This repository contains extensions for Metaflow that are in use at Netflix (or being tested at Netflix) and that are more cutting edge than what is included in the OSS Metaflow package.
You can find support for this extension on the usual Metaflow Slack.
NOTE: of you are within Netflix and are looking for the Netflix version of Metaflow, this is not it (this only contains a part of the Netflix internal extensions).
Netflix released Metaflow as OSS in 2019. Since then, development of Metaflow internally to Netflix has continued primarily around extensions to better support Netflix's infrastructure and provide a more seamless integration with the compute and orchestration platforms specific to Netflix. Netflix continues to collaboratively improve Metaflow's OSS capabilities in collaboration with OuterBounds and, as such, sometimes develops functionality that is not yet fully ready for inclusion in the community supported Metaflow as interest in the functionality may not be clear or there is not time in the community to properly integrate and fully test the functionality.
This repository will contain such functionality. While we do our best to ensure that the functionality present works, it does not have the same levels of support and backward compatibility guarantees that Metaflow does. Functionality present in this package is likely to end up in the main Metaflow package with, potentially, some modification (in which case it will be removed from this package) but that is not a guarantee. If you find this functionality useful and would like to see it make it to the main Metaflow package, let us know. Feedback is always welcome!
This extension is currently tested on python 3.7+.
If you have any question, feel free to open an issue here or contact us on the usual Metaflow slack channels.
This extension currently contains:
Version 1.0.0 is considered stable. Some UX changes have occurred compared to previous versions. Please see the docs for more information
Version 0.2.0 of this extension is not fully backward compatible with previous versions due to
where packages are cached. If you are using a previous version of the extension, it is recommended
that you change the CONDA_MAGIC_FILE_V2, CONDA_PACKAGES_DIRNAME and CONDA_ENVS_DIRNAME to
new values to be able to have both versions active at the same time.
It is likely to evolve primarily in its implementation as we do further testing. Feedback on what is working and what is not is most welcome.
This decorator improves several aspects of the included Conda decorator:
requirements.txt of environment.yml filesTo use, simply install this package alongside the metaflow package. This package
requires Metaflow v2.8.3 or later.
You have several configuration options that can be set in
metaflow_extensions/netflix_ext/config/mfextinit_netflixext.py. Due to limitations in
the OSS implementation of decorators such as batch and kubernetes, prior to Metaflow v2.10,
you should set these values directly in the mfextinit_netflixext.py configuration file and
not in an external configuration or through environment variables. This limitation is
removed in Metaflow v2.10.
The useful configuration values are listed below:
CONDA_S3ROOT/CONDA_AZUREROOT/CONDA_GSROOT: directory in S3/azure/gs containing
all the cached packages and environments
as well as eventual conda distributions to use. For safety, do not point this to the
same prefix as for the current Conda implementation.CONDA_DEPENDENCY_RESOLVER: mamba, conda or micromamba; mamba is recommended as
typically faster. micromamba is sometimes a bit more unstable but can be even fasterCONDA_PYPI_DEPENDENCY_RESOLVER: pip or None; if None, you will not be able to resolve
environments specifying only pypi dependencies.CONDA_MIXED_DEPENDENCY_RESOLVER: conda-lock or none; if none, you will not be able
to resolve environments specifying a mix of pypi and conda dependencies.CONDA_REMOTE_INSTALLER_DIRNAME: if set contains a prefix within
CONDA_S3ROOT/CONDA_AZUREROOT/CONDA_GSROOT
under which micromamba (or other similar executable) are cached. If not specified,
micromamba's latest version will be downloaded on remote environments when an
environment needs to be re-hydrated.CONDA_REMOTE_INSTALLER: if set architecture specific installer in
CONDA_REMOTE_INSTALLER_DIRNAME.CONDA_LOCAL_DIST_DIRNAME: if set contains a prefix within
CONDA_S3ROOT/CONDA_AZUREROOT/CONDA_GSROOT under
which fully created conda environments for local execution are cached. If not set,
the local machine's Conda installation is used.CONDA_PACKAGES_DIRNAME: directory within CONDA_S3ROOT/CONDA_AZUREROOT/CONDA_GSROOT
under which cached packages are stored (defaults to packages)CONDA_ENVS_DIRNAME: same thing a CONDA_PACKAGES_DIRNAME but for environments
(defaults to envs)CONDA_LOCAL_DIST: if set architecture specific tar ball in CONDA_LOCAL_DIST_DIRNAME.CONDA_LOCAL_PATH: if set, installs the tarball in CONDA_LOCAL_DIST in this path.CONDA_PREFERRED_FORMAT: .tar.bz2 or .conda or none (default).
