DOCKER_IMAGE_ZENML_CONFIG_PATH}"
)
lines.append("COPY . .")
lines.append("RUN chmod -R a+rw .")
if docker_settings.user:
# Change file ownership to specified user
lines.append(f"RUN chown -R {docker_settings.user} .")
# Switch back to specified user for subsequent instructions
lines.append(f"USER {docker_settings.user}")
if entrypoint:
lines.append(f"ENTRYPOINT {entrypoint}")
return "\n".join(lines)
build_docker_image(self, docker_settings, tag, stack, include_files, download_files, entrypoint=None, extra_files=None, code_repository=None)
Builds (and optionally pushes) a Docker image to run a pipeline.
Use the image name returned by this method whenever you need to uniquely reference the pushed image in order to pull or run it.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
docker_settings |
DockerSettings |
The settings for the image build. |
required |
tag |
str |
The tag to use for the image. |
required |
stack |
Stack |
The stack on which the pipeline will be deployed. |
required |
include_files |
bool |
Whether to include files in the build context. |
required |
download_files |
bool |
Whether to download files in the build context. |
required |
entrypoint |
Optional[str] |
Entrypoint to use for the final image. If left empty, no entrypoint will be included in the image. |
None |
extra_files |
Optional[Dict[str, str]] |
Extra files to add to the build context. Keys are the path inside the build context, values are either the file content or a file path. |
None |
code_repository |
Optional[BaseCodeRepository] |
The code repository from which files will be downloaded. |
None |
Returns:
| Type | Description |
|---|---|
A tuple (image_digest, dockerfile, requirements) |
|
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If the stack does not contain an image builder. |
ValueError |
If no Dockerfile and/or custom parent image is specified and the Docker configuration doesn't require an image build. |
Source code in zenml/utils/pipeline_docker_image_builder.py
def build_docker_image(
self,
docker_settings: "DockerSettings",
tag: str,
stack: "Stack",
include_files: bool,
download_files: bool,
entrypoint: Optional[str] = None,
extra_files: Optional[Dict[str, str]] = None,
code_repository: Optional["BaseCodeRepository"] = None,
) -> Tuple[str, Optional[str], Optional[str]]:
"""Builds (and optionally pushes) a Docker image to run a pipeline.
Use the image name returned by this method whenever you need to uniquely
reference the pushed image in order to pull or run it.
Args:
docker_settings: The settings for the image build.
tag: The tag to use for the image.
stack: The stack on which the pipeline will be deployed.
include_files: Whether to include files in the build context.
download_files: Whether to download files in the build context.
entrypoint: Entrypoint to use for the final image. If left empty,
no entrypoint will be included in the image.
extra_files: Extra files to add to the build context. Keys are the
path inside the build context, values are either the file
content or a file path.
code_repository: The code repository from which files will be
downloaded.
Returns:
A tuple (image_digest, dockerfile, requirements):
- The Docker image repo digest or local name, depending on whether
the image was pushed or is just stored locally.
- Dockerfile will contain the contents of the Dockerfile used to
build the image.
- Requirements is a string with a single pip requirement per line.
Raises:
RuntimeError: If the stack does not contain an image builder.
ValueError: If no Dockerfile and/or custom parent image is
specified and the Docker configuration doesn't require an
image build.
"""
requirements: Optional[str] = None
dockerfile: Optional[str] = None
if docker_settings.skip_build:
assert (
docker_settings.parent_image
) # checked via validator already
# Should we tag this here and push it to the container registry of
# the stack to make sure it's always accessible when running the
# pipeline?
return docker_settings.parent_image, dockerfile, requirements
stack.validate_image_builder()
image_builder = stack.image_builder
if not image_builder:
raise RuntimeError(
"Unable to build Docker images without an image builder in the "
f"stack `{stack.name}`."
)
container_registry = stack.container_registry
build_context_class = image_builder.build_context_class
target_image_name = self._get_target_image_name(
docker_settings=docker_settings,
tag=tag,
container_registry=container_registry,
)
requires_zenml_build = any(
[
docker_settings.requirements,
docker_settings.required_integrations,
docker_settings.required_hub_plugins,
docker_settings.replicate_local_python_environment,
docker_settings.install_stack_requirements,
docker_settings.apt_packages,
docker_settings.environment,
include_files,
download_files,
entrypoint,
extra_files,
]
)
# Fallback to the value defined on the stack component if the
# pipeline configuration doesn't have a configured value
parent_image = (
docker_settings.parent_image or DEFAULT_DOCKER_PARENT_IMAGE
)
if docker_settings.dockerfile:
if parent_image != DEFAULT_DOCKER_PARENT_IMAGE:
logger.warning(
"You've specified both a Dockerfile and a custom parent "
"image, ignoring the parent image."
)
push = (
not image_builder.is_building_locally
or not requires_zenml_build
)
if requires_zenml_build:
# We will build an additional image on top of this one later
# to include user files and/or install requirements. The image
# we build now will be used as the parent for the next build.
user_image_name = (
f"{docker_settings.target_repository}:"
f"{tag}-intermediate-build"
)
if push and container_registry:
user_image_name = (
f"{container_registry.config.uri}/{user_image_name}"
)
parent_image = user_image_name
else:
# The image we'll build from the custom Dockerfile will be
# used directly, so we tag it with the requested target name.
user_image_name = target_image_name
build_context = build_context_class(
root=docker_settings.build_context_root
)
build_context.add_file(
source=docker_settings.dockerfile, destination="Dockerfile"
)
logger.info("Building Docker image `%s`.", user_image_name)
image_name_or_digest = image_builder.build(
image_name=user_image_name,
build_context=build_context,
docker_build_options=docker_settings.build_options,
container_registry=container_registry if push else None,
)
elif not requires_zenml_build:
if parent_image == DEFAULT_DOCKER_PARENT_IMAGE:
raise ValueError(
"Unable to run a ZenML pipeline with the given Docker "
"settings: No Dockerfile or custom parent image "
"specified and no files will be copied or requirements "
"installed."
)
else:
# The parent image will be used directly to run the pipeline and
# needs to be tagged/pushed
docker_utils.tag_image(parent_image, target=target_image_name)
if container_registry:
image_name_or_digest = container_registry.push_image(
target_image_name
)
else:
image_name_or_digest = target_image_name
if requires_zenml_build:
logger.info("Building Docker image `%s`.", target_image_name)
# Leave the build context empty if we don't want to include any files
build_context_root = (
source_utils.get_source_root() if include_files else None
)
build_context = build_context_class(
root=build_context_root,
dockerignore_file=docker_settings.dockerignore,
)
requirements_files = self.gather_requirements_files(
docker_settings=docker_settings,
stack=stack,
# Only pass code repo to include its dependencies if we actually
# need to download code
code_repository=code_repository if download_files else None,
)
self._add_requirements_files(
requirements_files=requirements_files,
build_context=build_context,
)
requirements = (
"\n".join(
file_content for _, file_content, _ in requirements_files
)
or None
)
apt_packages = docker_settings.apt_packages.copy()
if docker_settings.install_stack_requirements:
apt_packages += stack.apt_packages
if apt_packages:
logger.info(
"Including apt packages: %s",
", ".join(f"`{p}`" for p in apt_packages),
)
if parent_image == DEFAULT_DOCKER_PARENT_IMAGE:
# The default parent image is static and doesn't require a pull
# each time
pull_parent_image = False
elif docker_settings.dockerfile and not container_registry:
# We built a custom parent image and there was no container
# registry in the stack to push to, this is a local image
pull_parent_image = False
elif not image_builder.is_building_locally:
# Remote image builders always need to pull the image
pull_parent_image = True
else:
# If the image is local, we don't need to pull it. Otherwise
# we play it safe and always pull in case the user pushed a new
# image for the given name and tag
pull_parent_image = not docker_utils.is_local_image(
parent_image
)
build_options = {"pull": pull_parent_image, "rm": False}
dockerfile = self._generate_zenml_pipeline_dockerfile(
parent_image=parent_image,
docker_settings=docker_settings,
download_files=download_files,
requirements_files=requirements_files,
apt_packages=apt_packages,
entrypoint=entrypoint,
)
build_context.add_file(destination="Dockerfile", source=dockerfile)
if extra_files:
for destination, source in extra_files.items():
build_context.add_file(
destination=destination, source=source
)
image_name_or_digest = image_builder.build(
image_name=target_image_name,
build_context=build_context,
docker_build_options=build_options,
container_registry=container_registry,
)
return image_name_or_digest, dockerfile, requirements
gather_requirements_files(docker_settings, stack, code_repository=None, log=True)
staticmethod
Gathers and/or generates pip requirements files.
This method is called in PipelineDockerImageBuilder.build_docker_image
but it is also called by other parts of the codebase, e.g. the
AzureMLStepOperator, which needs to upload the requirements files to
AzureML where the step image is then built.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
docker_settings |
DockerSettings |
Docker settings that specifies which requirements to install. |
required |
stack |
Stack |
The stack on which the pipeline will run. |
required |
code_repository |
Optional[BaseCodeRepository] |
The code repository from which files will be downloaded. |
None |
log |
bool |
If True, will log the requirements. |
True |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If the command to export the local python packages failed. |
FileNotFoundError |
If the specified requirements file does not exist. |
Returns:
| Type | Description |
|---|---|
List of tuples (filename, file_content, pip_options) of all
requirements files.
The files will be in the following order |
|
Source code in zenml/utils/pipeline_docker_image_builder.py
@staticmethod
def gather_requirements_files(
docker_settings: DockerSettings,
stack: "Stack",
code_repository: Optional["BaseCodeRepository"] = None,
log: bool = True,
) -> List[Tuple[str, str, List[str]]]:
"""Gathers and/or generates pip requirements files.
This method is called in `PipelineDockerImageBuilder.build_docker_image`
but it is also called by other parts of the codebase, e.g. the
`AzureMLStepOperator`, which needs to upload the requirements files to
AzureML where the step image is then built.
Args:
docker_settings: Docker settings that specifies which
requirements to install.
stack: The stack on which the pipeline will run.
code_repository: The code repository from which files will be
downloaded.
log: If True, will log the requirements.
Raises:
RuntimeError: If the command to export the local python packages
failed.
FileNotFoundError: If the specified requirements file does not
exist.
Returns:
List of tuples (filename, file_content, pip_options) of all
requirements files.
