wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-20 12:40:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add ray.util package and move libraries from experimental (#7100)

Browse Source
This commit is contained in:
Eric Liang
2020-02-18 13:43:19 -08:00
committed by GitHub
parent fae99ecb8e
commit 5df801605e
113 changed files with 305 additions and 637 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.travis.yml
+1 -1
View File
@@ -294,7 +294,7 @@ script:
- ./ci/suppress_output bazel test --build_tests_only --show_progress_rate_limit=100 --test_output=errors -- //:all -rllib/...
# ray serve tests
- if [ $RAY_CI_SERVE_AFFECTED == "1" ]; then ./ci/keep_alive bazel test --spawn_strategy=local --flaky_test_attempts=3 --nocache_test_results --test_verbose_timeout_warnings --progress_report_interval=100 --show_progress_rate_limit=100 --show_timestamps --test_output=errors --test_tag_filters=-jenkins_only python/ray/experimental/serve/...; fi
- if [ $RAY_CI_SERVE_AFFECTED == "1" ]; then ./ci/keep_alive bazel test --spawn_strategy=local --flaky_test_attempts=3 --nocache_test_results --test_verbose_timeout_warnings --progress_report_interval=100 --show_progress_rate_limit=100 --show_timestamps --test_output=errors --test_tag_filters=-jenkins_only python/ray/serve/...; fi
# ray operator tests
- cd ./deploy/ray-operator/
BUILD.bazel
+1
View File
@@ -1176,6 +1176,7 @@ filegroup(
"python/ray/core/generated/ray/protocol/__init__.py",
"python/ray/dashboard/dashboard.py",
"python/ray/experimental/*.py",
"python/ray/util/*.py",
"python/ray/internal/*.py",
"python/ray/projects/*.py",
"python/ray/projects/schema.json",
README.rst
+1 -1
View File
@@ -15,7 +15,7 @@ Ray is packaged with the following libraries for accelerating machine learning w
- `Tune`_: Scalable Hyperparameter Tuning
- `RLlib`_: Scalable Reinforcement Learning
- `Distributed Training <https://ray.readthedocs.io/en/latest/distributed_training.html>`__
- `RaySGD <https://ray.readthedocs.io/en/latest/raysgd/raysgd.html>`__: Distributed Training Wrappers
Install Ray with: ``pip install ray``. For nightly wheels, see the
`Installation page <https://ray.readthedocs.io/en/latest/installation.html>`__.
ci/jenkins_tests/run_tune_tests.sh
+15 -15
View File
@@ -144,49 +144,49 @@ $SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE}
######################## SGD TESTS #################################
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python -m pytest /ray/python/ray/experimental/sgd/tests
python -m pytest /ray/python/ray/util/sgd/tests
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/doc/examples/doc_code/raysgd_torch_signatures.py
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/train_example.py
python /ray/python/ray/util/sgd/pytorch/examples/train_example.py
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/train_example.py --num-replicas=2
python /ray/python/ray/util/sgd/pytorch/examples/train_example.py --num-replicas=2
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/tune_example.py
python /ray/python/ray/util/sgd/pytorch/examples/tune_example.py
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/tune_example.py --num-replicas=2
python /ray/python/ray/util/sgd/pytorch/examples/tune_example.py --num-replicas=2
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/cifar_pytorch_example.py --smoke-test
python /ray/python/ray/util/sgd/pytorch/examples/cifar_pytorch_example.py --smoke-test
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/cifar_pytorch_example.py --smoke-test --num-replicas=2
python /ray/python/ray/util/sgd/pytorch/examples/cifar_pytorch_example.py --smoke-test --num-replicas=2
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/cifar_pytorch_example.py --smoke-test --tune
python /ray/python/ray/util/sgd/pytorch/examples/cifar_pytorch_example.py --smoke-test --tune
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/pytorch/examples/dcgan.py --smoke-test --num-replicas=2
python /ray/python/ray/util/sgd/pytorch/examples/dcgan.py --smoke-test --num-replicas=2
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/tf/examples/tensorflow_train_example.py
python /ray/python/ray/util/sgd/tf/examples/tensorflow_train_example.py
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/tf/examples/tensorflow_train_example.py --num-replicas=2
python /ray/python/ray/util/sgd/tf/examples/tensorflow_train_example.py --num-replicas=2
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/tf/examples/tensorflow_train_example.py --tune
python /ray/python/ray/util/sgd/tf/examples/tensorflow_train_example.py --tune
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/tf/examples/cifar_tf_example.py --smoke-test
python /ray/python/ray/util/sgd/tf/examples/cifar_tf_example.py --smoke-test
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/tf/examples/cifar_tf_example.py --num-replicas 2 --smoke-test
python /ray/python/ray/util/sgd/tf/examples/cifar_tf_example.py --num-replicas 2 --smoke-test
$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} --memory-swap=-1 $DOCKER_SHA \
python /ray/python/ray/experimental/sgd/tf/examples/cifar_tf_example.py --num-replicas 2 --smoke-test --augment-data
python /ray/python/ray/util/sgd/tf/examples/cifar_tf_example.py --num-replicas 2 --smoke-test --augment-data
ci/long_running_tests/workloads/serve.py
+2 -2
View File
@@ -5,9 +5,9 @@ from subprocess import PIPE
import requests
import ray
from ray.experimental import serve
from ray import serve
from ray.cluster_utils import Cluster
from ray.experimental.serve.kv_store_service import RayInternalKVStore
from ray.serve.kv_store_service import RayInternalKVStore
num_redis_shards = 1
redis_max_memory = 10**8
ci/stress_tests/run_application_stress_tests.sh
+1 -1
View File
@@ -127,7 +127,7 @@ test_sgd(){
echo "Try running SGD stress test."
{
SGD_DIR=$ROOT_DIR/../../python/ray/experimental/sgd/
SGD_DIR=$ROOT_DIR/../../python/ray/util/sgd/
ray --logging-level=DEBUG up -y "$CLUSTER" &&
# TODO: fix submit so that args work
ray rsync_up "$CLUSTER" "$SGD_DIR/mnist_example.py" mnist_example.py &&
ci/travis/determine_tests_to_run.py
+1 -1
View File
@@ -65,7 +65,7 @@ if __name__ == "__main__":
RAY_CI_RLLIB_FULL_AFFECTED = 1
RAY_CI_LINUX_WHEELS_AFFECTED = 1
RAY_CI_MACOS_WHEELS_AFFECTED = 1
elif changed_file.startswith("python/ray/experimental/serve"):
elif changed_file.startswith("python/ray/serve"):
RAY_CI_SERVE_AFFECTED = 1
RAY_CI_LINUX_WHEELS_AFFECTED = 1
RAY_CI_MACOS_WHEELS_AFFECTED = 1
doc/examples/doc_code/raysgd_torch_signatures.py
+2 -2
View File
@@ -46,7 +46,7 @@ def optimizer_creator(model, config):
# __torch_data_start__
from ray.experimental.sgd.pytorch.examples.train_example import LinearDataset
from ray.util.sgd.pytorch.examples.train_example import LinearDataset
def data_creator(config):
"""Constructs torch.utils.data.Dataset objects.