Prefer .conda for speed gains; any
package not available in the preferred format will be transmuted to it automatically.
If left empty, whatever package is found will be used (ie: there is no preference)CONDA_DEFAULT_PYPI_SOURCE: mirror to use for PYPI.CONDA_USE_REMOTE_LATEST: by default, it is set to :none: which means that if a new
environment is not locally known (for example first time resolving it on the machine), it
will be re-resolved. You can also set it to :username:, :any: or a comma separated
list of usernames to tell Metaflow to go check if there is a cached environment that matches
the requested specification that has been resolved previously by either the current user,
any user or the set of users.For Azure, you need to do the following two steps once during setup:
CONDA_AZUREROOTStorage Blob Data Contributor role to the storage account to the service
principal or user accounts that will be accessing as described
here.Your local conda environment or the cached environment (in CONDA_LOCAL_DIST_DIRNAME)
needs to satisfy the following requirements:
condamamba>=1.4.0micromamba>=1.4.0If you want support for environments containing only pip packages, you will also need:
pip>=23.0If you want support for environments containing both pip and conda packages, you will also need:
conda-lock>=2.1.0.tar.bz2 and .conda packagesIf you set CONDA_PREFERRED_FORMAT to either .tar.bz2 or .conda, for some packages,
we will need to transmute them from one format to the other. For example if a package
is available for download as a .tar.bz2 package but you request .conda packages,
the system will transmute (convert) the .tar.bz2 package into one that ends in
.conda. To do so, you need to have one of the following package installed:
conda-package-handling>=1.9.0micromamba>=1.4.0 (not supported for cross-platform transmutation due to
https://github.com/mamba-org/mamba/issues/2328 or if you are transmuting to
.tar.bz2 files).Also due to a bug in conda and the way we use it, if your resolved environment
contains .conda packages and you do not have micromamba installed, the
environment creation will fail.
This plugin relies on conda, mamba, and micromamba. These technologies are being constantly improved and there are a few outstanding issues that we are aware of:
.conda and .tar.bz2 packages, conda/mamba
will fail to create it because we use it in "offline" mode
(see: https://github.com/conda/conda/issues/11775). The workaround is to have
micromamba available which does not have this issue and which Metaflow will use
if it is presentmicromamba is not supported for cross-platform
transmutes due to https://github.com/mamba-org/mamba/issues/2328. It also does
not work properly when transmuting from .conda packages to .tar.bz2 packages.
Install conda-package-handling as well to support this.Uninstalling this package will revert the behavior of the conda decorator to the one currently present in Metaflow. It is safe to switch back and forth and there should be no conflict between both implementations provided they do not share the same caching prefix in S3/azure/gs and that you do not use any of the new features.
Your current code with conda decorators will continue working as is. However, at this
time, there is no method to "convert" previously resolved environment to this new
implementation so the first time you run Metaflow with this package, your previously
resolved environments will be ignored and re-resolved.
Environments listed below are examples that can be resolved using Metaflow. The environments
given here are either in the requirements.txt format or environment.yml format and can,
for example, be passed to metaflow environment resolve using the -r or -f option
respectively. They highlight some of the functionalities present. Note that the same
environments can also be specified directly using the @conda or @pip decorators.
--conda-pkg ffmpeg
ffmpeg-python
The requirements.txt file above will create an environment with the Pip package
ffmpeg-python as well as the ffmpeg Conda executable. This is useful to have
a pure pip environment (and therefore use the underlying pip ecosystem without
conda-lock but still have other non Python packages installed.
