"""
The ``mlflow.fastai`` module provides an API for logging and loading fast.ai models. This module
exports fast.ai models with the following flavors:
fastai (native) format
This is the main flavor that can be loaded back into fastai.
:py:mod:`mlflow.pyfunc`
Produced for use by generic pyfunc-based deployment tools and batch inference.
.. _fastai.Learner:
https://docs.fast.ai/basic_train.html#Learner
.. _fastai.Learner.export:
https://docs.fast.ai/basic_train.html#Learner.export
"""
import os
import yaml
import tempfile
import shutil
import pandas as pd
import numpy as np
from mlflow import pyfunc
from mlflow.models import Model, ModelSignature, ModelInputExample
import mlflow.tracking
from mlflow.exceptions import MlflowException
from mlflow.models.utils import _save_example
from mlflow.models.model import MLMODEL_FILE_NAME
from mlflow.tracking.artifact_utils import _download_artifact_from_uri
from mlflow.utils.environment import (
_mlflow_conda_env,
_validate_env_arguments,
_process_pip_requirements,
_process_conda_env,
_CONDA_ENV_FILE_NAME,
_REQUIREMENTS_FILE_NAME,
_CONSTRAINTS_FILE_NAME,
)
from mlflow.utils.requirements_utils import _get_pinned_requirement
from mlflow.utils.file_utils import write_to
from mlflow.utils.docstring_utils import format_docstring, LOG_MODEL_PARAM_DOCS
from mlflow.utils.model_utils import _get_flavor_configuration
from mlflow.utils.annotations import experimental
from mlflow.utils.autologging_utils import (
try_mlflow_log,
log_fn_args_as_params,
safe_patch,
batch_metrics_logger,
autologging_integration,
)
from mlflow.tracking._model_registry import DEFAULT_AWAIT_MAX_SLEEP_SECONDS
FLAVOR_NAME = "fastai"
[docs]def get_default_pip_requirements(include_cloudpickle=False):
"""
:return: A list of default pip requirements for MLflow Models produced by this flavor.
Calls to :func:`save_model()` and :func:`log_model()` produce a pip environment
that, at minimum, contains these requirements.
"""
pip_deps = [_get_pinned_requirement("fastai")]
if include_cloudpickle:
pip_deps.append(_get_pinned_requirement("cloudpickle"))
return pip_deps
[docs]def get_default_conda_env(include_cloudpickle=False): # pylint: disable=unused-argument
"""
:return: The default Conda environment for MLflow Models produced by calls to
:func:`save_model()` and :func:`log_model()`.
"""
return _mlflow_conda_env(additional_pip_deps=get_default_pip_requirements())
[docs]@format_docstring(LOG_MODEL_PARAM_DOCS.format(package_name=FLAVOR_NAME))
def save_model(
fastai_learner,
path,
conda_env=None,
mlflow_model=None,
signature: ModelSignature = None,
input_example: ModelInputExample = None,
pip_requirements=None,
extra_pip_requirements=None,
**kwargs
):
"""
Save a fastai Learner to a path on the local file system.
:param fastai_learner: fastai Learner to be saved.
:param path: Local path where the model is to be saved.
:param conda_env: {{ conda_env }}
:param mlflow_model: MLflow model config this flavor is being added to.
:param signature: :py:class:`ModelSignature <mlflow.models.ModelSignature>`
describes model input and output :py:class:`Schema <mlflow.types.Schema>`.
The model signature can be :py:func:`inferred <mlflow.models.infer_signature>`
from datasets with valid model input (e.g. the training dataset with target
column omitted) and valid model output (e.g. model predictions generated on
the training dataset), for example:
.. code-block:: python
from mlflow.models.signature import infer_signature
train = df.drop_column("target_label")
predictions = ... # compute model predictions
signature = infer_signature(train, predictions)
:param input_example: Input example provides one or several instances of valid
model input. The example can be used as a hint of what data to feed the
model. The given example will be converted to a Pandas DataFrame and then
serialized to json using the Pandas split-oriented format. Bytes are
base64-encoded.