@@ -108,7 +108,7 @@ ray.init()
# __torch_ray_end__
# __torch_trainer_start__
from ray.experimental.sgd import PyTorchTrainer
from ray.util.sgd import PyTorchTrainer
trainer = PyTorchTrainer(
model_creator,
doc/examples/overview.rst
+8 -2
View File
@@ -1,5 +1,11 @@
Examples Overview
=================
Ray Tutorials and Examples
==========================
Get started with Ray, Tune, and RLlib with these notebooks that you can run online in CoLab or Binder:
* `Ray Tutorial Notebooks <https://github.com/ray-project/tutorial>`__
Example Gallery
---------------
.. customgalleryitem::
:tooltip: Build a simple parameter server using Ray.
doc/source/actors.rst
+5 -5
View File
@@ -197,19 +197,19 @@ If we instantiate an actor, we can pass the handle around to various tasks.
print(ray.get(counter.get_counter.remote()))
Actor Pool (Experimental)
-------------------------
Actor Pool
----------
The ``ray.experimental`` module contains a utility class, ``ActorPool``.
The ``ray.util`` module contains a utility class, ``ActorPool``.
This class is similar to multiprocessing.Pool and lets you schedule Ray tasks over a fixed pool of actors.
.. code-block::
from ray.experimental import ActorPool
from ray.util import ActorPool
a1, a2 = Actor.remote(), Actor.remote()
pool = ActorPool([a1, a2])
print(pool.map(lambda a, v: a.double.remote(v), [1, 2, 3, 4]))
# [2, 4, 6, 8]
See the `package reference <package-ref.html#ray.experimental.ActorPool>`_ for more information.
See the `package reference <package-ref.html#ray.util.ActorPool>`_ for more information.
doc/source/advanced.rst
+1 -1
View File
@@ -259,7 +259,7 @@ driver:
.. code-block:: python
counter = ray.experimental.get_actor("CounterActor")
counter = ray.util.get_actor("CounterActor")
print(ray.get(counter.get_counter.remote()))
Note that just creating a named actor is allowed, this actor will be cleaned
doc/source/async_api.rst
+2 -2
View File
@@ -1,5 +1,5 @@
Async API (Experimental)
========================
AsyncIO / Concurrency for Actors
================================
Since Python 3.5, it is possible to write concurrent code using the
``async/await`` `syntax <https://docs.python.org/3/library/asyncio.html>`__.
doc/source/cluster-index.rst
+1
View File
@@ -12,3 +12,4 @@ How to setup your cluster and use Ray most effectively.
deploy-on-yarn.rst
deploy-on-kubernetes.rst
deploying-on-slurm.rst
projects.rst
doc/source/development.rst
+2 -6
View File
@@ -237,12 +237,8 @@ scripts. Some of the examples include:
* ``bazel test --build_tests_only //:all``
* Ray serving test commands:
* ``python -m pytest python/ray/experimental/serve/tests``
* ``python python/ray/experimental/serve/examples/echo_full.py``
* Ray test commands:
* ``python/ray/experimental/test/async_test.py``
* ``python/ray/tests/py3_test.py``
* ``python -m pytest python/ray/serve/tests``
* ``python python/ray/serve/examples/echo_full.py``
If a Travis-CI build exception doesn't appear to be related to your change,
please visit `this link <https://ray-travis-tracker.herokuapp.com/>`_ to
doc/source/index.rst
+9 -14
View File
@@ -16,7 +16,7 @@ Ray is packaged with the following libraries for accelerating machine learning w
- `Tune`_: Scalable Hyperparameter Tuning
- `RLlib`_: Scalable Reinforcement Learning
- `RaySGD`_: Distributed Training
- `RaySGD`_: Distributed Training Wrappers
Star us on `on GitHub`_. You can also get started by visiting our `Tutorials <https://github.com/ray-project/tutorial>`_. For the latest wheels (nightlies), see the `installation page <installation.html>`__.
@@ -239,8 +239,7 @@ Getting Involved
using-ray.rst
configure.rst
cluster-index.rst
Tutorials <https://github.com/ray-project/tutorial>
Examples <auto_examples/overview.rst>
Tutorial and Examples <auto_examples/overview.rst>
package-ref.rst
.. toctree::
@@ -254,9 +253,9 @@ Getting Involved
tune-distributed.rst
tune-schedulers.rst
tune-searchalg.rst
tune-package-ref.rst
tune-design.rst
tune-examples.rst
tune-package-ref.rst
tune-contrib.rst
.. toctree::
@@ -272,31 +271,27 @@ Getting Involved
rllib-offline.rst
rllib-concepts.rst
rllib-examples.rst
rllib-dev.rst
rllib-package-ref.rst
rllib-dev.rst
.. toctree::
:maxdepth: -1
:caption: RaySGD
:caption: Ray SGD
raysgd/raysgd.rst
raysgd/raysgd_pytorch.rst
raysgd/raysgd_tensorflow.rst
raysgd/raysgd_ref.rst
.. toctree::
:maxdepth: -1
:caption: Experimental
:caption: Other Libraries
pandas_on_ray.rst
projects.rst
signals.rst
async_api.rst
serve.rst
iter.rst
multiprocessing.rst
joblib.rst
iter.rst
pandas_on_ray.rst
serve.rst
.. toctree::
:maxdepth: -1
doc/source/iter.rst
+21 -15
View File
@@ -1,13 +1,19 @@
Parallel Iterator API (Experimental)
====================================
Distributed Iterators
=====================
``ray.experimental.iter`` provides a parallel iterator API for simple data ingest
and processing. It can be thought of as syntactic sugar around Ray actors and ``ray.wait`` loops.
.. _`issue on GitHub`: https://github.com/ray-project/ray/issues
``ray.util.iter`` provides a parallel iterator API for simple data ingest and processing. It can be thought of as syntactic sugar around Ray actors and ``ray.wait`` loops.
Parallel iterators are lazy and can operate over infinite sequences of items. Iterator
transformations are only executed when the user calls ``next()`` to fetch the next output
item from the iterator.
.. note::
This API is new and may be revised in future Ray releases. If you encounter
any bugs, please file an `issue on GitHub`_.
Concepts
--------
@@ -18,20 +24,20 @@ create a worker actor that produces the data for each shard of the iterator:
.. code-block:: python
# Create an iterator with 2 worker actors over the list [1, 2, 3, 4].
>>> it = ray.experimental.iter.from_items([1, 2, 3, 4], num_shards=2)
>>> it = ray.util.iter.from_items([1, 2, 3, 4], num_shards=2)
ParallelIterator[from_items[int, 4, shards=2]]
# Create an iterator with 32 worker actors over range(1000000).
>>> it = ray.experimental.iter.from_range(1000000, num_shards=32)
>>> it = ray.util.iter.from_range(1000000, num_shards=32)
ParallelIterator[from_range[1000000, shards=32]]
# Create an iterator over two range(10) generators.
>>> it = ray.experimental.iter.from_iterators([range(10), range(10)])
>>> it = ray.util.iter.from_iterators([range(10), range(10)])
ParallelIterator[from_iterators[shards=2]]