The files will be in the following order:
- Packages installed in the local Python environment
- User-defined requirements
- Requirements defined by user-defined and/or stack integrations
"""
requirements_files: List[Tuple[str, str, List[str]]] = []
# Generate requirements file for the local environment if configured
if docker_settings.replicate_local_python_environment:
if isinstance(
docker_settings.replicate_local_python_environment,
PythonEnvironmentExportMethod,
):
command = (
docker_settings.replicate_local_python_environment.command
)
else:
command = " ".join(
docker_settings.replicate_local_python_environment
)
try:
local_requirements = subprocess.check_output(
command, shell=True # nosec
).decode()
except subprocess.CalledProcessError as e:
raise RuntimeError(
"Unable to export local python packages."
) from e
requirements_files.append(
(".zenml_local_requirements", local_requirements, [])
)
if log:
logger.info(
"- Including python packages from local environment"
)
# Generate/Read requirements file for user-defined requirements
if isinstance(docker_settings.requirements, str):
path = os.path.abspath(docker_settings.requirements)
try:
user_requirements = io_utils.read_file_contents_as_string(path)
except FileNotFoundError as e:
raise FileNotFoundError(
f"Requirements file {path} does not exist."
) from e
if log:
logger.info(
"- Including user-defined requirements from file `%s`",
path,
)
elif isinstance(docker_settings.requirements, List):
user_requirements = "\n".join(docker_settings.requirements)
if log:
logger.info(
"- Including user-defined requirements: %s",
", ".join(f"`{r}`" for r in docker_settings.requirements),
)
else:
user_requirements = None
if user_requirements:
requirements_files.append(
(".zenml_user_requirements", user_requirements, [])
)
# Generate requirements file for all required integrations
integration_requirements = set(
itertools.chain.from_iterable(
integration_registry.select_integration_requirements(
integration_name=integration,
target_os=OperatingSystemType.LINUX,
)
for integration in docker_settings.required_integrations
)
)
if docker_settings.install_stack_requirements:
integration_requirements.update(stack.requirements())
if code_repository:
integration_requirements.update(code_repository.requirements)
if integration_requirements:
integration_requirements_list = sorted(integration_requirements)
integration_requirements_file = "\n".join(
integration_requirements_list
)
requirements_files.append(
(
".zenml_integration_requirements",
integration_requirements_file,
[],
)
)
if log:
logger.info(
"- Including integration requirements: %s",
", ".join(f"`{r}`" for r in integration_requirements_list),
)
# Generate requirements files for all ZenML Hub plugins
if docker_settings.required_hub_plugins:
(
hub_internal_requirements,
hub_pypi_requirements,
) = PipelineDockerImageBuilder._get_hub_requirements(
docker_settings.required_hub_plugins
)
# Plugin packages themselves
for i, (index, packages) in enumerate(
hub_internal_requirements.items()
):
file_name = f".zenml_hub_internal_requirements_{i}"
file_lines = [f"-i {index}", *packages]
file_contents = "\n".join(file_lines)
requirements_files.append(
(file_name, file_contents, ["--no-deps"])
)
if log:
logger.info(
"- Including internal hub packages from index `%s`: %s",
index,
", ".join(f"`{r}`" for r in packages),
)
# PyPI requirements of plugin packages
if hub_pypi_requirements:
file_name = ".zenml_hub_pypi_requirements"
file_contents = "\n".join(hub_pypi_requirements)
requirements_files.append((file_name, file_contents, []))
if log:
logger.info(
"- Including hub requirements from PyPI: %s",
", ".join(f"`{r}`" for r in hub_pypi_requirements),
)
return requirements_files
proxy_utils
Proxy design pattern utils.
make_proxy_class(interface, attribute)
Proxy class decorator.
Use this decorator to transform the decorated class into a proxy that
forwards all calls defined in the interface interface to the attribute
class attribute that implements the same interface.
This class is useful in cases where you need to have a base class that acts as a proxy or facade for one or more other classes. Both the decorated class and the class attribute must inherit from the same ABC interface for this to work. Only regular methods are supported, not class methods or attributes.
Example: Let's say you have an interface called BodyBuilder, a base class
called FatBob and another class called BigJim. BigJim implements the
BodyBuilder interface, but FatBob does not. And let's say you want
FatBob to look as if it implements the BodyBuilder interface, but in
fact it just forwards all calls to BigJim. You could do this:
from abc import ABC, abstractmethod
class BodyBuilder(ABC):
@abstractmethod
def build_body(self):
pass
class BigJim(BodyBuilder):
def build_body(self):
print("Looks fit!")
class FatBob(BodyBuilder)
def __init__(self):
self.big_jim = BigJim()
def build_body(self):
self.big_jim.build_body()
fat_bob = FatBob()
fat_bob.build_body()
But this leads to a lot of boilerplate code with bigger interfaces and makes everything harder to maintain. This is where the proxy class decorator comes in handy. Here's how to use it:
from zenml.utils.proxy_utils import make_proxy_class
from typing import Optional
@make_proxy_class(BodyBuilder, "big_jim")
class FatBob(BodyBuilder)
big_jim: Optional[BodyBuilder] = None
def __init__(self):
self.big_jim = BigJim()
fat_bob = FatBob()
fat_bob.build_body()
This is the same as implementing FatBob to call BigJim explicitly, but it has the advantage that you don't need to write a lot of boilerplate code of modify the FatBob class every time you change something in the BodyBuilder interface.
This proxy decorator also allows to extend classes dynamically at runtime:
if the attribute class attribute is set to None, the proxy class
will assume that the interface is not implemented by the class and will
raise a NotImplementedError:
@make_proxy_class(BodyBuilder, "big_jim")
class FatBob(BodyBuilder)
big_jim: Optional[BodyBuilder] = None
def __init__(self):
self.big_jim = None
fat_bob = FatBob()
# Raises NotImplementedError, class not extended yet:
fat_bob.build_body()
fat_bob.big_jim = BigJim()
# Now it works:
fat_bob.build_body()
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
interface |
Type[abc.ABC] |
The interface to implement. |
required |
attribute |
str |
The attribute of the base class to forward calls to. |
required |
Returns:
| Type | Description |
|---|---|
Callable[[~C], ~C] |
The proxy class. |
Source code in zenml/utils/proxy_utils.py
def make_proxy_class(interface: Type[ABC], attribute: str) -> Callable[[C], C]:
"""Proxy class decorator.
Use this decorator to transform the decorated class into a proxy that
forwards all calls defined in the `interface` interface to the `attribute`
class attribute that implements the same interface.
This class is useful in cases where you need to have a base class that acts
as a proxy or facade for one or more other classes. Both the decorated class
and the class attribute must inherit from the same ABC interface for this to
work. Only regular methods are supported, not class methods or attributes.
Example: Let's say you have an interface called `BodyBuilder`, a base class
called `FatBob` and another class called `BigJim`. `BigJim` implements the
`BodyBuilder` interface, but `FatBob` does not. And let's say you want
`FatBob` to look as if it implements the `BodyBuilder` interface, but in
fact it just forwards all calls to `BigJim`. You could do this:
```python
from abc import ABC, abstractmethod
class BodyBuilder(ABC):
@abstractmethod
def build_body(self):
pass
class BigJim(BodyBuilder):
def build_body(self):
print("Looks fit!")
class FatBob(BodyBuilder)
def __init__(self):
self.big_jim = BigJim()
def build_body(self):
self.big_jim.build_body()
fat_bob = FatBob()
fat_bob.build_body()
```
But this leads to a lot of boilerplate code with bigger interfaces and
makes everything harder to maintain. This is where the proxy class
decorator comes in handy. Here's how to use it:
```python
from zenml.utils.proxy_utils import make_proxy_class
from typing import Optional
@make_proxy_class(BodyBuilder, "big_jim")
class FatBob(BodyBuilder)
big_jim: Optional[BodyBuilder] = None
def __init__(self):
self.big_jim = BigJim()
fat_bob = FatBob()
fat_bob.build_body()
```
This is the same as implementing FatBob to call BigJim explicitly, but it
has the advantage that you don't need to write a lot of boilerplate code
of modify the FatBob class every time you change something in the
BodyBuilder interface.
This proxy decorator also allows to extend classes dynamically at runtime:
if the `attribute` class attribute is set to None, the proxy class
will assume that the interface is not implemented by the class and will
raise a NotImplementedError:
```python
@make_proxy_class(BodyBuilder, "big_jim")
class FatBob(BodyBuilder)
big_jim: Optional[BodyBuilder] = None
def __init__(self):
self.big_jim = None
fat_bob = FatBob()
# Raises NotImplementedError, class not extended yet:
fat_bob.build_body()
fat_bob.big_jim = BigJim()
# Now it works:
fat_bob.build_body()
```
Args:
interface: The interface to implement.
attribute: The attribute of the base class to forward calls to.
Returns:
The proxy class.
"""
def make_proxy_method(cls: C, _method: F) -> F:
"""Proxy method decorator.
Used to transform a method into a proxy that forwards all calls to the
given class attribute.
Args:
cls: The class to use as the base.
_method: The method to replace.
Returns:
The proxy method.
"""
@wraps(_method)
def proxy_method(*args: Any, **kw: Any) -> Any:
"""Proxy method.
Args:
*args: The arguments to pass to the method.
**kw: The keyword arguments to pass to the method.
Returns:
The return value of the proxied method.
Raises:
TypeError: If the class does not have the attribute specified
in the decorator or if the attribute does not implement
the specified interface.
NotImplementedError: If the attribute specified in the
decorator is None, i.e. the interface is not implemented.
"""
self = args[0]
if not hasattr(self, attribute):
raise TypeError(
f"Class '{cls.__name__}' does not have a '{attribute}' "
f"as specified in the 'make_proxy_class' decorator."
)
proxied_obj = getattr(self, attribute)
if proxied_obj is None:
raise NotImplementedError(
f"This '{cls.__name__}' instance does not implement the "
f"'{interface.__name__}' interface."
)
if not isinstance(proxied_obj, interface):
raise TypeError(
f"Interface '{interface.__name__}' must be implemented by "
f"the '{cls.__name__}' '{attribute}' attribute."
)
proxied_method = getattr(proxied_obj, _method.__name__)
return proxied_method(*args[1:], **kw)
return cast(F, proxy_method)
def _inner_decorator(_cls: C) -> C:
"""Inner proxy class decorator.
Args:
_cls: The class to decorate.
Returns:
The decorated class.
Raises:
TypeError: If the decorated class does not implement the specified
interface.
"""
if not issubclass(_cls, interface):
raise TypeError(
f"Interface '{interface.__name__}' must be implemented by "
f"the '{_cls.__name__}' class."