:param pip_requirements: {{ pip_requirements }}
:param extra_pip_requirements: {{ extra_pip_requirements }}
:param kwargs: kwargs to pass to ``Learner.save`` method.
.. code-block:: python
:caption: Example
import os
import mlflow.fastai
# Create a fastai Learner model
model = ...
# Start MLflow session and save model to current working directory
with mlflow.start_run():
model.fit(epochs, learning_rate)
mlflow.fastai.save_model(model, 'model')
# Load saved model for inference
model_uri = "{}/{}".format(os.getcwd(), 'model')
loaded_model = mlflow.fastai.load_model(model_uri)
results = loaded_model.predict(predict_data)
"""
import fastai
from fastai.callback.all import ParamScheduler
from pathlib import Path
_validate_env_arguments(conda_env, pip_requirements, extra_pip_requirements)
path = os.path.abspath(path)
if os.path.exists(path):
raise MlflowException("Path '{}' already exists".format(path))
model_data_subpath = "model.fastai"
model_data_path = os.path.join(path, model_data_subpath)
model_data_path = Path(model_data_path)
os.makedirs(path)
if mlflow_model is None:
mlflow_model = Model()
if signature is not None:
mlflow_model.signature = signature
if input_example is not None:
_save_example(mlflow_model, input_example, path)
# ParamScheduler currently is not pickable
# hence it is been removed before export and added again after export
cbs = [c for c in fastai_learner.cbs if isinstance(c, ParamScheduler)]
fastai_learner.remove_cbs(cbs)
fastai_learner.export(model_data_path, **kwargs)
fastai_learner.add_cbs(cbs)
pyfunc.add_to_model(
mlflow_model,
loader_module="mlflow.fastai",
data=model_data_subpath,
env=_CONDA_ENV_FILE_NAME,
)
mlflow_model.add_flavor(FLAVOR_NAME, fastai_version=fastai.__version__, data=model_data_subpath)
mlflow_model.save(os.path.join(path, MLMODEL_FILE_NAME))
if conda_env is None:
if pip_requirements is None:
default_reqs = get_default_pip_requirements()
# To ensure `_load_pyfunc` can successfully load the model during the dependency
# inference, `mlflow_model.save` must be called beforehand to save an MLmodel file.
inferred_reqs = mlflow.models.infer_pip_requirements(
path, FLAVOR_NAME, fallback=default_reqs,
)
default_reqs = sorted(set(inferred_reqs).union(default_reqs))
else:
default_reqs = None
conda_env, pip_requirements, pip_constraints = _process_pip_requirements(
default_reqs, pip_requirements, extra_pip_requirements,
)
else:
conda_env, pip_requirements, pip_constraints = _process_conda_env(conda_env)
with open(os.path.join(path, _CONDA_ENV_FILE_NAME), "w") as f:
yaml.safe_dump(conda_env, stream=f, default_flow_style=False)
# Save `constraints.txt` if necessary
if pip_constraints:
write_to(os.path.join(path, _CONSTRAINTS_FILE_NAME), "\n".join(pip_constraints))
# Save `requirements.txt`
write_to(os.path.join(path, _REQUIREMENTS_FILE_NAME), "\n".join(pip_requirements))
[docs]@format_docstring(LOG_MODEL_PARAM_DOCS.format(package_name=FLAVOR_NAME))
def log_model(
fastai_learner,
artifact_path,
conda_env=None,
registered_model_name=None,
signature: ModelSignature = None,
input_example: ModelInputExample = None,
await_registration_for=DEFAULT_AWAIT_MAX_SLEEP_SECONDS,
pip_requirements=None,
extra_pip_requirements=None,
**kwargs
):
"""
Log a fastai model as an MLflow artifact for the current run.
:param fastai_learner: Fastai model (an instance of `fastai.Learner`_) to be saved.
:param artifact_path: Run-relative artifact path.
:param conda_env: {{ conda_env }}
:param registered_model_name: This argument may change or be removed in a
future release without warning. If given, create a model
version under ``registered_model_name``, also creating a
registered model if one with the given name does not exist.
:param signature: :py:class:`ModelSignature <mlflow.models.ModelSignature>`
describes model input and output :py:class:`Schema <mlflow.types.Schema>`.