# Create an iterator from existing worker actors. These actors must
# implement the ParallelIteratorWorker interface.
>>> it = ray.experimental.iter.from_actors([a1, a2, a3, a4])
>>> it = ray.util.iter.from_actors([a1, a2, a3, a4])
ParallelIterator[from_actors[shards=4]]
Simple transformations can be chained on the iterator, such as mapping,
@@ -58,7 +64,7 @@ correspond to ``ray.get`` and ``ray.wait`` loops over the actors respectively:
.. code-block:: python
# Gather items synchronously (deterministic round robin across shards):
>>> it = ray.experimental.iter.from_range(1000000, 1)
>>> it = ray.util.iter.from_range(1000000, 1)
>>> it = it.gather_sync()
LocalIterator[ParallelIterator[from_range[1000000, shards=1]].gather_sync()]
@@ -72,7 +78,7 @@ correspond to ``ray.get`` and ``ray.wait`` loops over the actors respectively:
[0, 2, 4, 6, 8]
# Async gather can be used for better performance, but it is non-deterministic.
>>> it = ray.experimental.iter.from_range(1000, 4).gather_async()
>>> it = ray.util.iter.from_range(1000, 4).gather_async()
>>> it.take(5)
[0, 250, 500, 750, 1]
@@ -83,7 +89,7 @@ each shard should only be read by one process at a time:
.. code-block:: python
# Get local iterators representing the shards of this ParallelIterator:
>>> it = ray.experimental.iter.from_range(10000, 3)
>>> it = ray.util.iter.from_range(10000, 3)
>>> [s0, s1, s2] = it.shards()
[LocalIterator[from_range[10000, shards=3].shard[0]],
LocalIterator[from_range[10000, shards=3].shard[1]],
@@ -122,7 +128,7 @@ This means that you can pass a stateful callable to ``.foreach()``:
self.total += x
return (self.total, x)
it = ray.experimental.iter.from_range(5, 1)
it = ray.util.iter.from_range(5, 1)
for x in it.for_each(CumulativeSum()).gather_sync():
print(x)
@@ -150,7 +156,7 @@ streaming grep:
file_list = glob.glob("/var/log/syslog*.gz")
it = (
ray.experimental.iter.from_items(file_list, num_shards=4)
ray.util.iter.from_items(file_list, num_shards=4)
.for_each(lambda f: gzip.open(f).readlines())
.flatten()
.for_each(lambda line: line.decode("utf-8"))
@@ -184,7 +190,7 @@ distributed training:
print("train on", batch) # perform model update with batch
it = (
ray.experimental.iter.from_range(1000000, num_shards=4, repeat=True)
ray.util.iter.from_range(1000000, num_shards=4, repeat=True)
.batch(1024)
.for_each(np.array)
)
@@ -195,7 +201,7 @@ distributed training:
API Reference
-------------
.. automodule:: ray.experimental.iter
.. automodule:: ray.util.iter
:members:
:show-inheritance:
:special-members:
doc/source/joblib.rst
+9 -11
View File
@@ -1,12 +1,5 @@
sklearn Ray Backend API (Experimental)
=======================================
.. warning::
Support for running scikit-learn on Ray is an experimental feature,
so it may be changed at any time without warning. If you encounter any
bugs/shortcomings/incompatibilities, please file an `issue on GitHub`_.
Contributions are always welcome!
Distributed Scikit-learn / Joblib
=================================
.. _`issue on GitHub`: https://github.com/ray-project/ray/issues
@@ -15,6 +8,11 @@ implementing a Ray backend for `joblib`_ using `Ray Actors <actors.html>`__
instead of local processes. This makes it easy to scale existing applications
that use scikit-learn from a single node to a cluster.
.. note::
This API is new and may be revised in future Ray releases. If you encounter
any bugs, please file an `issue on GitHub`_.
.. _`joblib`: https://joblib.readthedocs.io
.. _`scikit-learn`: https://scikit-learn.org
@@ -22,7 +20,7 @@ Quickstart
----------
To get started, first `install Ray <installation.html>`__, then use
``from ray.experimental.joblib import register_ray`` and run ``register_ray()``.
``from ray.util.joblib import register_ray`` and run ``register_ray()``.
This will register Ray as a joblib backend for scikit-learn to use.
Then run your original scikit-learn code inside
``with joblib.parallel_backend('ray')``. This will start a local Ray cluster.
@@ -46,7 +44,7 @@ a multi-node Ray cluster instead.
search = RandomizedSearchCV(model, param_space, cv=5, n_iter=300, verbose=10)
import joblib
from ray.experimental.joblib import register_ray
from ray.util.joblib import register_ray
register_ray()
with joblib.parallel_backend('ray'):
search.fit(digits.data, digits.target)
doc/source/memory-management.rst
+4 -3
View File
@@ -84,10 +84,11 @@ Heap memory quota
When Ray starts, it queries the available memory on a node / container not reserved for Redis and the object store or being used by other applications. This is considered "available memory" that actors and tasks can request memory out of. You can also set ``memory=<bytes>`` on Ray init to tell Ray explicitly how much memory is available.
.. note::
.. important::
Setting available memory for the node does not impose any limits on memory usage
of tasks. To set per-task limits, see the following sections.
Setting available memory for the node does NOT impose any limits on memory usage
unless you specify memory resource requirements in decorators. By default, tasks
and actors request no memory (and hence have no limit).
To tell the Ray scheduler a task or actor requires a certain amount of available memory to run, set the ``memory`` argument. The Ray scheduler will then reserve the specified amount of available memory during scheduling, similar to how it handles CPU and GPU resources:
doc/source/multiprocessing.rst
+7 -9
View File
@@ -1,12 +1,5 @@
multiprocessing.Pool API
========================
.. warning::
Support for the multiprocessing.Pool API on Ray is an experimental feature,
so it may be changed at any time without warning. If you encounter any
bugs/shortcomings/incompatibilities, please file an `issue on GitHub`_.
Contributions are always welcome!
Distributed multiprocessing.Pool
================================
.. _`issue on GitHub`: https://github.com/ray-project/ray/issues
@@ -15,6 +8,11 @@ using `Ray Actors <actors.html>`__ instead of local processes. This makes it eas
to scale existing applications that use ``multiprocessing.Pool`` from a single node
to a cluster.
.. note::
This API is new and may be revised in future Ray releases. If you encounter
any bugs, please file an `issue on GitHub`_.
.. _`multiprocessing.Pool API`: https://docs.python.org/3/library/multiprocessing.html#module-multiprocessing.pool
Quickstart
doc/source/raysgd/raysgd.rst
+11 -8
View File
@@ -1,9 +1,7 @@
RaySGD: Distributed Deep Learning
=================================
RaySGD: Distributed Training Wrappers
=====================================
.. image:: raysgdlogo.png
:scale: 20%
:align: center
.. _`issue on GitHub`: https://github.com/ray-project/ray/issues
RaySGD is a lightweight library for distributed deep learning, providing thin wrappers around PyTorch and TensorFlow native modules for data parallel training.
@@ -13,7 +11,10 @@ The main features are:
- **Composability**: RaySGD is built on top of the Ray Actor API, enabling seamless integration with existing Ray applications such as RLlib, Tune, and Ray.Serve.
- **Scale up and down**: Start on single CPU. Scale up to multi-node, multi-CPU, or multi-GPU clusters by changing 2 lines of code.
.. tip:: We need your feedback! RaySGD is currently early in its development, and we're hoping to get feedback from people using or considering it. We'd love `to get in touch <https://forms.gle/26EMwdahdgm7Lscy9>`_!