)
for method_name in interface.__abstractmethods__:
original_method = getattr(_cls, method_name)
method_proxy = make_proxy_method(_cls, original_method)
# Make sure the proxy method is not considered abstract.
method_proxy.__isabstractmethod__ = False
setattr(_cls, method_name, method_proxy)
# Remove the abstract methods in the interface from the decorated class.
_cls.__abstractmethods__ = frozenset(
method_name
for method_name in _cls.__abstractmethods__
if method_name not in interface.__abstractmethods__
)
return cast(C, _cls)
return _inner_decorator
pydantic_utils
Utilities for pydantic models.
TemplateGenerator
Class to generate templates for pydantic models or classes.
Source code in zenml/utils/pydantic_utils.py
class TemplateGenerator:
"""Class to generate templates for pydantic models or classes."""
def __init__(
self, instance_or_class: Union[BaseModel, Type[BaseModel]]
) -> None:
"""Initializes the template generator.
Args:
instance_or_class: The pydantic model or model class for which to
generate a template.
"""
self.instance_or_class = instance_or_class
def run(self) -> Dict[str, Any]:
"""Generates the template.
Returns:
The template dictionary.
"""
if isinstance(self.instance_or_class, BaseModel):
template = self._generate_template_for_model(
self.instance_or_class
)
else:
template = self._generate_template_for_model_class(
self.instance_or_class
)
# Convert to json in an intermediate step so we can leverage Pydantic's
# encoder to support types like UUID and datetime
json_string = json.dumps(template, default=pydantic_encoder)
return cast(Dict[str, Any], json.loads(json_string))
def _generate_template_for_model(self, model: BaseModel) -> Dict[str, Any]:
"""Generates a template for a pydantic model.
Args:
model: The model for which to generate the template.
Returns:
The model template.
"""
template = self._generate_template_for_model_class(model.__class__)
for name in model.__fields_set__:
value = getattr(model, name)
template[name] = self._generate_template_for_value(value)
return template
def _generate_template_for_model_class(
self,
model_class: Type[BaseModel],
) -> Dict[str, Any]:
"""Generates a template for a pydantic model class.
Args:
model_class: The model class for which to generate the template.
Returns:
The model class template.
"""
template: Dict[str, Any] = {}
for name, field in model_class.__fields__.items():
if self._is_model_class(field.outer_type_):
template[name] = self._generate_template_for_model_class(
field.outer_type_
)
elif field.outer_type_ is Optional and self._is_model_class(
field.type_
):
template[name] = self._generate_template_for_model_class(
field.type_
)
else:
template[name] = field._type_display()
return template
def _generate_template_for_value(self, value: Any) -> Any:
"""Generates a template for an arbitrary value.
Args:
value: The value for which to generate the template.
Returns:
The value template.
"""
if isinstance(value, Dict):
return {
k: self._generate_template_for_value(v)
for k, v in value.items()
}
elif sequence_like(value):
return [self._generate_template_for_value(v) for v in value]
elif isinstance(value, BaseModel):
return self._generate_template_for_model(value)
else:
return value
@staticmethod
def _is_model_class(value: Any) -> bool:
"""Checks if the given value is a pydantic model class.
Args:
value: The value to check.
Returns:
If the value is a pydantic model class.
"""
return isinstance(value, type) and issubclass(value, BaseModel)
__init__(self, instance_or_class)
special
Initializes the template generator.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
instance_or_class |
Union[pydantic.main.BaseModel, Type[pydantic.main.BaseModel]] |
The pydantic model or model class for which to generate a template. |
required |
Source code in zenml/utils/pydantic_utils.py
def __init__(
self, instance_or_class: Union[BaseModel, Type[BaseModel]]
) -> None:
"""Initializes the template generator.
Args:
instance_or_class: The pydantic model or model class for which to
generate a template.
"""
self.instance_or_class = instance_or_class
run(self)
Generates the template.
Returns:
| Type | Description |
|---|---|
Dict[str, Any] |
The template dictionary. |
Source code in zenml/utils/pydantic_utils.py
def run(self) -> Dict[str, Any]:
"""Generates the template.
Returns:
The template dictionary.
"""
if isinstance(self.instance_or_class, BaseModel):
template = self._generate_template_for_model(
self.instance_or_class
)
else:
template = self._generate_template_for_model_class(
self.instance_or_class
)
# Convert to json in an intermediate step so we can leverage Pydantic's
# encoder to support types like UUID and datetime
json_string = json.dumps(template, default=pydantic_encoder)
return cast(Dict[str, Any], json.loads(json_string))
YAMLSerializationMixin (BaseModel)
pydantic-model
Class to serialize/deserialize pydantic models to/from YAML.
Source code in zenml/utils/pydantic_utils.py
class YAMLSerializationMixin(BaseModel):
"""Class to serialize/deserialize pydantic models to/from YAML."""
def yaml(self, sort_keys: bool = False, **kwargs: Any) -> str:
"""YAML string representation..
Args:
sort_keys: Whether to sort the keys in the YAML representation.
**kwargs: Kwargs to pass to the pydantic json(...) method.
Returns:
YAML string representation.
"""
dict_ = json.loads(self.json(**kwargs, sort_keys=sort_keys))
return yaml.dump(dict_, sort_keys=sort_keys)
@classmethod
def from_yaml(cls: Type[M], path: str) -> M:
"""Creates an instance from a YAML file.
Args:
path: Path to a YAML file.
Returns:
The model instance.
"""
dict_ = yaml_utils.read_yaml(path)
return cls.parse_obj(dict_)
from_yaml(path)
classmethod
Creates an instance from a YAML file.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path |
str |
Path to a YAML file. |
required |
Returns:
| Type | Description |
|---|---|
~M |
The model instance. |
Source code in zenml/utils/pydantic_utils.py
@classmethod
def from_yaml(cls: Type[M], path: str) -> M:
"""Creates an instance from a YAML file.
Args:
path: Path to a YAML file.
Returns:
The model instance.
"""
dict_ = yaml_utils.read_yaml(path)
return cls.parse_obj(dict_)
yaml(self, sort_keys=False, **kwargs)
YAML string representation..
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
sort_keys |
bool |
Whether to sort the keys in the YAML representation. |
False |
**kwargs |
Any |
Kwargs to pass to the pydantic json(...) method. |
{} |
Returns:
| Type | Description |
|---|---|
str |
YAML string representation. |
Source code in zenml/utils/pydantic_utils.py
def yaml(self, sort_keys: bool = False, **kwargs: Any) -> str:
"""YAML string representation..
Args:
sort_keys: Whether to sort the keys in the YAML representation.
**kwargs: Kwargs to pass to the pydantic json(...) method.
Returns:
YAML string representation.
"""
dict_ = json.loads(self.json(**kwargs, sort_keys=sort_keys))
return yaml.dump(dict_, sort_keys=sort_keys)
update_model(original, update, recursive=True, exclude_none=True)
Updates a pydantic model.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
original |
~M |
The model to update. |
required |
update |
Union[BaseModel, Dict[str, Any]] |
The update values. |
required |
recursive |
bool |
If |
True |
exclude_none |
bool |
If |
True |
Returns:
| Type | Description |
|---|---|
~M |
The updated model. |
Source code in zenml/utils/pydantic_utils.py
def update_model(
original: M,
update: Union["BaseModel", Dict[str, Any]],
recursive: bool = True,
exclude_none: bool = True,
) -> M:
"""Updates a pydantic model.
Args:
original: The model to update.
update: The update values.
recursive: If `True`, dictionary values will be updated recursively.
exclude_none: If `True`, `None` values in the update dictionary
will be removed.
Returns:
The updated model.
"""
if isinstance(update, Dict):
if exclude_none:
update_dict = dict_utils.remove_none_values(
update, recursive=recursive
)
else:
update_dict = update
else:
update_dict = update.dict(exclude_unset=True)
original_dict = original.dict(exclude_unset=True)
if recursive:
values = dict_utils.recursive_update(original_dict, update_dict)
else:
values = {**original_dict, **update_dict}
return original.__class__(**values)
validate_function_args(__func, __config, *args, **kwargs)
Validates arguments passed to a function.
This function validates that all arguments to call the function exist and that the types match.
It raises a pydantic.ValidationError if the validation fails.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
__func |
Callable[..., Any] |
The function for which the arguments are passed. |
required |
__config |
Dict[str, Any] |
The pydantic config for the underlying model that is created to validate the types of the arguments. |
required |
*args |
Any |
Function arguments. |
() |
**kwargs |
Any |
Function keyword arguments. |
{} |
Returns:
| Type | Description |
|---|---|
Dict[str, Any] |
The validated arguments. |
Source code in zenml/utils/pydantic_utils.py
def validate_function_args(
__func: Callable[..., Any],
__config: Dict[str, Any],
*args: Any,
**kwargs: Any,
) -> Dict[str, Any]:
"""Validates arguments passed to a function.
This function validates that all arguments to call the function exist and
that the types match.
It raises a pydantic.ValidationError if the validation fails.
Args:
__func: The function for which the arguments are passed.
__config: The pydantic config for the underlying model that is created
to validate the types of the arguments.
*args: Function arguments.
**kwargs: Function keyword arguments.
Returns:
The validated arguments.
"""
parameter_prefix = "zenml__"
signature = inspect.signature(__func)
parameters = [
param.replace(name=f"{parameter_prefix}{param.name}")
for param in signature.parameters.values()
]
signature = signature.replace(parameters=parameters)
def f() -> None:
pass
# We create a dummy function with the original function signature, but
# add a prefix to all arguments to avoid potential clashes with pydantic
# BaseModel attributes
f.__signature__ = signature # type: ignore[attr-defined]
f.__annotations__ = {
f"{parameter_prefix}{key}": annotation
for key, annotation in __func.__annotations__.items()
}
validation_func = ValidatedFunction(f, config=__config)
kwargs = {
f"{parameter_prefix}{key}": value for key, value in kwargs.items()
}
model = validation_func.init_model_instance(*args, **kwargs)
validated_args = {
k[len(parameter_prefix) :]: v
for k, v in model._iter()
if k in model.__fields_set__
or model.__fields__[k].default_factory
or model.__fields__[k].default
}
return validated_args
secret_utils
Utility functions for secrets and secret references.
SecretReference (tuple)
Class representing a secret reference.
Attributes:
| Name | Type | Description |
|---|---|---|
name |
str |
The secret name. |
key |
str |
The secret key. |
Source code in zenml/utils/secret_utils.py
class SecretReference(NamedTuple):
"""Class representing a secret reference.
Attributes:
name: The secret name.
key: The secret key.
"""
name: str
key: str
__getnewargs__(self)
special
Return self as a plain tuple. Used by copy and pickle.