The model signature can be :py:func:`inferred <mlflow.models.infer_signature>`
from datasets with valid model input (e.g. the training dataset with target
column omitted) and valid model output (e.g. model predictions generated on
the training dataset), for example:
.. code-block:: python
from mlflow.models.signature import infer_signature
train = df.drop_column("target_label")
predictions = ... # compute model predictions
signature = infer_signature(train, predictions)
:param input_example: Input example provides one or several instances of valid
model input. The example can be used as a hint of what data to feed the
model. The given example will be converted to a Pandas DataFrame and then
serialized to json using the Pandas split-oriented format. Bytes are
base64-encoded.
:param kwargs: kwargs to pass to `fastai.Learner.export`_ method.
:param await_registration_for: Number of seconds to wait for the model version to finish
being created and is in ``READY`` status. By default, the function
waits for five minutes. Specify 0 or None to skip waiting.
:param pip_requirements: {{ pip_requirements }}
:param extra_pip_requirements: {{ extra_pip_requirements }}
.. code-block:: python
:caption: Example
import fastai.vision as vis
import mlflow.fastai
from mlflow.tracking import MlflowClient
def main(epochs=5, learning_rate=0.01):
# Download and untar the MNIST data set
path = vis.untar_data(vis.URLs.MNIST_SAMPLE)
# Prepare, transform, and normalize the data
data = vis.ImageDataBunch.from_folder(path, ds_tfms=(vis.rand_pad(2, 28), []), bs=64)
data.normalize(vis.imagenet_stats)
# Create the CNN Learner model
model = vis.cnn_learner(data, vis.models.resnet18, metrics=vis.accuracy)
# Start MLflow session and log model
with mlflow.start_run() as run:
model.fit(epochs, learning_rate)
mlflow.fastai.log_model(model, 'model')
# fetch the logged model artifacts
artifacts = [f.path for f in MlflowClient().list_artifacts(run.info.run_id, 'model')]
print("artifacts: {}".format(artifacts))
main()
.. code-block:: text
:caption: Output
artifacts: ['model/MLmodel', 'model/conda.yaml', 'model/model.fastai']
"""
Model.log(
artifact_path=artifact_path,
flavor=mlflow.fastai,
registered_model_name=registered_model_name,
fastai_learner=fastai_learner,
conda_env=conda_env,
signature=signature,
input_example=input_example,
await_registration_for=await_registration_for,
pip_requirements=pip_requirements,
extra_pip_requirements=extra_pip_requirements,
**kwargs,
)
def _load_model(path):
from fastai.learner import load_learner
return load_learner(os.path.abspath(path))
class _FastaiModelWrapper:
def __init__(self, learner):
self.learner = learner
def predict(self, dataframe):
dl = self.learner.dls.test_dl(dataframe)
preds, _ = self.learner.get_preds(dl=dl)
return pd.DataFrame(map(np.array, preds.numpy()), columns=["predictions"])
def _load_pyfunc(path):
"""
Load PyFunc implementation. Called by ``pyfunc.load_pyfunc``.
:param path: Local filesystem path to the MLflow Model with the ``fastai`` flavor.
"""
return _FastaiModelWrapper(_load_model(path))
[docs]def load_model(model_uri):
"""
Load a fastai model from a local file or a run.
:param model_uri: The location, in URI format, of the MLflow model. For example:
- ``/Users/me/path/to/local/model``
- ``relative/path/to/local/model``
- ``s3://my_bucket/path/to/model``
- ``runs:/<mlflow_run_id>/run-relative/path/to/model``
For more information about supported URI schemes, see
`Referencing Artifacts <https://www.mlflow.org/docs/latest/tracking.html#
artifact-locations>`_.
:return: A fastai model (an instance of `fastai.Learner`_).
.. code-block:: python
:caption: Example
import mlflow.fastai
# Define the Learner model
model = ...