.. note::
This API is new and may be revised in future Ray releases. If you encounter
any bugs, please file an `issue on GitHub`_.
Getting Started
@@ -28,8 +29,8 @@ You can start a ``PyTorchTrainer`` with the following:
import torch.nn as nn
from torch import distributed
from ray.experimental.sgd import PyTorchTrainer
from ray.experimental.sgd.examples.train_example import LinearDataset
from ray.util.sgd import PyTorchTrainer
from ray.util.sgd.examples.train_example import LinearDataset
def model_creator(config):
@@ -61,3 +62,5 @@ You can start a ``PyTorchTrainer`` with the following:
print(stats)
trainer1.shutdown()
print("success!")
.. tip:: Get in touch with us if you're using or considering using `RaySGD <https://forms.gle/26EMwdahdgm7Lscy9>`_!
doc/source/raysgd/raysgd_pytorch.rst
+8 -17
View File
@@ -1,8 +1,5 @@
RaySGD Pytorch
==============
.. image:: raysgd-pytorch.svg
:align: center
Distributed PyTorch
===================
The RaySGD ``PyTorchTrainer`` simplifies distributed model training for PyTorch. The ``PyTorchTrainer`` is a wrapper around ``torch.distributed.launch`` with a Python API to easily incorporate distributed training into a larger Python application, as opposed to needing to wrap your training code in bash scripts.
@@ -16,7 +13,7 @@ For end to end examples leveraging RaySGD PyTorchTrainer, jump to :ref:`raysgd-p
Setting up training
-------------------
.. tip:: We need your feedback! RaySGD is currently early in its development, and we're hoping to get feedback from people using or considering it. We'd love `to get in touch <https://forms.gle/26EMwdahdgm7Lscy9>`_!
.. tip:: Get in touch with us if you're using or considering using `RaySGD <https://forms.gle/26EMwdahdgm7Lscy9>`_!
The ``PyTorchTrainer`` can be constructed with functions that wrap components of the training script. Specifically, it requires constructors for the Model, Data, Optimizer, Loss, and ``lr_scheduler`` to create replicated copies across different devices and machines.
@@ -154,8 +151,6 @@ After training, you may want to reappropriate the Ray cluster. To release Ray re
Initialization Functions
------------------------
.. warning:: This is still an experimental API and is subject to change without warning.
You may want to run some initializers on each worker when they are started. This may be something like setting an environment variable or downloading some data. You can do this via the ``initialization_hook`` parameter:
.. code-block:: python
@@ -305,11 +300,9 @@ Users can set ``checkpoint="auto"`` to always checkpoint the current model befor
Advanced: Hyperparameter Tuning
-------------------------------
.. warning:: This is still an experimental API and is subject to change without warning.
``PyTorchTrainer`` naturally integrates with Tune via the ``PyTorchTrainable`` interface. The same arguments to ``PyTorchTrainer`` should be passed into the ``tune.run(config=...)`` as shown below.
.. literalinclude:: ../../../python/ray/experimental/sgd/pytorch/examples/tune_example.py
.. literalinclude:: ../../../python/ray/util/sgd/pytorch/examples/tune_example.py
:language: python
:start-after: __torch_tune_example__
@@ -321,7 +314,7 @@ In certain scenarios such as training GANs, you may want to use multiple models
If multiple models, optimizers, or schedulers are returned, you will need to provide a custom training function (and custom validation function if you plan to call ``validate``).
You can see the `DCGAN script <https://github.com/ray-project/ray/blob/master/python/ray/experimental/sgd/pytorch/examples/dcgan.py>`_ for an end-to-end example.
You can see the `DCGAN script <https://github.com/ray-project/ray/blob/master/python/ray/util/sgd/pytorch/examples/dcgan.py>`_ for an end-to-end example.
.. code-block:: python
@@ -362,8 +355,6 @@ You can see the `DCGAN script <https://github.com/ray-project/ray/blob/master/py
Custom Training and Validation Functions
----------------------------------------
.. warning:: This is still an experimental API and is subject to change in the near future.
``PyTorchTrainer`` allows you to run a custom training and validation step in parallel on each worker, providing a flexibility similar to using PyTorch natively. This is done via the ``train_function`` and ``validation_function`` parameters.
Note that this is needed if the model creator returns multiple models, optimizers, or schedulers.
@@ -477,13 +468,13 @@ PyTorchTrainer Examples
Here are some examples of using RaySGD for training PyTorch models. If you'd like
to contribute an example, feel free to create a `pull request here <https://github.com/ray-project/ray/>`_.
- `PyTorch training example <https://github.com/ray-project/ray/blob/master/python/ray/experimental/sgd/pytorch/examples/train_example.py>`__:
- `PyTorch training example <https://github.com/ray-project/ray/blob/master/python/ray/util/sgd/pytorch/examples/train_example.py>`__:
Simple example of using Ray's PyTorchTrainer.
- `CIFAR10 example <https://github.com/ray-project/ray/blob/master/python/ray/experimental/sgd/pytorch/examples/cifar_pytorch_example.py>`__:
- `CIFAR10 example <https://github.com/ray-project/ray/blob/master/python/ray/util/sgd/pytorch/examples/cifar_pytorch_example.py>`__:
Training a ResNet18 model on CIFAR10. It uses a custom training
function, a custom validation function, and custom initialization code for each worker.
- `DCGAN example <https://github.com/ray-project/ray/blob/master/python/ray/experimental/sgd/pytorch/examples/dcgan.py>`__:
- `DCGAN example <https://github.com/ray-project/ray/blob/master/python/ray/util/sgd/pytorch/examples/dcgan.py>`__:
Training a Deep Convolutional GAN on MNIST. It constructs
two models and two optimizers and uses a custom training and validation function.
doc/source/raysgd/raysgd_ref.rst
+3 -3
View File
@@ -6,7 +6,7 @@ Package Reference
PyTorchTrainer
--------------
.. autoclass:: ray.experimental.sgd.pytorch.PyTorchTrainer
.. autoclass:: ray.util.sgd.pytorch.PyTorchTrainer
:members:
.. automethod:: __init__
@@ -15,13 +15,13 @@ PyTorchTrainer
PyTorchTrainable
----------------
.. autoclass:: ray.experimental.sgd.pytorch.PyTorchTrainable
.. autoclass:: ray.util.sgd.pytorch.PyTorchTrainable
:members:
TFTrainer
---------
.. autoclass:: ray.experimental.sgd.tf.TFTrainer
.. autoclass:: ray.util.sgd.tf.TFTrainer
:members:
.. automethod:: __init__
doc/source/raysgd/raysgd_tensorflow.rst
+7 -4
View File
@@ -1,9 +1,12 @@
RaySGD TensorFlow
=================
Distributed TensorFlow
======================
RaySGD's ``TFTrainer`` simplifies distributed model training for Tensorflow. The ``TFTrainer`` is a wrapper around ``MultiWorkerMirroredStrategy`` with a Python API to easily incorporate distributed training into a larger Python application, as opposed to write custom logic of setting environments and starting separate processes.
.. important:: This API has only been tested with TensorFlow2.0rc and is still highly experimental. Please file bug reports if you run into any - thanks!