Source code in zenml/utils/secret_utils.py
def __getnewargs__(self):
'Return self as a plain tuple. Used by copy and pickle.'
return _tuple(self)
__new__(_cls, name, key)
special
staticmethod
Create new instance of SecretReference(name, key)
__repr__(self)
special
Return a nicely formatted representation string
Source code in zenml/utils/secret_utils.py
def __repr__(self):
'Return a nicely formatted representation string'
return self.__class__.__name__ + repr_fmt % self
ClearTextField(*args, **kwargs)
Marks a pydantic field to prevent secret references.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
*args |
Any |
Positional arguments which will be forwarded
to |
() |
**kwargs |
Any |
Keyword arguments which will be forwarded to
|
{} |
Returns:
| Type | Description |
|---|---|
Any |
Pydantic field info. |
Source code in zenml/utils/secret_utils.py
def ClearTextField(*args: Any, **kwargs: Any) -> Any:
"""Marks a pydantic field to prevent secret references.
Args:
*args: Positional arguments which will be forwarded
to `pydantic.Field(...)`.
**kwargs: Keyword arguments which will be forwarded to
`pydantic.Field(...)`.
Returns:
Pydantic field info.
"""
kwargs[PYDANTIC_CLEAR_TEXT_FIELD_MARKER] = True
return Field(*args, **kwargs) # type: ignore[pydantic-field]
SecretField(*args, **kwargs)
Marks a pydantic field as something containing sensitive information.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
*args |
Any |
Positional arguments which will be forwarded
to |
() |
**kwargs |
Any |
Keyword arguments which will be forwarded to
|
{} |
Returns:
| Type | Description |
|---|---|
Any |
Pydantic field info. |
Source code in zenml/utils/secret_utils.py
def SecretField(*args: Any, **kwargs: Any) -> Any:
"""Marks a pydantic field as something containing sensitive information.
Args:
*args: Positional arguments which will be forwarded
to `pydantic.Field(...)`.
**kwargs: Keyword arguments which will be forwarded to
`pydantic.Field(...)`.
Returns:
Pydantic field info.
"""
kwargs[PYDANTIC_SENSITIVE_FIELD_MARKER] = True
return Field(*args, **kwargs) # type: ignore[pydantic-field]
is_clear_text_field(field)
Returns whether a pydantic field prevents secret references or not.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
field |
ModelField |
The field to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
|
Source code in zenml/utils/secret_utils.py
def is_clear_text_field(field: "ModelField") -> bool:
"""Returns whether a pydantic field prevents secret references or not.
Args:
field: The field to check.
Returns:
`True` if the field prevents secret references, `False` otherwise.
"""
return field.field_info.extra.get(PYDANTIC_CLEAR_TEXT_FIELD_MARKER, False)
is_secret_field(field)
Returns whether a pydantic field contains sensitive information or not.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
field |
ModelField |
The field to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
|
Source code in zenml/utils/secret_utils.py
def is_secret_field(field: "ModelField") -> bool:
"""Returns whether a pydantic field contains sensitive information or not.
Args:
field: The field to check.
Returns:
`True` if the field contains sensitive information, `False` otherwise.
"""
return field.field_info.extra.get(PYDANTIC_SENSITIVE_FIELD_MARKER, False)
is_secret_reference(value)
Checks whether any value is a secret reference.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
value |
Any |
The value to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
|
Source code in zenml/utils/secret_utils.py
def is_secret_reference(value: Any) -> bool:
"""Checks whether any value is a secret reference.
Args:
value: The value to check.
Returns:
`True` if the value is a secret reference, `False` otherwise.
"""
if not isinstance(value, str):
return False
return bool(_secret_reference_expression.fullmatch(value))
parse_secret_reference(reference)
Parses a secret reference.
This function assumes the input string is a valid secret reference and does not perform any additional checks. If you pass an invalid secret reference here, this will most likely crash.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
reference |
str |
The string representing a valid secret reference. |
required |
Returns:
| Type | Description |
|---|---|
SecretReference |
The parsed secret reference. |
Source code in zenml/utils/secret_utils.py
def parse_secret_reference(reference: str) -> SecretReference:
"""Parses a secret reference.
This function assumes the input string is a valid secret reference and
**does not** perform any additional checks. If you pass an invalid secret
reference here, this will most likely crash.
Args:
reference: The string representing a **valid** secret reference.
Returns:
The parsed secret reference.
"""
reference = reference[2:]
reference = reference[:-2]
secret_name, secret_key = reference.split(".", 1)
secret_name, secret_key = secret_name.strip(), secret_key.strip()
return SecretReference(name=secret_name, key=secret_key)
settings_utils
Utility functions for ZenML settings.
get_flavor_setting_key(flavor)
Gets the setting key for a flavor.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
flavor |
Flavor |
The flavor for which to get the key. |
required |
Returns:
| Type | Description |
|---|---|
str |
The setting key for the flavor. |
Source code in zenml/utils/settings_utils.py
def get_flavor_setting_key(flavor: "Flavor") -> str:
"""Gets the setting key for a flavor.
Args:
flavor: The flavor for which to get the key.
Returns:
The setting key for the flavor.
"""
return f"{flavor.type}.{flavor.name}"
get_general_settings()
Returns all general settings.
Returns:
| Type | Description |
|---|---|
Dict[str, Type[BaseSettings]] |
Dictionary mapping general settings keys to their type. |
Source code in zenml/utils/settings_utils.py
def get_general_settings() -> Dict[str, Type["BaseSettings"]]:
"""Returns all general settings.
Returns:
Dictionary mapping general settings keys to their type.
"""
from zenml.config import DockerSettings, ResourceSettings
return {
DOCKER_SETTINGS_KEY: DockerSettings,
RESOURCE_SETTINGS_KEY: ResourceSettings,
}
get_stack_component_for_settings_key(key, stack)
Gets the stack component of a stack for a given settings key.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key |
str |
The settings key for which to get the component. |
required |
stack |
Stack |
The stack from which to get the component. |
required |
Exceptions:
| Type | Description |
|---|---|
ValueError |
If the key is invalid or the stack does not contain a component of the correct flavor. |
Returns:
| Type | Description |
|---|---|
StackComponent |
The stack component. |
Source code in zenml/utils/settings_utils.py
def get_stack_component_for_settings_key(
key: str, stack: "Stack"
) -> "StackComponent":
"""Gets the stack component of a stack for a given settings key.
Args:
key: The settings key for which to get the component.
stack: The stack from which to get the component.
Raises:
ValueError: If the key is invalid or the stack does not contain a
component of the correct flavor.
Returns:
The stack component.
"""
if not is_stack_component_setting_key(key):
raise ValueError(
f"Settings key {key} does not refer to a stack component."
)
component_type, flavor = key.split(".", 1)
stack_component = stack.components.get(StackComponentType(component_type))
if not stack_component or stack_component.flavor != flavor:
raise ValueError(
f"Component of type {component_type} in stack {stack} is not "
f"of the flavor {flavor} specified by the settings key {key}."
)
return stack_component
get_stack_component_setting_key(stack_component)
Gets the setting key for a stack component.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
stack_component |
StackComponent |
The stack component for which to get the key. |
required |
Returns:
| Type | Description |
|---|---|
str |
The setting key for the stack component. |
Source code in zenml/utils/settings_utils.py
def get_stack_component_setting_key(stack_component: "StackComponent") -> str:
"""Gets the setting key for a stack component.
Args:
stack_component: The stack component for which to get the key.
Returns:
The setting key for the stack component.
"""
return f"{stack_component.type}.{stack_component.flavor}"
is_general_setting_key(key)
Checks whether the key refers to a general setting.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key |
str |
The key to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
If the key refers to a general setting. |
Source code in zenml/utils/settings_utils.py
def is_general_setting_key(key: str) -> bool:
"""Checks whether the key refers to a general setting.
Args:
key: The key to check.
Returns:
If the key refers to a general setting.
"""
return key in get_general_settings()
is_stack_component_setting_key(key)
Checks whether a settings key refers to a stack component.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key |
str |
The key to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
If the key refers to a stack component. |
Source code in zenml/utils/settings_utils.py
def is_stack_component_setting_key(key: str) -> bool:
"""Checks whether a settings key refers to a stack component.
Args:
key: The key to check.
Returns:
If the key refers to a stack component.
"""
return bool(STACK_COMPONENT_REGEX.fullmatch(key))
is_valid_setting_key(key)
Checks whether a settings key is valid.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key |
str |
The key to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
If the key is valid. |
Source code in zenml/utils/settings_utils.py
def is_valid_setting_key(key: str) -> bool:
"""Checks whether a settings key is valid.
Args:
key: The key to check.
Returns:
If the key is valid.
"""
return is_general_setting_key(key) or is_stack_component_setting_key(key)
validate_setting_keys(setting_keys)
Validates settings keys.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
setting_keys |
Sequence[str] |
The keys to validate. |
required |
Exceptions:
| Type | Description |
|---|---|
ValueError |
If any key is invalid. |
Source code in zenml/utils/settings_utils.py
def validate_setting_keys(setting_keys: Sequence[str]) -> None:
"""Validates settings keys.
Args:
setting_keys: The keys to validate.
Raises:
ValueError: If any key is invalid.
"""
for key in setting_keys:
if not is_valid_setting_key(key):
raise ValueError(
f"Invalid setting key `{key}`. Setting keys can either refer "
"to general settings (available keys: "
f"{set(get_general_settings())}) or stack component specific "
"settings. Stack component specific keys are of the format "
"`<STACK_COMPONENT_TYPE>.<STACK_COMPONENT_FLAVOR>`."
)
singleton
Utility class to turn classes into singleton classes.
SingletonMetaClass (type)
Singleton metaclass.
Use this metaclass to make any class into a singleton class:
class OneRing(metaclass=SingletonMetaClass):
def __init__(self, owner):
self._owner = owner
@property
def owner(self):
return self._owner
the_one_ring = OneRing('Sauron')
the_lost_ring = OneRing('Frodo')
print(the_lost_ring.owner) # Sauron
OneRing._clear() # ring destroyed
Source code in zenml/utils/singleton.py
class SingletonMetaClass(type):
"""Singleton metaclass.
Use this metaclass to make any class into a singleton class:
```python
class OneRing(metaclass=SingletonMetaClass):
def __init__(self, owner):
self._owner = owner
@property
def owner(self):
return self._owner
the_one_ring = OneRing('Sauron')
the_lost_ring = OneRing('Frodo')
print(the_lost_ring.owner) # Sauron
OneRing._clear() # ring destroyed
```
"""
def __init__(cls, *args: Any, **kwargs: Any) -> None:
"""Initialize a singleton class.
Args:
*args: Additional arguments.