# log the fastai Leaner model
with mlflow.start_run() as run:
model.fit(epochs, learning_rate)
mlflow.fastai.log_model(model, "model")
# Load the model for scoring
model_uri = "runs:/{}/model".format(run.info.run_id)
loaded_model = mlflow.fastai.load_model(model_uri)
results = loaded_model.predict(predict_data)
"""
local_model_path = _download_artifact_from_uri(artifact_uri=model_uri)
flavor_conf = _get_flavor_configuration(model_path=local_model_path, flavor_name=FLAVOR_NAME)
model_file_path = os.path.join(local_model_path, flavor_conf.get("data", "model.fastai"))
return _load_model(path=model_file_path)
[docs]@experimental
@autologging_integration(FLAVOR_NAME)
def autolog(
log_models=True,
disable=False,
exclusive=False,
disable_for_unsupported_versions=False,
silent=False,
): # pylint: disable=unused-argument
"""
Enable automatic logging from Fastai to MLflow.
Logs loss and any other metrics specified in the fit
function, and optimizer data as parameters. Model checkpoints
are logged as artifacts to a 'models' directory.
MLflow will also log the parameters of the
`EarlyStoppingCallback <https://docs.fast.ai/callback.tracker.html#EarlyStoppingCallback>`_
and `OneCycleScheduler <https://docs.fast.ai/callback.schedule.html#ParamScheduler>`_ callbacks
:param log_models: If ``True``, trained models are logged as MLflow model artifacts.
If ``False``, trained models are not logged.
:param disable: If ``True``, disables the Fastai autologging integration. If ``False``,
enables the Fastai autologging integration.
:param exclusive: If ``True``, autologged content is not logged to user-created fluent runs.
If ``False``, autologged content is logged to the active fluent run,
which may be user-created.
:param disable_for_unsupported_versions: If ``True``, disable autologging for versions of
fastai that have not been tested against this version of the MLflow client
or are incompatible.
:param silent: If ``True``, suppress all event logs and warnings from MLflow during Fastai
autologging. If ``False``, show all events and warnings during Fastai
autologging.
.. code-block:: python
:caption: Example
# This is a modified example from
# https://github.com/mlflow/mlflow/tree/master/examples/fastai
# demonstrating autolog capabilites.
import fastai.vision as vis
import mlflow.fastai
from mlflow.tracking import MlflowClient
def print_auto_logged_info(r):
tags = {k: v for k, v in r.data.tags.items() if not k.startswith("mlflow.")}
artifacts = [f.path for f in MlflowClient().list_artifacts(r.info.run_id, "model")]
print("run_id: {}".format(r.info.run_id))
print("artifacts: {}".format(artifacts))
print("params: {}".format(r.data.params))
print("metrics: {}".format(r.data.metrics))
print("tags: {}".format(tags))
def main(epochs=5, learning_rate=0.01):
# Download and untar the MNIST data set
path = vis.untar_data(vis.URLs.MNIST_SAMPLE)
# Prepare, transform, and normalize the data
data = vis.ImageDataBunch.from_folder(path, ds_tfms=(vis.rand_pad(2, 28), []), bs=64)
data.normalize(vis.imagenet_stats)
# Create CNN the Learner model
model = vis.cnn_learner(data, vis.models.resnet18, metrics=vis.accuracy)
# Enable auto logging
mlflow.fastai.autolog()
# Start MLflow session
with mlflow.start_run() as run:
model.fit(epochs, learning_rate)
# fetch the auto logged parameters, metrics, and artifacts
print_auto_logged_info(mlflow.get_run(run_id=run.info.run_id))
main()
.. code-block:: text
:caption: output
run_id: 5a23dcbcaa334637814dbce7a00b2f6a
artifacts: ['model/MLmodel', 'model/conda.yaml', 'model/model.fastai']
params: {'wd': 'None',
'bn_wd': 'True',
'opt_func': 'Adam',
'epochs': '5', '
train_bn': 'True',
'num_layers': '60',
'lr': '0.01',
'true_wd': 'True'}