Under the hood, ``TFTrainer`` will create *replicas* of your model (controlled by ``num_replicas``), each of which is managed by a Ray actor.
.. image:: raysgd-actors.svg
:align: center
.. tip:: We need your feedback! RaySGD is currently early in its development, and we're hoping to get feedback from people using or considering it. We'd love `to get in touch <https://forms.gle/26EMwdahdgm7Lscy9>`_!
@@ -69,6 +72,6 @@ TFTrainer Example
Below is an example of using Ray's TFTrainer. Under the hood, ``TFTrainer`` will create *replicas* of your model (controlled by ``num_replicas``) which are each managed by a worker.
.. literalinclude:: ../../../python/ray/experimental/sgd/tf/examples/tensorflow_train_example.py
.. literalinclude:: ../../../python/ray/util/sgd/tf/examples/tensorflow_train_example.py
:language: python
doc/source/rllib-dev.rst
+2 -2
View File
@@ -1,5 +1,5 @@
RLlib Development
=================
Contributing to RLlib
=====================
Development Install
-------------------
doc/source/rllib-training.rst
+1 -1
View File
@@ -442,7 +442,7 @@ Sometimes, it is necessary to coordinate between pieces of code that live in dif
.. code-block:: python
from ray.experimental import named_actors
from ray.util import named_actors
@ray.remote
class Counter:
doc/source/serve.rst
+10 -7
View File
@@ -1,20 +1,23 @@
Ray Serve (Experimental)
========================
Ray Serve
=========
.. _`issue on GitHub`: https://github.com/ray-project/ray/issues
Ray Serve is a serving library that exposes python function/classes to HTTP.
It has built-in support for flexible traffic policy. This means you can easy
split incoming traffic to multiple implementations.
With Ray Serve, you can deploy your services at any scale.
.. warning::
Ray Serve is Python 3 only.
Ray Serve is under development and its API may be revised in future Ray releases. If you encounter any bugs, please file an `issue on GitHub`_.
With Ray Serve, you can deploy your services at any scale.
Quickstart
----------
.. literalinclude:: ../../python/ray/experimental/serve/examples/echo_full.py
.. literalinclude:: ../../python/ray/serve/examples/echo_full.py
API
---
.. automodule:: ray.experimental.serve
.. automodule:: ray.serve
:members:
doc/source/signals.rst
-168
View File
@@ -1,168 +0,0 @@
Signal API (Experimental)
=========================
This experimental API allows tasks and actors to generate signals which can
be received by other tasks and actors. In addition, task failures and actor
method failures generate error signals. The error signals enable applications
to detect failures and potentially recover from failures.
.. autofunction:: ray.experimental.signal.send
Here is a simple example of a remote function that sends a user-defined signal.
.. code-block:: python
import ray.experimental.signal as signal
# Define an application level signal.
class UserSignal(signal.Signal):
def __init__(self, value):
self.value = value
def get_value(self):
return self.value
# Define a remote function that sends a user-defined signal.
@ray.remote
def send_signal(value):
signal.send(UserSignal(value))
.. autofunction:: ray.experimental.signal.receive
Here is a simple example of how to receive signals from an actor or task identified
by ``a``. Note that an actor is identified by its handle, and a task by one of its
object ID return values.
.. code-block:: python
import ray.experimental.signal as signal
# This returns a possibly empty list of all signals that have been sent by 'a'
# since the last invocation of signal.receive from within this process. If 'a'
# did not send any signals, then this will wait for up to 10 seconds to receive
# a signal from 'a'.
signal_list = signal.receive([a], timeout=10)
.. autofunction:: ray.experimental.signal.reset
Example: sending a user signal
------------------------------
The code below show a simple example in which a task, called ``send_signal()``
sends a user signal and the driver gets it by invoking ``signal.receive()``.
.. code-block:: python
import ray.experimental.signal as signal
# Define a user signal.
class UserSignal(signal.Signal):
def __init__(self, value):
self.value = value
def get_value(self):
return self.value
@ray.remote
def send_signal(value):
signal.send(UserSignal(value))
return
signal_value = 'simple signal'
object_id = send_signal.remote(signal_value)
# Wait up to 10sec to receive a signal from the task. Note the task is
# identified by the object_id it returns.
result_list = signal.receive([object_id], timeout=10)
# Print signal values. This should print "simple_signal".
# Note that result_list[0] is the signal we expect from the task.
# The signal is a tuple where the first element is the first object ID
# returned by the task and the second element is the signal object.
print(result_list[0][1].get_value())
Example: Getting an error signals
---------------------------------
This is a simple example in which a driver gets an error signal caused
by the failure of ``task()``.
.. code-block:: python
@ray.remote
def task():
raise Exception('exception message')
object_id = task.remote()
try:
ray.get(object_id)
except Exception as e:
pass
finally:
result_list = signal.receive([object_id], timeout=10)
# Expected signal is 'ErrorSignal'.
assert type(result_list[0][1]) == signal.ErrorSignal
# Print the error.
print(result_list[0][1].get_error())
Example: Sending signals between multiple actors
------------------------------------------------
This is a more involved example in which two actors ``a1`` and ``a2`` each
generate five signals, and another actor ``b`` waits to receive all signals
generated by ``a1`` and ``a2``, respectively. Note that ``b`` recursively calls
its own method ``get_signals()`` until it gets all signals it expects.
.. code-block:: python
@ray.remote
class ActorSendSignals(object):
def send_signals(self, value, count):
for i in range(count):
signal.send(UserSignal(value + str(i)))
@ray.remote
class ActorGetAllSignals(object):
def __init__(self, num_expected_signals, *source_ids):
self.received_signals = []
self.num_expected_signals = num_expected_signals
self.source_ids = source_ids
def register_handle(self, handle):
self.this_actor = handle
def get_signals(self):
new_signals = signal.receive(self.source_ids, timeout=10)
self.received_signals.extend(new_signals)
if len(self.received_signals) < self.num_expected_signals:
self.this_actor.get_signals.remote()
def get_count(self):
return len(self.received_signals)
# Create two actors to send signals.
a1 = ActorSendSignals.remote()
a2 = ActorSendSignals.remote()
signal_value = 'simple signal'
count = 5
# Each actor sends five signals.
a1.send_signals.remote(signal_value, count)
a2.send_signals.remote(signal_value, count)
# Create an actor that waits for all five signals sent by each actor.
b = ActorGetAllSignals.remote(2 * count, *[a1, a2])
# Provide actor to its own handle, so it can recursively call itself
# to get all signals from a1, and a2, respectively. This enables the actor
# execute other methods if needed.
ray.get(b.register_handle.remote(b))
b.get_signals.remote()
# Print total number of signals. This should be 2*count = 10.
print(ray.get(b.get_count.remote()))
Note
----
A failed actor (e.g., an actor that crashed) generates an error message only
when another actor or task invokes one of its methods.
Please `let us know <https://github.com/ray-project/ray/issues>`__ any issues you encounter.
doc/source/tune.rst
+3 -3
View File
@@ -1,11 +1,11 @@
Tune: A Scalable Hyperparameter Tuning Library
==============================================
Tune: Scalable Hyperparameter Tuning
====================================
.. image:: images/tune.png
:scale: 30%
:align: center
Tune is a Python library for hyperparameter tuning at any scale. Core features:
Tune is a Python library for experiment execution and hyperparameter tuning at any scale. Core features:
* Launch a multi-node `distributed hyperparameter sweep <tune-distributed.html>`_ in less than 10 lines of code.