**kwargs: Additional keyword arguments.
"""
super().__init__(*args, **kwargs)
cls.__singleton_instance: Optional["SingletonMetaClass"] = None
def __call__(cls, *args: Any, **kwargs: Any) -> "SingletonMetaClass":
"""Create or return the singleton instance.
Args:
*args: Additional arguments.
**kwargs: Additional keyword arguments.
Returns:
The singleton instance.
"""
if not cls.__singleton_instance:
cls.__singleton_instance = cast(
"SingletonMetaClass", super().__call__(*args, **kwargs)
)
return cls.__singleton_instance
def _clear(cls) -> None:
"""Clear the singleton instance."""
cls.__singleton_instance = None
def _exists(cls) -> bool:
"""Check if the singleton instance exists.
Returns:
`True` if the singleton instance exists, `False` otherwise.
"""
return cls.__singleton_instance is not None
__call__(cls, *args, **kwargs)
special
Create or return the singleton instance.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
*args |
Any |
Additional arguments. |
() |
**kwargs |
Any |
Additional keyword arguments. |
{} |
Returns:
| Type | Description |
|---|---|
SingletonMetaClass |
The singleton instance. |
Source code in zenml/utils/singleton.py
def __call__(cls, *args: Any, **kwargs: Any) -> "SingletonMetaClass":
"""Create or return the singleton instance.
Args:
*args: Additional arguments.
**kwargs: Additional keyword arguments.
Returns:
The singleton instance.
"""
if not cls.__singleton_instance:
cls.__singleton_instance = cast(
"SingletonMetaClass", super().__call__(*args, **kwargs)
)
return cls.__singleton_instance
__init__(cls, *args, **kwargs)
special
Initialize a singleton class.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
*args |
Any |
Additional arguments. |
() |
**kwargs |
Any |
Additional keyword arguments. |
{} |
Source code in zenml/utils/singleton.py
def __init__(cls, *args: Any, **kwargs: Any) -> None:
"""Initialize a singleton class.
Args:
*args: Additional arguments.
**kwargs: Additional keyword arguments.
"""
super().__init__(*args, **kwargs)
cls.__singleton_instance: Optional["SingletonMetaClass"] = None
source_code_utils
Utilities for getting the source code of objects.
get_hashed_source_code(value)
Returns a hash of the objects source code.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
value |
Any |
object to get source from. |
required |
Returns:
| Type | Description |
|---|---|
str |
Hash of source code. |
Exceptions:
| Type | Description |
|---|---|
TypeError |
If unable to compute the hash. |
Source code in zenml/utils/source_code_utils.py
def get_hashed_source_code(value: Any) -> str:
"""Returns a hash of the objects source code.
Args:
value: object to get source from.
Returns:
Hash of source code.
Raises:
TypeError: If unable to compute the hash.
"""
try:
source_code = get_source_code(value)
except TypeError:
raise TypeError(
f"Unable to compute the hash of source code of object: {value}."
)
return hashlib.sha256(source_code.encode("utf-8")).hexdigest()
get_source_code(value)
Returns the source code of an object.
If executing within a IPython kernel environment, then this monkey-patches
inspect module temporarily with a workaround to get source from the cell.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
value |
Any |
object to get source from. |
required |
Returns:
| Type | Description |
|---|---|
str |
Source code of object. |
Source code in zenml/utils/source_code_utils.py
def get_source_code(value: Any) -> str:
"""Returns the source code of an object.
If executing within a IPython kernel environment, then this monkey-patches
`inspect` module temporarily with a workaround to get source from the cell.
Args:
value: object to get source from.
Returns:
Source code of object.
"""
if Environment.in_notebook():
# Monkey patch inspect.getfile temporarily to make getsource work.
# Source: https://stackoverflow.com/questions/51566497/
def _new_getfile(
object: Any,
_old_getfile: Callable[
[
Union[
ModuleType,
Type[Any],
MethodType,
FunctionType,
TracebackType,
FrameType,
CodeType,
Callable[..., Any],
]
],
str,
] = inspect.getfile,
) -> Any:
if not inspect.isclass(object):
return _old_getfile(object)
# Lookup by parent module (as in current inspect)
if hasattr(object, "__module__"):
object_ = sys.modules.get(object.__module__)
if hasattr(object_, "__file__"):
return object_.__file__ # type: ignore[union-attr]
# If parent module is __main__, lookup by methods
for _, member in inspect.getmembers(object):
if (
inspect.isfunction(member)
and object.__qualname__ + "." + member.__name__
== member.__qualname__
):
return inspect.getfile(member)
else:
raise TypeError(f"Source for {object!r} not found.")
# Monkey patch, compute source, then revert monkey patch.
_old_getfile = inspect.getfile
inspect.getfile = _new_getfile
try:
src = inspect.getsource(value)
finally:
inspect.getfile = _old_getfile
else:
# Use standard inspect if running outside a notebook
src = inspect.getsource(value)
return src
source_utils
Utilities for loading/resolving objects.
get_source_root()
Get the source root.
The source root will be determined in the following order: - The manually specified custom source root if it was set. - The ZenML repository directory if one exists in the current working directory or any parent directories. - The parent directory of the main module file.
Returns:
| Type | Description |
|---|---|
str |
The source root. |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If the main module file can't be found. |
Source code in zenml/utils/source_utils.py
def get_source_root() -> str:
"""Get the source root.
The source root will be determined in the following order:
- The manually specified custom source root if it was set.
- The ZenML repository directory if one exists in the current working
directory or any parent directories.
- The parent directory of the main module file.
Returns:
The source root.
Raises:
RuntimeError: If the main module file can't be found.
"""
if _CUSTOM_SOURCE_ROOT:
logger.debug("Using custom source root: %s", _CUSTOM_SOURCE_ROOT)
return _CUSTOM_SOURCE_ROOT
from zenml.client import Client
repo_root = Client.find_repository()
if repo_root:
logger.debug("Using repository root as source root: %s", repo_root)
return str(repo_root.resolve())
main_module = sys.modules.get("__main__")
if main_module is None:
raise RuntimeError(
"Unable to determine source root because the main module could not "
"be found."
)
if not hasattr(main_module, "__file__") or not main_module.__file__:
raise RuntimeError(
"Unable to determine source root because the main module does not "
"have an associated file. This could be because you're running in "
"an interactive Python environment."
)
path = Path(main_module.__file__).resolve().parent
logger.debug("Using main module parent directory as source root: %s", path)
return str(path)
get_source_type(module)
Get the type of a source.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
module |
module |
The module for which to get the source type. |
required |
Returns:
| Type | Description |
|---|---|
SourceType |
The source type. |
Source code in zenml/utils/source_utils.py
def get_source_type(module: ModuleType) -> SourceType:
"""Get the type of a source.
Args:
module: The module for which to get the source type.
Returns:
The source type.
"""
try:
file_path = inspect.getfile(module)
except (TypeError, OSError):
if module.__name__ == "__main__" and Environment.in_notebook():
return SourceType.USER
return SourceType.BUILTIN
if is_internal_module(module_name=module.__name__):
return SourceType.INTERNAL
if is_standard_lib_file(file_path=file_path):
return SourceType.BUILTIN
if is_distribution_package_file(
file_path=file_path, module_name=module.__name__
):
return SourceType.DISTRIBUTION_PACKAGE
# Make sure to check for distribution packages before this to catch the
# case when a virtual environment is inside our source root
if is_user_file(file_path=file_path):
return SourceType.USER
return SourceType.UNKNOWN
is_distribution_package_file(file_path, module_name)
Checks if a file/module belongs to a distribution package.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
The file path to check. |
required |
module_name |
str |
The module name. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if the file/module belongs to a distribution package, False otherwise. |
Source code in zenml/utils/source_utils.py
def is_distribution_package_file(file_path: str, module_name: str) -> bool:
"""Checks if a file/module belongs to a distribution package.
Args:
file_path: The file path to check.
module_name: The module name.
Returns:
True if the file/module belongs to a distribution package, False
otherwise.
"""
absolute_file_path = Path(file_path).resolve()
for path in site.getsitepackages() + [site.getusersitepackages()]:
if Path(path).resolve() in absolute_file_path.parents:
return True
# TODO: The previous check does not detect editable installs because
# the site packages dir only contains a reference to the source files,
# not the actual files. That means currently editable installs get a
# source type UNKNOWN which might or might not lead to issues.
return False
is_internal_module(module_name)
Checks if a module is internal (=part of the zenml package).
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
module_name |
str |
Name of the module to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if the module is internal, False otherwise. |
Source code in zenml/utils/source_utils.py
def is_internal_module(module_name: str) -> bool:
"""Checks if a module is internal (=part of the zenml package).
Args:
module_name: Name of the module to check.
Returns:
True if the module is internal, False otherwise.
"""
return module_name.split(".", maxsplit=1)[0] == "zenml"
is_standard_lib_file(file_path)
Checks if a file belongs to the Python standard library.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
The file path to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if the file belongs to the Python standard library, False otherwise. |
Source code in zenml/utils/source_utils.py
def is_standard_lib_file(file_path: str) -> bool:
"""Checks if a file belongs to the Python standard library.
Args:
file_path: The file path to check.
Returns:
True if the file belongs to the Python standard library, False
otherwise.
"""
stdlib_root = get_python_lib(standard_lib=True)
return Path(stdlib_root).resolve() in Path(file_path).resolve().parents
is_user_file(file_path)
Checks if a file is a user file.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
The file path to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if the file is a user file, False otherwise. |
Source code in zenml/utils/source_utils.py
def is_user_file(file_path: str) -> bool:
"""Checks if a file is a user file.
Args:
file_path: The file path to check.
Returns:
True if the file is a user file, False otherwise.
"""
source_root = get_source_root()
return Path(source_root) in Path(file_path).resolve().parents
load(source)
Load a source or import path.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
source |
Union[zenml.config.source.Source, str] |
The source to load. |
required |
Returns:
| Type | Description |
|---|---|
Any |
The loaded object. |
Source code in zenml/utils/source_utils.py
def load(source: Union[Source, str]) -> Any:
"""Load a source or import path.
Args:
source: The source to load.
Returns:
The loaded object.