--conda-pkg git-lfs
# Needs LFS to build
transnetv2 @ git+https://github.com/soCzech/TransNetV2.git#main
# GIT repo
clip @ git+https://github.com/openai/CLIP.git@d50d76daa670286dd6cacf3bcd80b5e4823fc8e1
# Source only distribution
outlier-detector==0.0.3
# Local package
foo @ file:///tmp/build_foo_pkg
The above requirements.txt shows that it is possible to specify repositories directly.
Note that this does not work cross platform. Behind the scenes, Metaflow will build wheel
packages and cache them.
dependencies:
- pandas = >=1.0.0
- pip:
- tensorflow = 2.7.4
- apache-airflow[aiobotocore]
The above environment.yml shows that it is possible to mix and match pip and conda
packages. You can specify packages using "extras" but you cannot, in this form,
specify pip packages that come from git repositories or from your local file-system.
Pypi packages that are available as wheels or source tar balls are acceptable.
In general, the following restrictions are applicable:
osx-arm64 from a linux-64 machine). This
restriction will not be removed as this would potentially require cross-platform build which
can be tricky and error-prone.For additional documentation, please refer to the documentation which contains more detailed documentation.
This section dives a bit more in the technical aspects of this implementation.
An environment can either be un-resolved or resolved. An un-resolved environment is simply defined by the set of high-level user-requirements that the environment must satisfy. Typically, this is a list of Conda and/or Pypi packages and version constraints on them. In our case, we also include the set of channels (Conda) or sources (Pip). A resolved environment contains the concrete list of packages that are to be installed to meet the aforementioned requirements. In a resolved environment, all packages are pinned to a single unique version.
In Metaflow, two hashes identify environments and EnvID (from env_descr.py)
encapsulates these hashes:
req_id. This hash encapsulates the packages and version constraints as well
as the channels or sources. The packages are sorted to provide a stable hash for
identical set of requirements.full_id.We also associate the architecture for which the environment was resolved to form the
complete EnvID.
Environments are named as metaflow_<req_id>_<full_id>. Note that environments that
are resolved versions of the same un-resolved environment therefore have the same
prefix.
This implementation of Conda clearly separates out the phases needed to execute a Metaflow task in a Conda environment:
resolving the environment: this is the step needed to go from an un-resolved environment to a fully resolved one. It does not require the downloading of packages (for the most part) nor the creation of an environment.
caching the environment: this is an optional step which stores all the packages as well as the description of the environment in S3/azure/gs for later retrieval on environment creation. During this step, packages may be downloaded (from the web for example) but an environment is still not created.
creating the environment: in this step, the exact set of packages needed are downloaded (if needed) and an environment is created from there. At this point, there is no resolution (we know the exact set of packages needed).
env_descr.py contains a simple way to encode all the information needed for all the
above steps, specifically it contains a set of ResolvedEnvironment which, in turn,
contain the ID for the environment and information about each package. Each package,
in turn, contains information about where it can be located on the web as well
as caching information (where it is located in the cache). Each package can also
support multiple formats (Conda uses either .tar.bz2 or .conda -- note that this
is meant to support equivalent formats and not .whl versus .tar.gz for Pypi
packages for example).
Very little effort is made to remove duplicate information (packages may for example
be present in several resolved environments) as modularity is favored (ie: each
ResolvedEnvironment is fully self contained).
The conda_flow_decorator.py and conda_step_decorator.py files simply contain
trivial logic to convert the specification passed to those decorators (effectively
information needed to construct the requirement ID of the environment) to something
that is understandable by the rest of the system. In effect, they are mostly
transformers that take user-information and convert it to the set of packages the
user wants to have present in their environment.
The conda_environment.py file contains methods to effectively:
conda_step_decorator.py that has to do with starting a task locally).The actual work is all handled in the conda.py file which contains the crux of the
logic.
All environments are resolved in parallel and independently. To do so, we either use
conda-lock or mamba/conda using the --dry-run option. The processing
for this takes place in resolve_environment in the conda.py file.
The input to this step is a set of user-level requirements and the output is a set
of ResolvedEnvironment. At this point, no package has been downloaded and the
ResolvedEnvironment is most likely missing any information about caching.
The cache_environments method in the conda.py file implements this.