* Supports any machine learning framework, including PyTorch, XGBoost, MXNet, and Keras. See `examples here <tune-examples.html>`_.
doc/source/using-ray.rst
+1
View File
@@ -12,6 +12,7 @@ Finally, we've also included some content on using core Ray APIs with `Tensorflo
starting-ray.rst
actors.rst
async_api.rst
using-ray-with-gpus.rst
serialization.rst
memory-management.rst
python/ray/__init__.py
+2
View File
@@ -96,6 +96,7 @@ import ray.actor # noqa: F401
from ray.actor import method # noqa: E402
from ray.runtime_context import _get_runtime_context # noqa: E402
from ray.cross_language import java_function, java_actor_class # noqa: E402
from ray import util # noqa: E402
# Ray version string.
__version__ = "0.9.0.dev0"
@@ -140,6 +141,7 @@ __all__ = [
"Language",
"java_function",
"java_actor_class",
"util",
]
# ID types
python/ray/actor.py
+3 -3
View File
@@ -483,14 +483,14 @@ class ActorClass:
# Check whether the name is already taken.
if name is not None:
try:
ray.experimental.get_actor(name)
ray.util.get_actor(name)
except ValueError: # name is not taken, expected.
pass
else:
raise ValueError(
"The name {name} is already taken. Please use "
"a different name or get existing actor using "
"ray.experimental.get_actor('{name}')".format(name=name))
"ray.util.get_actor('{name}')".format(name=name))
# Set the actor's default resources if not already set. First three
# conditions are to check that no resources were specified in the
@@ -585,7 +585,7 @@ class ActorClass:
original_handle=True)
if name is not None:
ray.experimental.register_actor(name, actor_handle)
ray.util.register_actor(name, actor_handle)
return actor_handle
python/ray/experimental/__init__.py
-20
View File
@@ -1,28 +1,8 @@
from .gcs_flush_policy import (set_flushing_policy, GcsFlushPolicy,
SimpleGcsFlushPolicy)
from .named_actors import get_actor, register_actor
from .api import get, wait
from .actor_pool import ActorPool
from .dynamic_resources import set_resource
from . import iter
def TensorFlowVariables(*args, **kwargs):
raise DeprecationWarning(
"'ray.experimental.TensorFlowVariables' is deprecated. Instead, please"
" do 'from ray.experimental.tf_utils import TensorFlowVariables'.")
__all__ = [
"TensorFlowVariables",
"get_actor",
"register_actor",
"get",
"wait",
"set_flushing_policy",
"GcsFlushPolicy",
"SimpleGcsFlushPolicy",
"set_resource",
"ActorPool",
"iter",
]
python/ray/experimental/gcs_flush_policy.py
-89
View File
@@ -1,89 +0,0 @@
import os
import time
import ray
import ray.cloudpickle as pickle
class GcsFlushPolicy:
"""Experimental: a policy to control GCS flushing.
Used by Monitor to enable automatic control of memory usage.
"""
def should_flush(self, redis_client):
"""Returns a bool, whether a flush request should be issued."""
pass
def num_entries_to_flush(self):
"""Returns an upper bound for number of entries to flush next."""
pass
def record_flush(self):
"""Must be called after a flush has been performed."""
pass
class SimpleGcsFlushPolicy(GcsFlushPolicy):
"""A simple policy with constant flush rate, after a warmup period.
Example policy values:
flush_when_at_least_bytes 2GB
flush_period_secs 10s
flush_num_entries_each_time 10k
This means: (1) If the GCS shard uses less than 2GB of memory,
no flushing would take place. This should cover most Ray runs. (2) The
GCS shard will only honor a flush request, if it's issued after 10
seconds since the last processed flush. In particular this means it's
okay for the Monitor to issue requests more frequently than this param.
(3) When processing a flush, the shard will flush at most 10k entries.
This is to control the latency of each request.
Note, flush rate == (flush period) * (num entries each time). So
applications that have a heavier GCS load can tune these params.
"""
def __init__(self,
flush_when_at_least_bytes=(1 << 31),
flush_period_secs=10,
flush_num_entries_each_time=10000):
self.flush_when_at_least_bytes = flush_when_at_least_bytes
self.flush_period_secs = flush_period_secs
self.flush_num_entries_each_time = flush_num_entries_each_time
self.last_flush_timestamp = time.time()
def should_flush(self, redis_client):
if time.time() - self.last_flush_timestamp < self.flush_period_secs:
return False
used_memory = redis_client.info("memory")["used_memory"]
assert used_memory > 0
return used_memory >= self.flush_when_at_least_bytes
def num_entries_to_flush(self):
return self.flush_num_entries_each_time
def record_flush(self):
self.last_flush_timestamp = time.time()
def serialize(self):
return pickle.dumps(self)
def set_flushing_policy(flushing_policy):
"""Serialize this policy for Monitor to pick up."""
if "RAY_USE_NEW_GCS" not in os.environ:
raise Exception(
"set_flushing_policy() is only available when environment "
"variable RAY_USE_NEW_GCS is present at both compile and run time."
)
ray.worker.global_worker.check_connected()
redis_client = ray.worker.global_worker.redis_client
serialized = pickle.dumps(flushing_policy)
redis_client.set("gcs_flushing_policy", serialized)
python/ray/experimental/sgd/__init__.py
-4
View File
@@ -1,4 +0,0 @@
from ray.experimental.sgd.pytorch import PyTorchTrainer
from ray.experimental.sgd.tf import TFTrainer
__all__ = ["PyTorchTrainer", "TFTrainer"]
python/ray/experimental/sgd/tf/__init__.py
-3
View File
@@ -1,3 +0,0 @@
from ray.experimental.sgd.tf.tf_trainer import (TFTrainer, TFTrainable)
__all__ = ["TFTrainer", "TFTrainable"]
python/ray/monitor.py
-64
View File
@@ -5,11 +5,8 @@ import time
import traceback
import json
import redis
import ray
from ray.autoscaler.autoscaler import LoadMetrics, StandardAutoscaler
import ray.cloudpickle as pickle
import ray.gcs_utils
import ray.utils
import ray.ray_constants as ray_constants
@@ -54,34 +51,6 @@ class Monitor:
self.autoscaler = None
self.autoscaling_config = None
# Experimental feature: GCS flushing.
self.issue_gcs_flushes = "RAY_USE_NEW_GCS" in os.environ
self.gcs_flush_policy = None
if self.issue_gcs_flushes:
# Data is stored under the first data shard, so we issue flushes to
# that redis server.
addr_port = self.redis.lrange("RedisShards", 0, -1)
if len(addr_port) > 1:
logger.warning(
"Monitor: "
"TODO: if launching > 1 redis shard, flushing needs to "
"touch shards in parallel.")
self.issue_gcs_flushes = False
else:
addr_port = addr_port[0].split(b":")
self.redis_shard = redis.StrictRedis(
host=addr_port[0],
port=addr_port[1],
password=redis_password)
try:
self.redis_shard.execute_command("HEAD.FLUSH 0")
except redis.exceptions.ResponseError as e:
logger.info(
"Monitor: "
"Turning off flushing due to exception: {}".format(
str(e)))
self.issue_gcs_flushes = False
def __del__(self):
"""Destruct the monitor object."""