"""
if isinstance(source, str):
source = Source.from_import_path(source)
if source.import_path == NoneTypeSource.import_path:
# The class of the `None` object doesn't exist in the `builtin` module
# so we need to manually handle it here
return NoneType
import_root = None
if source.type == SourceType.CODE_REPOSITORY:
source = CodeRepositorySource.parse_obj(source)
_warn_about_potential_source_loading_issues(source=source)
import_root = get_source_root()
elif source.type == SourceType.DISTRIBUTION_PACKAGE:
source = DistributionPackageSource.parse_obj(source)
if source.version:
current_package_version = _get_package_version(
package_name=source.package_name
)
if current_package_version != source.version:
logger.warning(
"The currently installed version `%s` of package `%s` "
"does not match the source version `%s`. This might lead "
"to unexpected behavior when using the source object `%s`.",
current_package_version,
source.package_name,
source.version,
source.import_path,
)
elif source.type in {SourceType.USER, SourceType.UNKNOWN}:
# Unknown source might also refer to a user file, include source
# root in python path just to be sure
import_root = get_source_root()
module = _load_module(module_name=source.module, import_root=import_root)
if source.attribute:
obj = getattr(module, source.attribute)
else:
obj = module
return obj
load_and_validate_class(source, expected_class)
Loads a source class and validates its class.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
source |
Union[str, zenml.config.source.Source] |
The source. |
required |
expected_class |
Type[Any] |
The class that the source should resolve to. |
required |
Exceptions:
| Type | Description |
|---|---|
TypeError |
If the source does not resolve to the expected class. |
Returns:
| Type | Description |
|---|---|
Type[Any] |
The resolved source class. |
Source code in zenml/utils/source_utils.py
def load_and_validate_class(
source: Union[str, Source], expected_class: Type[Any]
) -> Type[Any]:
"""Loads a source class and validates its class.
Args:
source: The source.
expected_class: The class that the source should resolve to.
Raises:
TypeError: If the source does not resolve to the expected class.
Returns:
The resolved source class.
"""
obj = load(source)
if isinstance(obj, type) and issubclass(obj, expected_class):
return obj
else:
raise TypeError(
f"Error while loading `{source}`. Expected class "
f"{expected_class.__name__}, got {obj} instead."
)
prepend_python_path(path)
Context manager to temporarily prepend a path to the python path.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path |
str |
Path that will be prepended to sys.path for the duration of the context manager. |
required |
Yields:
| Type | Description |
|---|---|
Iterator[NoneType] |
None |
Source code in zenml/utils/source_utils.py
@contextlib.contextmanager
def prepend_python_path(path: str) -> Iterator[None]:
"""Context manager to temporarily prepend a path to the python path.
Args:
path: Path that will be prepended to sys.path for the duration of
the context manager.
Yields:
None
"""
try:
sys.path.insert(0, path)
yield
finally:
sys.path.remove(path)
resolve(obj, skip_validation=False)
Resolve an object.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
obj |
Union[Type[Any], Callable[..., Any], module] |
The object to resolve. |
required |
skip_validation |
bool |
If True, the validation that the object exist in the module is skipped. |
False |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If the object can't be resolved. |
Returns:
| Type | Description |
|---|---|
Source |
The source of the resolved object. |
Source code in zenml/utils/source_utils.py
def resolve(
obj: Union[Type[Any], Callable[..., Any], ModuleType, NoneType],
skip_validation: bool = False,
) -> Source:
"""Resolve an object.
Args:
obj: The object to resolve.
skip_validation: If True, the validation that the object exist in the
module is skipped.
Raises:
RuntimeError: If the object can't be resolved.
Returns:
The source of the resolved object.
"""
if obj is NoneType: # type: ignore[comparison-overlap]
# The class of the `None` object doesn't exist in the `builtin` module
# so we need to manually handle it here
return NoneTypeSource
elif isinstance(obj, ModuleType):
module = obj
attribute_name = None
else:
module = sys.modules[obj.__module__]
attribute_name = obj.__name__ # type: ignore[union-attr]
if (
not skip_validation
and attribute_name
and getattr(module, attribute_name, None) is not obj
):
raise RuntimeError(
f"Unable to resolve object `{obj}`. For the resolving to work, the "
"class or function must be defined as top-level code (= it must "
"get defined when importing the module) and not inside a function/"
f"if-condition. Please make sure that your `{module.__name__}` "
f"module has a top-level attribute `{attribute_name}` that "
"holds the object you want to resolve."
)
module_name = module.__name__
if module_name == "__main__":
module_name = _resolve_module(module)
source_type = get_source_type(module=module)
if source_type == SourceType.USER:
from zenml.utils import code_repository_utils
local_repo_context = (
code_repository_utils.find_active_code_repository()
)
if local_repo_context and not local_repo_context.has_local_changes:
module_name = _resolve_module(module)
source_root = get_source_root()
subdir = PurePath(source_root).relative_to(local_repo_context.root)
return CodeRepositorySource(
repository_id=local_repo_context.code_repository_id,
commit=local_repo_context.current_commit,
subdirectory=subdir.as_posix(),
module=module_name,
attribute=attribute_name,
)
module_name = _resolve_module(module)
elif source_type == SourceType.DISTRIBUTION_PACKAGE:
package_name = _get_package_for_module(module_name=module_name)
if package_name:
package_version = _get_package_version(package_name=package_name)
return DistributionPackageSource(
module=module_name,
attribute=attribute_name,
package_name=package_name,
version=package_version,
type=source_type,
)
else:
# Fallback to an unknown source if we can't find the package
source_type = SourceType.UNKNOWN
return Source(
module=module_name, attribute=attribute_name, type=source_type
)
set_custom_source_root(source_root)
Sets a custom source root.
If set this has the highest priority and will always be used as the source root.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
source_root |
Optional[str] |
The source root to use. |
required |
Source code in zenml/utils/source_utils.py
def set_custom_source_root(source_root: Optional[str]) -> None:
"""Sets a custom source root.
If set this has the highest priority and will always be used as the source
root.
Args:
source_root: The source root to use.
"""
logger.debug("Setting custom source root: %s", source_root)
global _CUSTOM_SOURCE_ROOT
_CUSTOM_SOURCE_ROOT = source_root
validate_source_class(source, expected_class)
Validates that a source resolves to a certain class.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
source |
Union[zenml.config.source.Source, str] |
The source to validate. |
required |
expected_class |
Type[Any] |
The class that the source should resolve to. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if the source resolves to the expected class, False otherwise. |
Source code in zenml/utils/source_utils.py
def validate_source_class(
source: Union[Source, str], expected_class: Type[Any]
) -> bool:
"""Validates that a source resolves to a certain class.
Args:
source: The source to validate.
expected_class: The class that the source should resolve to.
Returns:
True if the source resolves to the expected class, False otherwise.
"""
try:
obj = load(source)
except Exception:
return False
if isinstance(obj, type) and issubclass(obj, expected_class):
return True
else:
return False
string_utils
Utils for strings.
b64_decode(input_)
Returns a decoded string of the base 64 encoded input string.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
input_ |
str |
Base64 encoded string. |
required |
Returns:
| Type | Description |
|---|---|
str |
Decoded string. |
Source code in zenml/utils/string_utils.py
def b64_decode(input_: str) -> str:
"""Returns a decoded string of the base 64 encoded input string.
Args:
input_: Base64 encoded string.
Returns:
Decoded string.
"""
encoded_bytes = input_.encode()
decoded_bytes = base64.b64decode(encoded_bytes)
return decoded_bytes.decode()
b64_encode(input_)
Returns a base 64 encoded string of the input string.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
input_ |
str |
The input to encode. |
required |
Returns:
| Type | Description |
|---|---|
str |
Base64 encoded string. |
Source code in zenml/utils/string_utils.py
def b64_encode(input_: str) -> str:
"""Returns a base 64 encoded string of the input string.
Args:
input_: The input to encode.
Returns:
Base64 encoded string.
"""
input_bytes = input_.encode()
encoded_bytes = base64.b64encode(input_bytes)
return encoded_bytes.decode()
get_human_readable_filesize(bytes_)
Convert a file size in bytes into a human-readable string.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
bytes_ |
int |
The number of bytes to convert. |
required |
Returns:
| Type | Description |
|---|---|
str |
A human-readable string. |
Source code in zenml/utils/string_utils.py
def get_human_readable_filesize(bytes_: int) -> str:
"""Convert a file size in bytes into a human-readable string.
Args:
bytes_: The number of bytes to convert.
Returns:
A human-readable string.
"""
size = abs(float(bytes_))
for unit in ["B", "KiB", "MiB", "GiB"]:
if size < 1024.0 or unit == "GiB":
break
size /= 1024.0
return f"{size:.2f} {unit}"
get_human_readable_time(seconds)
Convert seconds into a human-readable string.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
seconds |
float |
The number of seconds to convert. |
required |
Returns:
| Type | Description |
|---|---|
str |
A human-readable string. |
Source code in zenml/utils/string_utils.py
def get_human_readable_time(seconds: float) -> str:
"""Convert seconds into a human-readable string.
Args:
seconds: The number of seconds to convert.
Returns:
A human-readable string.
"""
prefix = "-" if seconds < 0 else ""
seconds = abs(seconds)
int_seconds = int(seconds)
days, int_seconds = divmod(int_seconds, 86400)
hours, int_seconds = divmod(int_seconds, 3600)
minutes, int_seconds = divmod(int_seconds, 60)
if days > 0:
time_string = f"{days}d{hours}h{minutes}m{int_seconds}s"
elif hours > 0:
time_string = f"{hours}h{minutes}m{int_seconds}s"
elif minutes > 0:
time_string = f"{minutes}m{int_seconds}s"
else:
time_string = f"{seconds:.3f}s"
return prefix + time_string
random_str(length)
Generate a random human readable string of given length.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
length |
int |
Length of string |
required |
Returns:
| Type | Description |
|---|---|
str |
Random human-readable string. |
Source code in zenml/utils/string_utils.py
def random_str(length: int) -> str:
"""Generate a random human readable string of given length.
Args:
length: Length of string
Returns:
Random human-readable string.
"""
random.seed()
return "".join(random.choices(string.ascii_letters, k=length))
terraform_utils
Terraform utilities.
verify_terraform_installation()
Verifies the Terraform installation.
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If Terraform is not installed or ZenML was installed
without the |
Source code in zenml/utils/terraform_utils.py
def verify_terraform_installation() -> None:
"""Verifies the Terraform installation.
Raises:
RuntimeError: If Terraform is not installed or ZenML was installed
without the `terraform` extra.
"""
try:
import python_terraform
except ImportError:
raise RuntimeError(
"Missing Terraform python library. This is probably because ZenML "
"was installed without the optional `terraform` extra. To install "
"the missing dependencies, run: \n\n"
f'`pip install "zenml[terraform]=={zenml.__version__}"`.'
)
try:
python_terraform.Terraform().cmd("-version")
except FileNotFoundError:
raise RuntimeError(
"Missing Terraform installation. Check out "
"https://developer.hashicorp.com/terraform/downloads for how to "
"install it on your operating system."