metrics: {'train_loss': 0.024,
'accuracy': 0.99214,
'valid_loss': 0.021}
# Tags model summary omitted too long
tags: {...}
.. figure:: ../_static/images/fastai_autolog.png
Fastai autologged MLflow entities
"""
from fastai.learner import Learner
from fastai.callback.hook import module_summary, layer_info, find_bs, _print_shapes
from fastai.callback.all import EarlyStoppingCallback, TrackerCallback
def getFastaiCallback(metrics_logger, is_fine_tune=False):
from mlflow.fastai.callback import __MlflowFastaiCallback
return __MlflowFastaiCallback(
metrics_logger=metrics_logger, log_models=log_models, is_fine_tune=is_fine_tune,
)
def _find_callback_of_type(callback_type, callbacks):
for callback in callbacks:
if isinstance(callback, callback_type):
return callback
return None
def _log_early_stop_callback_params(callback):
if callback:
try:
earlystopping_params = {
"early_stop_monitor": callback.monitor,
"early_stop_min_delta": callback.min_delta,
"early_stop_patience": callback.patience,
"early_stop_comp": callback.comp.__name__,
}
try_mlflow_log(mlflow.log_params, earlystopping_params)
except Exception: # pylint: disable=W0703
return
def _log_model_info(learner):
# The process excuted here, are incompatible with TrackerCallback
# Hence it is removed and add again after the execution
remove_cbs = [cb for cb in learner.cbs if isinstance(cb, TrackerCallback)]
if remove_cbs:
learner.remove_cbs(remove_cbs)
xb = learner.dls.train.one_batch()[: learner.dls.train.n_inp]
infos = layer_info(learner, *xb)
bs = find_bs(xb)
inp_sz = _print_shapes(map(lambda x: x.shape, xb), bs)
try_mlflow_log(mlflow.log_param, "input_size", inp_sz)
try_mlflow_log(mlflow.log_param, "num_layers", len(infos))
summary = module_summary(learner, *xb)
# Add again TrackerCallback
if remove_cbs:
learner.add_cbs(remove_cbs)
tempdir = tempfile.mkdtemp()
try:
summary_file = os.path.join(tempdir, "module_summary.txt")
with open(summary_file, "w") as f:
f.write(summary)
try_mlflow_log(mlflow.log_artifact, local_path=summary_file)
finally:
shutil.rmtree(tempdir)
def _run_and_log_function(self, original, args, kwargs, unlogged_params, is_fine_tune=False):
# Check if is trying to fit while fine tuning or not
mlflow_cbs = [cb for cb in self.cbs if cb.name == "___mlflow_fastai"]
fit_in_fine_tune = (
original.__name__ == "fit" and len(mlflow_cbs) > 0 and mlflow_cbs[0].is_fine_tune
)
if not fit_in_fine_tune:
log_fn_args_as_params(original, list(args), kwargs, unlogged_params)
run_id = mlflow.active_run().info.run_id
with batch_metrics_logger(run_id) as metrics_logger:
if not fit_in_fine_tune:
early_stop_callback = _find_callback_of_type(EarlyStoppingCallback, self.cbs)
_log_early_stop_callback_params(early_stop_callback)
# First try to remove if any already registered callback
self.remove_cbs(mlflow_cbs)
# Log information regarding model and data without bar and print-out
with self.no_bar(), self.no_logging():
try_mlflow_log(_log_model_info, learner=self)
mlflowFastaiCallback = getFastaiCallback(
metrics_logger=metrics_logger, is_fine_tune=is_fine_tune
)
# Add the new callback
self.add_cb(mlflowFastaiCallback)
result = original(self, *args, **kwargs)
return result
def fit(original, self, *args, **kwargs):
unlogged_params = ["self", "cbs", "learner", "lr", "lr_max", "wd"]
return _run_and_log_function(
self, original, args, kwargs, unlogged_params, is_fine_tune=False
)
safe_patch(FLAVOR_NAME, Learner, "fit", fit, manage_run=True)
def fine_tune(original, self, *args, **kwargs):
unlogged_params = ["self", "cbs", "learner", "lr", "lr_max", "wd"]
return _run_and_log_function(
self, original, args, kwargs, unlogged_params, is_fine_tune=True
)
safe_patch(FLAVOR_NAME, Learner, "fine_tune", fine_tune, manage_run=True)