# We close the pubsub client to avoid leaking file descriptors.
@@ -288,37 +257,6 @@ class Monitor:
ip_address += ":" + str(raylet_info["NodeManagerPort"])
self.raylet_id_to_ip_map[node_id] = ip_address
def _maybe_flush_gcs(self):
"""Experimental: issue a flush request to the GCS.
The purpose of this feature is to control GCS memory usage.
To activate this feature, Ray must be compiled with the flag
RAY_USE_NEW_GCS set, and Ray must be started at run time with the flag
as well.
"""
if not self.issue_gcs_flushes:
return
if self.gcs_flush_policy is None:
serialized = self.redis.get("gcs_flushing_policy")
if serialized is None:
# Client has not set any policy; by default flushing is off.
return
self.gcs_flush_policy = pickle.loads(serialized)
if not self.gcs_flush_policy.should_flush(self.redis_shard):
return
max_entries_to_flush = self.gcs_flush_policy.num_entries_to_flush()
num_flushed = self.redis_shard.execute_command(
"HEAD.FLUSH {}".format(max_entries_to_flush))
logger.info("Monitor: num_flushed {}".format(num_flushed))
# This flushes event log and log files.
ray.experimental.flush_redis_unsafe(self.redis)
self.gcs_flush_policy.record_flush()
def _run(self):
"""Run the monitor.
@@ -346,8 +284,6 @@ class Monitor:
if self.autoscaler:
self.autoscaler.update()
self._maybe_flush_gcs()
# Process a round of messages.
self.process_messages()
python/ray/scripts/scripts.py
+1 -1
View File
@@ -977,7 +977,7 @@ cli.add_command(project_cli)
cli.add_command(session_cli)
try:
from ray.experimental.serve.scripts import serve_cli
from ray.serve.scripts import serve_cli
cli.add_command(serve_cli)
except Exception as e:
logger.debug(
python/ray/experimental/serve/BUILD → python/ray/serve/BUILD
View File
python/ray/experimental/serve/__init__.py → python/ray/serve/__init__.py
+3 -3
View File
@@ -1,6 +1,6 @@
from ray.experimental.serve.backend_config import BackendConfig
from ray.experimental.serve.policy import RoutePolicy
from ray.experimental.serve.api import (
from ray.serve.backend_config import BackendConfig
from ray.serve.policy import RoutePolicy
from ray.serve.api import (
init, create_backend, create_endpoint, link, split, get_handle, stat,
set_backend_config, get_backend_config, accept_batch, route) # noqa: E402
python/ray/experimental/serve/api.py → python/ray/serve/api.py
+13 -14
View File
@@ -7,18 +7,17 @@ from multiprocessing import cpu_count
import numpy as np
import ray
from ray.experimental.serve.constants import (
DEFAULT_HTTP_HOST, DEFAULT_HTTP_PORT, SERVE_NURSERY_NAME)
from ray.experimental.serve.global_state import (GlobalState,
start_initial_state)
from ray.experimental.serve.kv_store_service import SQLiteKVStore
from ray.experimental.serve.task_runner import RayServeMixin, TaskRunnerActor
from ray.experimental.serve.utils import (block_until_http_ready,
get_random_letters, expand)
from ray.experimental.serve.exceptions import RayServeException
from ray.experimental.serve.backend_config import BackendConfig
from ray.experimental.serve.policy import RoutePolicy
from ray.experimental.serve.queues import Query
from ray.serve.constants import (DEFAULT_HTTP_HOST, DEFAULT_HTTP_PORT,
SERVE_NURSERY_NAME)
from ray.serve.global_state import (GlobalState, start_initial_state)
from ray.serve.kv_store_service import SQLiteKVStore
from ray.serve.task_runner import RayServeMixin, TaskRunnerActor
from ray.serve.utils import (block_until_http_ready, get_random_letters,
expand)
from ray.serve.exceptions import RayServeException
from ray.serve.backend_config import BackendConfig
from ray.serve.policy import RoutePolicy
from ray.serve.queues import Query
global_state = None
@@ -114,7 +113,7 @@ def init(kv_store_connector=None,
# Try to get serve nursery if there exists
try:
ray.experimental.get_actor(SERVE_NURSERY_NAME)
ray.util.get_actor(SERVE_NURSERY_NAME)
global_state = GlobalState()
return
except ValueError:
@@ -440,7 +439,7 @@ def get_handle(endpoint_name, relative_slo_ms=None, absolute_slo_ms=None):
global_state.route_table.list_service(include_headless=True).values())
# Delay import due to it's dependency on global_state
from ray.experimental.serve.handle import RayServeHandle
from ray.serve.handle import RayServeHandle
return RayServeHandle(global_state.init_or_get_router(), endpoint_name,
relative_slo_ms, absolute_slo_ms)
python/ray/experimental/serve/backend_config.py → python/ray/serve/backend_config.py
View File
python/ray/experimental/serve/constants.py → python/ray/serve/constants.py
View File
python/ray/experimental/serve/context.py → python/ray/serve/context.py
+1 -1
View File
@@ -1,6 +1,6 @@
from enum import IntEnum
from ray.experimental.serve.exceptions import RayServeException
from ray.serve.exceptions import RayServeException
class TaskContext(IntEnum):
python/ray/experimental/serve/examples/benchmark.py → python/ray/serve/examples/benchmark.py
+2 -2
View File
@@ -1,5 +1,5 @@
from ray.experimental import serve
from ray.experimental.serve.constants import DEFAULT_HTTP_ADDRESS
from ray import serve
from ray.serve.constants import DEFAULT_HTTP_ADDRESS
import requests
import time
import pandas as pd
python/ray/experimental/serve/examples/echo.py → python/ray/serve/examples/echo.py
+2 -2
View File
@@ -6,8 +6,8 @@ import time
import requests
from ray.experimental import serve
from ray.experimental.serve.utils import pformat_color_json
from ray import serve
from ray.serve.utils import pformat_color_json
def echo(flask_request):
python/ray/experimental/serve/examples/echo_actor.py → python/ray/serve/examples/echo_actor.py
+2 -2
View File
@@ -10,8 +10,8 @@ import time
import requests
import ray
from ray.experimental import serve
from ray.experimental.serve.utils import pformat_color_json
from ray import serve
from ray.serve.utils import pformat_color_json
class MagicCounter:
python/ray/experimental/serve/examples/echo_actor_batch.py → python/ray/serve/examples/echo_actor_batch.py
+3 -3
View File
@@ -10,9 +10,9 @@ import time
import requests
import ray
from ray.experimental import serve
from ray.experimental.serve.utils import pformat_color_json
from ray.experimental.serve import BackendConfig
from ray import serve
from ray.serve.utils import pformat_color_json
from ray.serve import BackendConfig
class MagicCounter:
python/ray/experimental/serve/examples/echo_batching.py → python/ray/serve/examples/echo_batching.py
+2 -2
View File
@@ -3,8 +3,8 @@ This example has backend which has batching functionality enabled.