)
typed_model
Utility classes for adding type information to Pydantic models.
BaseTypedModel (BaseModel)
pydantic-model
Typed Pydantic model base class.
Use this class as a base class instead of BaseModel to automatically
add a type literal attribute to the model that stores the name of the
class.
This can be useful when serializing models to JSON and then de-serializing them as part of a submodel union field, e.g.:
class BluePill(BaseTypedModel):
...
class RedPill(BaseTypedModel):
...
class TheMatrix(BaseTypedModel):
choice: Union[BluePill, RedPill] = Field(..., discriminator='type')
matrix = TheMatrix(choice=RedPill())
d = matrix.dict()
new_matrix = TheMatrix.parse_obj(d)
assert isinstance(new_matrix.choice, RedPill)
It can also facilitate de-serializing objects when their type isn't known:
matrix = TheMatrix(choice=RedPill())
d = matrix.dict()
new_matrix = BaseTypedModel.from_dict(d)
assert isinstance(new_matrix.choice, RedPill)
Source code in zenml/utils/typed_model.py
class BaseTypedModel(BaseModel, metaclass=BaseTypedModelMeta):
"""Typed Pydantic model base class.
Use this class as a base class instead of BaseModel to automatically
add a `type` literal attribute to the model that stores the name of the
class.
This can be useful when serializing models to JSON and then de-serializing
them as part of a submodel union field, e.g.:
```python
class BluePill(BaseTypedModel):
...
class RedPill(BaseTypedModel):
...
class TheMatrix(BaseTypedModel):
choice: Union[BluePill, RedPill] = Field(..., discriminator='type')
matrix = TheMatrix(choice=RedPill())
d = matrix.dict()
new_matrix = TheMatrix.parse_obj(d)
assert isinstance(new_matrix.choice, RedPill)
```
It can also facilitate de-serializing objects when their type isn't known:
```python
matrix = TheMatrix(choice=RedPill())
d = matrix.dict()
new_matrix = BaseTypedModel.from_dict(d)
assert isinstance(new_matrix.choice, RedPill)
```
"""
@classmethod
def from_dict(
cls,
model_dict: Dict[str, Any],
) -> "BaseTypedModel":
"""Instantiate a Pydantic model from a serialized JSON-able dict representation.
Args:
model_dict: the model attributes serialized as JSON-able dict.
Returns:
A BaseTypedModel created from the serialized representation.
Raises:
RuntimeError: if the model_dict contains an invalid type.
"""
model_type = model_dict.get("type")
if not model_type:
raise RuntimeError(
"`type` information is missing from the serialized model dict."
)
cls = source_utils.load(model_type)
if not issubclass(cls, BaseTypedModel):
raise RuntimeError(
f"Class `{cls}` is not a ZenML BaseTypedModel subclass."
)
return cls.parse_obj(model_dict)
@classmethod
def from_json(
cls,
json_str: str,
) -> "BaseTypedModel":
"""Instantiate a Pydantic model from a serialized JSON representation.
Args:
json_str: the model attributes serialized as JSON.
Returns:
A BaseTypedModel created from the serialized representation.
"""
model_dict = json.loads(json_str)
return cls.from_dict(model_dict)
from_dict(model_dict)
classmethod
Instantiate a Pydantic model from a serialized JSON-able dict representation.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_dict |
Dict[str, Any] |
the model attributes serialized as JSON-able dict. |
required |
Returns:
| Type | Description |
|---|---|
BaseTypedModel |
A BaseTypedModel created from the serialized representation. |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
if the model_dict contains an invalid type. |
Source code in zenml/utils/typed_model.py
@classmethod
def from_dict(
cls,
model_dict: Dict[str, Any],
) -> "BaseTypedModel":
"""Instantiate a Pydantic model from a serialized JSON-able dict representation.
Args:
model_dict: the model attributes serialized as JSON-able dict.
Returns:
A BaseTypedModel created from the serialized representation.
Raises:
RuntimeError: if the model_dict contains an invalid type.
"""
model_type = model_dict.get("type")
if not model_type:
raise RuntimeError(
"`type` information is missing from the serialized model dict."
)
cls = source_utils.load(model_type)
if not issubclass(cls, BaseTypedModel):
raise RuntimeError(
f"Class `{cls}` is not a ZenML BaseTypedModel subclass."
)
return cls.parse_obj(model_dict)
from_json(json_str)
classmethod
Instantiate a Pydantic model from a serialized JSON representation.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
json_str |
str |
the model attributes serialized as JSON. |
required |
Returns:
| Type | Description |
|---|---|
BaseTypedModel |
A BaseTypedModel created from the serialized representation. |
Source code in zenml/utils/typed_model.py
@classmethod
def from_json(
cls,
json_str: str,
) -> "BaseTypedModel":
"""Instantiate a Pydantic model from a serialized JSON representation.
Args:
json_str: the model attributes serialized as JSON.
Returns:
A BaseTypedModel created from the serialized representation.
"""
model_dict = json.loads(json_str)
return cls.from_dict(model_dict)
BaseTypedModelMeta (ModelMetaclass)
Metaclass responsible for adding type information to Pydantic models.
Source code in zenml/utils/typed_model.py
class BaseTypedModelMeta(ModelMetaclass):
"""Metaclass responsible for adding type information to Pydantic models."""
def __new__(
mcs, name: str, bases: Tuple[Type[Any], ...], dct: Dict[str, Any]
) -> "BaseTypedModelMeta":
"""Creates a Pydantic BaseModel class.
This includes a hidden attribute that reflects the full class
identifier.
Args:
name: The name of the class.
bases: The base classes of the class.
dct: The class dictionary.
Returns:
A Pydantic BaseModel class that includes a hidden attribute that
reflects the full class identifier.
Raises:
TypeError: If the class is not a Pydantic BaseModel class.
"""
if "type" in dct:
raise TypeError(
"`type` is a reserved attribute name for BaseTypedModel "
"subclasses"
)
type_name = f"{dct['__module__']}.{dct['__qualname__']}"
type_ann = Literal[type_name] # type: ignore[valid-type]
type = Field(type_name)
dct.setdefault("__annotations__", dict())["type"] = type_ann
dct["type"] = type
cls = cast(
Type["BaseTypedModel"], super().__new__(mcs, name, bases, dct)
)
return cls
__new__(mcs, name, bases, dct)
special
staticmethod
Creates a Pydantic BaseModel class.
This includes a hidden attribute that reflects the full class identifier.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
name |
str |
The name of the class. |
required |
bases |
Tuple[Type[Any], ...] |
The base classes of the class. |
required |
dct |
Dict[str, Any] |
The class dictionary. |
required |
Returns:
| Type | Description |
|---|---|
BaseTypedModelMeta |
A Pydantic BaseModel class that includes a hidden attribute that reflects the full class identifier. |
Exceptions:
| Type | Description |
|---|---|
TypeError |
If the class is not a Pydantic BaseModel class. |
Source code in zenml/utils/typed_model.py
def __new__(
mcs, name: str, bases: Tuple[Type[Any], ...], dct: Dict[str, Any]
) -> "BaseTypedModelMeta":
"""Creates a Pydantic BaseModel class.
This includes a hidden attribute that reflects the full class
identifier.
Args:
name: The name of the class.
bases: The base classes of the class.
dct: The class dictionary.
Returns:
A Pydantic BaseModel class that includes a hidden attribute that
reflects the full class identifier.
Raises:
TypeError: If the class is not a Pydantic BaseModel class.
"""
if "type" in dct:
raise TypeError(
"`type` is a reserved attribute name for BaseTypedModel "
"subclasses"
)
type_name = f"{dct['__module__']}.{dct['__qualname__']}"
type_ann = Literal[type_name] # type: ignore[valid-type]
type = Field(type_name)
dct.setdefault("__annotations__", dict())["type"] = type_ann
dct["type"] = type
cls = cast(
Type["BaseTypedModel"], super().__new__(mcs, name, bases, dct)
)
return cls
uuid_utils
Utility functions for handling UUIDs.
generate_uuid_from_string(value)
Deterministically generates a UUID from a string seed.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
value |
str |
The string from which to generate the UUID. |
required |
Returns:
| Type | Description |
|---|---|
UUID |
The generated UUID. |
Source code in zenml/utils/uuid_utils.py
def generate_uuid_from_string(value: str) -> UUID:
"""Deterministically generates a UUID from a string seed.
Args:
value: The string from which to generate the UUID.
Returns:
The generated UUID.
"""
hash_ = hashlib.md5() # nosec
hash_.update(value.encode("utf-8"))
return UUID(hex=hash_.hexdigest(), version=4)
is_valid_uuid(value, version=4)
Checks if a string is a valid UUID.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
value |
Any |
String to check. |
required |
version |
int |
Version of UUID to check for. |
4 |
Returns:
| Type | Description |
|---|---|
bool |
True if string is a valid UUID, False otherwise. |
Source code in zenml/utils/uuid_utils.py
def is_valid_uuid(value: Any, version: int = 4) -> bool:
"""Checks if a string is a valid UUID.
Args:
value: String to check.
version: Version of UUID to check for.
Returns:
True if string is a valid UUID, False otherwise.
"""
if isinstance(value, UUID):
return True
if isinstance(value, str):
try:
UUID(value, version=version)
return True
except ValueError:
return False
return False
parse_name_or_uuid(name_or_id)
Convert a "name or id" string value to a string or UUID.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
name_or_id |
Optional[str] |
Name or id to convert. |
required |
Returns:
| Type | Description |
|---|---|
Union[str, uuid.UUID] |
A UUID if name_or_id is a UUID, string otherwise. |
Source code in zenml/utils/uuid_utils.py
def parse_name_or_uuid(
name_or_id: Optional[str],
) -> Optional[Union[str, UUID]]:
"""Convert a "name or id" string value to a string or UUID.
Args:
name_or_id: Name or id to convert.
Returns:
A UUID if name_or_id is a UUID, string otherwise.
"""
if name_or_id:
try:
return UUID(name_or_id)
except ValueError:
return name_or_id
else:
return name_or_id
visualization_utils
Utility functions for dashboard visualizations.
format_csv_visualization_as_html(csv_visualization, max_rows=10, max_cols=10)
Formats a CSV visualization as an HTML table.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
csv_visualization |
str |
CSV visualization as a string. |
required |
max_rows |
int |
Maximum number of rows to display. Remaining rows will be replaced by an ellipsis in the middle of the table. |
10 |
max_cols |
int |
Maximum number of columns to display. Remaining columns will be replaced by an ellipsis at the end of each row. |
10 |
Returns:
| Type | Description |
|---|---|
str |
HTML table as a string. |
Source code in zenml/utils/visualization_utils.py
def format_csv_visualization_as_html(
csv_visualization: str, max_rows: int = 10, max_cols: int = 10
) -> str:
"""Formats a CSV visualization as an HTML table.