There are several steps here. We perform these steps for all resolved environments that need their cache information updated at once to be able to exploit the fact that several environments may refer to the same package:
lazy_download_packages method. We do this per architecture.
The basic concept of this function is to locate the "nearest" source of the package.
In order, we look for:
The ResolvedEnvironment, now with updated cache information, is also cached to S3/azure/gs to
promote sharing.
This is the easiest step of all and simply consists of fetching all packages (again
using the lazy_download_packages method which will not download any package that
is already present) and then using micromamba (or mamba/conda) to simply install
all packages.
There are two main things that are cached:
ResolvedEnvironment in JSON format)There are also two levels of caching:
conda_v2.cnd which
caches all environments already resolved. This allows us to reuse the same
environment for similar user-level requirements (which is typically what the user
wants).pkgs directory of the Conda installation.
They may be either fully expanded directories or archives.conda_v2.cnd file.pkgs directory. The
implementation takes care of properly updating the urls.txt file to make it
transparent to Conda (allowing it to operate in an "offline" mode effectively).This extension allows user's to seamlessly debug their executed steps in an isolated Jupyter notebook instance with appropriate dependencies by leveraging the conda extension described above (note, this extension currently only works with the version of Conda in this package).
Let's say you have a step called fit_gbrt_for_given_param in your flow, and on executing it, the pathspec for
this step/task is HousePricePredictionFlow/1199/fit_gbrt_for_given_param/150671013. To debug this step, you can run the command:
metaflow debug task <HousePricePredictionFlow/1199/fit_gbrt_for_given_param/150671013> --metaflow-root-dir ~/notebooks/debug_task`
Note that you can specify a partial pathspec as long as it can be resolved to a unique task:
Running the above command will:
It will additionally generate a notebook in the defined directory where you can debug the execution of your step line by line. For the given step definition:
@step
def fit_gbrt_for_given_param(self):
"""
Fit GBRT with given parameters
"""
from sklearn import ensemble
from sklearn.model_selection import cross_val_score
import numpy as np
estimator = ensemble.GradientBoostingRegressor( n_estimators = self.input['n_estimators'], learning_rate = self.input['learning_rate'],
max_depth = self.input['max_depth'], min_samples_split = 2, loss = 'ls')
estimator.fit(self.features, self.labels)
mses = cross_val_score(estimator, self.features, self.labels, cv = 5, scoring='neg_mean_squared_error')
rmse = np.sqrt(-mses).mean()
self.fit = dict(
index=int(self.index),
params=self.input,
rmse=rmse,
estimator=estimator
)
self.next(self.select_best_model)
You will be able to access the artifacts/inputs in your generated notebook directly:
>>> print(self.input['n_estimators']) # You can access objects using `self` as we imported a stub for it in the notebook
>>> print(self.input['learning_rate'])
You can also execute the whole function again:
>>> from sklearn import ensemble # imports work seamlessly due to conda extension
>>> from sklearn.model_selection import cross_val_score
>>> import numpy as np
>>> estimator = ensemble.GradientBoostingRegressor( n_estimators = self.input['n_estimators'], learning_rate = self.input['learning_rate'],
max_depth = self.input['max_depth'], min_samples_split = 2, loss = 'ls')
>>> estimator.fit(self.features, self.labels)
>>> mses = cross_val_score(estimator, self.features, self.labels, cv = 5, scoring='neg_mean_squared_error')
>>> rmse = np.sqrt(-mses).mean()
>>> self.fit = dict(
index=int(self.index),
params=self.input,
rmse=rmse,
estimator=estimator
)
You can examine the effects of other hyper-parameters live by modifying the min_samples_split = 3 and re-executing the steps on the same data.
FAQs
Metaflow extensions from Netflix
We found that metaflow-netflixext demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
PyPI confirms no security flaws were exploited in the Ultralytics supply chain attack and highlights improvements for safer package publishing.
Security News
Research
The Socket Research Team breaks down a malicious wrapper package that uses obfuscation to harvest credentials and exfiltrate sensitive data.
Research
Security News
Attackers used a malicious npm package typosquatting a popular ESLint plugin to steal sensitive data, execute commands, and exploit developer systems.