Args:
csv_visualization: CSV visualization as a string.
max_rows: Maximum number of rows to display. Remaining rows will be
replaced by an ellipsis in the middle of the table.
max_cols: Maximum number of columns to display. Remaining columns will
be replaced by an ellipsis at the end of each row.
Returns:
HTML table as a string.
"""
rows = csv_visualization.splitlines()
html = ""
# If there are fewer rows than the maximum, print all rows
if len(rows) <= max_rows:
for row in rows:
html += _format_csv_row_as_html(row, max_cols=max_cols)
else:
# Else, replace middle rows with ellipsis
half_max_rows = max_rows // 2
# Print first half of rows
for row in rows[:half_max_rows]:
html += _format_csv_row_as_html(row, max_cols=max_cols)
# Print ellipsis
if len(rows) > max_rows:
html += "<tr><td>...</td></tr>"
# Print last half of rows
for row in rows[-half_max_rows:]:
html += _format_csv_row_as_html(row, max_cols=max_cols)
return "<table>" + html + "</table>"
visualize_artifact(artifact, title=None)
Visualize an artifact in notebook environments.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
artifact |
ArtifactResponseModel |
The artifact to visualize. |
required |
title |
Optional[str] |
Optional title to show before the visualizations. |
None |
Exceptions:
| Type | Description |
|---|---|
RuntimeError |
If not in a notebook environment. |
Source code in zenml/utils/visualization_utils.py
def visualize_artifact(
artifact: ArtifactResponseModel, title: Optional[str] = None
) -> None:
"""Visualize an artifact in notebook environments.
Args:
artifact: The artifact to visualize.
title: Optional title to show before the visualizations.
Raises:
RuntimeError: If not in a notebook environment.
"""
if not Environment.in_notebook():
raise RuntimeError(
"The `output.visualize()` method is only available in Jupyter "
"notebooks. In all other runtime environments, please open "
"your ZenML dashboard using `zenml up` and view the "
"visualizations by clicking on the respective artifacts in the "
"pipeline run DAG instead."
)
if not artifact.visualizations:
return
if title:
display(Markdown(f"### {title}"))
for i in range(len(artifact.visualizations)):
visualization = load_artifact_visualization(artifact, index=i)
if visualization.type == VisualizationType.IMAGE:
display(Image(visualization.value))
elif visualization.type == VisualizationType.HTML:
display(HTML(visualization.value))
elif visualization.type == VisualizationType.MARKDOWN:
display(Markdown(visualization.value))
elif visualization.type == VisualizationType.CSV:
assert isinstance(visualization.value, str)
table = format_csv_visualization_as_html(visualization.value)
display(HTML(table))
else:
display(visualization.value)
yaml_utils
Utility functions to help with YAML files and data.
UUIDEncoder (JSONEncoder)
JSON encoder for UUID objects.
Source code in zenml/utils/yaml_utils.py
class UUIDEncoder(json.JSONEncoder):
"""JSON encoder for UUID objects."""
def default(self, obj: Any) -> Any:
"""Default UUID encoder for JSON.
Args:
obj: Object to encode.
Returns:
Encoded object.
"""
if isinstance(obj, UUID):
# if the obj is uuid, we simply return the value of uuid
return obj.hex
return json.JSONEncoder.default(self, obj)
default(self, obj)
Default UUID encoder for JSON.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
obj |
Any |
Object to encode. |
required |
Returns:
| Type | Description |
|---|---|
Any |
Encoded object. |
Source code in zenml/utils/yaml_utils.py
def default(self, obj: Any) -> Any:
"""Default UUID encoder for JSON.
Args:
obj: Object to encode.
Returns:
Encoded object.
"""
if isinstance(obj, UUID):
# if the obj is uuid, we simply return the value of uuid
return obj.hex
return json.JSONEncoder.default(self, obj)
append_yaml(file_path, contents)
Append contents to a YAML file at file_path.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to YAML file. |
required |
contents |
Dict[Any, Any] |
Contents of YAML file as dict. |
required |
Exceptions:
| Type | Description |
|---|---|
FileNotFoundError |
if directory does not exist. |
Source code in zenml/utils/yaml_utils.py
def append_yaml(file_path: str, contents: Dict[Any, Any]) -> None:
"""Append contents to a YAML file at file_path.
Args:
file_path: Path to YAML file.
contents: Contents of YAML file as dict.
Raises:
FileNotFoundError: if directory does not exist.
"""
file_contents = read_yaml(file_path) or {}
file_contents.update(contents)
if not io_utils.is_remote(file_path):
dir_ = str(Path(file_path).parent)
if not fileio.isdir(dir_):
raise FileNotFoundError(f"Directory {dir_} does not exist.")
io_utils.write_file_contents_as_string(file_path, yaml.dump(file_contents))
comment_out_yaml(yaml_string)
Comments out a yaml string.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
yaml_string |
str |
The yaml string to comment out. |
required |
Returns:
| Type | Description |
|---|---|
str |
The commented out yaml string. |
Source code in zenml/utils/yaml_utils.py
def comment_out_yaml(yaml_string: str) -> str:
"""Comments out a yaml string.
Args:
yaml_string: The yaml string to comment out.
Returns:
The commented out yaml string.
"""
lines = yaml_string.splitlines(keepends=True)
lines = ["# " + line for line in lines]
return "".join(lines)
is_json_serializable(obj)
Checks whether an object is JSON serializable.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
obj |
Any |
The object to check. |
required |
Returns:
| Type | Description |
|---|---|
bool |
Whether the object is JSON serializable using pydantics encoder class. |
Source code in zenml/utils/yaml_utils.py
def is_json_serializable(obj: Any) -> bool:
"""Checks whether an object is JSON serializable.
Args:
obj: The object to check.
Returns:
Whether the object is JSON serializable using pydantics encoder class.
"""
from pydantic.json import pydantic_encoder
try:
json.dumps(obj, default=pydantic_encoder)
return True
except TypeError:
return False
is_yaml(file_path)
Returns True if file_path is YAML, else False.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to YAML file. |
required |
Returns:
| Type | Description |
|---|---|
bool |
True if is yaml, else False. |
Source code in zenml/utils/yaml_utils.py
def is_yaml(file_path: str) -> bool:
"""Returns True if file_path is YAML, else False.
Args:
file_path: Path to YAML file.
Returns:
True if is yaml, else False.
"""
if file_path.endswith("yaml") or file_path.endswith("yml"):
return True
return False
read_json(file_path)
Read JSON on file path and returns contents as dict.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to JSON file. |
required |
Returns:
| Type | Description |
|---|---|
Any |
Contents of the file in a dict. |
Exceptions:
| Type | Description |
|---|---|
FileNotFoundError |
if file does not exist. |
Source code in zenml/utils/yaml_utils.py
def read_json(file_path: str) -> Any:
"""Read JSON on file path and returns contents as dict.
Args:
file_path: Path to JSON file.
Returns:
Contents of the file in a dict.
Raises:
FileNotFoundError: if file does not exist.
"""
if fileio.exists(file_path):
contents = io_utils.read_file_contents_as_string(file_path)
return json.loads(contents)
else:
raise FileNotFoundError(f"{file_path} does not exist.")
read_yaml(file_path)
Read YAML on file path and returns contents as dict.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to YAML file. |
required |
Returns:
| Type | Description |
|---|---|
Any |
Contents of the file in a dict. |
Exceptions:
| Type | Description |
|---|---|
FileNotFoundError |
if file does not exist. |
Source code in zenml/utils/yaml_utils.py
def read_yaml(file_path: str) -> Any:
"""Read YAML on file path and returns contents as dict.
Args:
file_path: Path to YAML file.
Returns:
Contents of the file in a dict.
Raises:
FileNotFoundError: if file does not exist.
"""
if fileio.exists(file_path):
contents = io_utils.read_file_contents_as_string(file_path)
# TODO: [LOW] consider adding a default empty dict to be returned
# instead of None
return yaml.safe_load(contents)
else:
raise FileNotFoundError(f"{file_path} does not exist.")
write_json(file_path, contents, encoder=None)
Write contents as JSON format to file_path.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to JSON file. |
required |
contents |
Any |
Contents of JSON file. |
required |
encoder |
Optional[Type[json.encoder.JSONEncoder]] |
Custom JSON encoder to use when saving json. |
None |
Exceptions:
| Type | Description |
|---|---|
FileNotFoundError |
if directory does not exist. |
Source code in zenml/utils/yaml_utils.py
def write_json(
file_path: str,
contents: Any,
encoder: Optional[Type[json.JSONEncoder]] = None,
) -> None:
"""Write contents as JSON format to file_path.
Args:
file_path: Path to JSON file.
contents: Contents of JSON file.
encoder: Custom JSON encoder to use when saving json.
Raises:
FileNotFoundError: if directory does not exist.
"""
if not io_utils.is_remote(file_path):
dir_ = str(Path(file_path).parent)
if not fileio.isdir(dir_):
# Check if it is a local path, if it doesn't exist, raise Exception.
raise FileNotFoundError(f"Directory {dir_} does not exist.")
io_utils.write_file_contents_as_string(
file_path,
json.dumps(
contents,
cls=encoder,
),
)
write_yaml(file_path, contents, sort_keys=True)
Write contents as YAML format to file_path.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
file_path |
str |
Path to YAML file. |
required |
contents |
Union[Dict[Any, Any], List[Any]] |
Contents of YAML file as dict or list. |
required |
sort_keys |
bool |
If |
True |
Exceptions:
| Type | Description |
|---|---|
FileNotFoundError |
if directory does not exist. |
Source code in zenml/utils/yaml_utils.py
def write_yaml(
file_path: str,
contents: Union[Dict[Any, Any], List[Any]],
sort_keys: bool = True,
) -> None:
"""Write contents as YAML format to file_path.
Args:
file_path: Path to YAML file.
contents: Contents of YAML file as dict or list.
sort_keys: If `True`, keys are sorted alphabetically. If `False`,
the order in which the keys were inserted into the dict will
be preserved.
Raises:
FileNotFoundError: if directory does not exist.
"""
if not io_utils.is_remote(file_path):
dir_ = str(Path(file_path).parent)
if not fileio.isdir(dir_):
raise FileNotFoundError(f"Directory {dir_} does not exist.")
io_utils.write_file_contents_as_string(
file_path, yaml.dump(contents, sort_keys=sort_keys)
)