"""
import ray
from ray.experimental import serve
from ray.experimental.serve import BackendConfig
from ray import serve
from ray.serve import BackendConfig
class MagicCounter:
python/ray/experimental/serve/examples/echo_error.py → python/ray/serve/examples/echo_error.py
+2 -2
View File
@@ -18,8 +18,8 @@ import time
import requests
import ray
from ray.experimental import serve
from ray.experimental.serve.utils import pformat_color_json
from ray import serve
from ray.serve.utils import pformat_color_json
def echo(_):
python/ray/experimental/serve/examples/echo_fixed_packing.py → python/ray/serve/examples/echo_fixed_packing.py
+2 -2
View File
@@ -10,8 +10,8 @@ import time
import requests
from ray.experimental import serve
from ray.experimental.serve.utils import pformat_color_json
from ray import serve
from ray.serve.utils import pformat_color_json
def echo_v1(_):
python/ray/experimental/serve/examples/echo_full.py → python/ray/serve/examples/echo_full.py
+2 -2
View File
@@ -7,8 +7,8 @@ import time
import requests
import ray
import ray.experimental.serve as serve
from ray.experimental.serve.utils import pformat_color_json
import ray.serve as serve
from ray.serve.utils import pformat_color_json
# initialize ray serve system.
# blocking=True will wait for HTTP server to be ready to serve request.
python/ray/experimental/serve/examples/echo_pipeline.py → python/ray/serve/examples/echo_pipeline.py
+1 -1
View File
@@ -2,7 +2,7 @@
Ray serve pipeline example
"""
import ray
import ray.experimental.serve as serve
import ray.serve as serve
import time
# initialize ray serve system.
python/ray/experimental/serve/examples/echo_round_robin.py → python/ray/serve/examples/echo_round_robin.py
+2 -2
View File
@@ -6,8 +6,8 @@ import time
import requests
from ray.experimental import serve
from ray.experimental.serve.utils import pformat_color_json
from ray import serve
from ray.serve.utils import pformat_color_json
def echo_v1(_):
python/ray/experimental/serve/examples/echo_slo_reverse.py → python/ray/serve/examples/echo_slo_reverse.py
+1 -1
View File
@@ -7,7 +7,7 @@ import time
import requests
import ray
import ray.experimental.serve as serve
import ray.serve as serve
# initialize ray serve system.
# blocking=True will wait for HTTP server to be ready to serve request.
python/ray/experimental/serve/examples/echo_split.py → python/ray/serve/examples/echo_split.py
+2 -2
View File
@@ -6,8 +6,8 @@ import time
import requests
from ray.experimental import serve
from ray.experimental.serve.utils import pformat_color_json
from ray import serve
from ray.serve.utils import pformat_color_json
def echo_v1(_):
python/ray/experimental/serve/exceptions.py → python/ray/serve/exceptions.py
View File
python/ray/experimental/serve/global_state.py → python/ray/serve/global_state.py
+11 -13
View File
@@ -1,19 +1,18 @@
import ray
from ray.experimental.serve.constants import (
BOOTSTRAP_KV_STORE_CONN_KEY, DEFAULT_HTTP_HOST, DEFAULT_HTTP_PORT,
SERVE_NURSERY_NAME, ASYNC_CONCURRENCY)
from ray.experimental.serve.kv_store_service import (
BackendTable, RoutingTable, TrafficPolicyTable)
from ray.experimental.serve.metric import (MetricMonitor,
start_metric_monitor_loop)
from ray.serve.constants import (BOOTSTRAP_KV_STORE_CONN_KEY,
DEFAULT_HTTP_HOST, DEFAULT_HTTP_PORT,
SERVE_NURSERY_NAME, ASYNC_CONCURRENCY)
from ray.serve.kv_store_service import (BackendTable, RoutingTable,
TrafficPolicyTable)
from ray.serve.metric import (MetricMonitor, start_metric_monitor_loop)
from ray.experimental.serve.policy import RoutePolicy
from ray.experimental.serve.server import HTTPActor
from ray.serve.policy import RoutePolicy
from ray.serve.server import HTTPActor
def start_initial_state(kv_store_connector):
nursery_handle = ActorNursery.remote()
ray.experimental.register_actor(SERVE_NURSERY_NAME, nursery_handle)
ray.util.register_actor(SERVE_NURSERY_NAME, nursery_handle)
ray.get(
nursery_handle.store_bootstrap_state.remote(
@@ -92,8 +91,7 @@ class GlobalState:
def __init__(self, actor_nursery_handle=None):
# Get actor nursery handle
if actor_nursery_handle is None:
actor_nursery_handle = ray.experimental.get_actor(
SERVE_NURSERY_NAME)
actor_nursery_handle = ray.util.get_actor(SERVE_NURSERY_NAME)
self.actor_nursery_handle = actor_nursery_handle
# Connect to all the table
@@ -126,7 +124,7 @@ class GlobalState:
return_policy = default_policy
# check if there is already a queue_actor running
# with policy as p.name for the case where
# serve nursery exists: ray.experimental.get_actor(SERVE_NURSERY_NAME)
# serve nursery exists: ray.util.get_actor(SERVE_NURSERY_NAME)
for p in RoutePolicy:
queue_actor_tag = "queue_actor::" + p.name
if queue_actor_tag in self.actor_handle_cache:
python/ray/experimental/serve/handle.py → python/ray/serve/handle.py
+5 -5
View File
@@ -1,8 +1,8 @@
from ray.experimental import serve
from ray.experimental.serve.context import TaskContext
from ray.experimental.serve.exceptions import RayServeException
from ray.experimental.serve.constants import DEFAULT_HTTP_ADDRESS
from ray.experimental.serve.request_params import RequestMetadata
from ray import serve
from ray.serve.context import TaskContext
from ray.serve.exceptions import RayServeException
from ray.serve.constants import DEFAULT_HTTP_ADDRESS
from ray.serve.request_params import RequestMetadata
class RayServeHandle:
python/ray/experimental/serve/http_util.py → python/ray/serve/http_util.py
View File
python/ray/experimental/serve/kv_store_service.py → python/ray/serve/kv_store_service.py
+3 -4
View File
@@ -1,13 +1,12 @@
import json
import sqlite3
from abc import ABC
from typing import Union
from ray import cloudpickle as pickle
import ray.experimental.internal_kv as ray_kv
from ray.experimental.serve.utils import logger
from typing import Union
from ray.experimental.serve.constants import NO_ROUTE_KEY
from ray.serve.utils import logger
from ray.serve.constants import NO_ROUTE_KEY
class NamespacedKVStore(ABC):
python/ray/experimental/serve/metric.py → python/ray/serve/metric.py
View File
python/ray/experimental/serve/policy.py → python/ray/serve/policy.py
+2 -2
View File
@@ -4,8 +4,8 @@ import itertools
import numpy as np
import ray
from ray.experimental.serve.queues import (CentralizedQueues)
from ray.experimental.serve.utils import logger
from ray.serve.queues import (CentralizedQueues)
from ray.serve.utils import logger
class RandomPolicyQueue(CentralizedQueues):
python/ray/experimental/serve/queues.py → python/ray/serve/queues.py
+1 -1
View File
@@ -13,7 +13,7 @@ import pickle
import blist
import ray
from ray.experimental.serve.utils import logger
from ray.serve.utils import logger
class Query:
python/ray/experimental/serve/request_params.py → python/ray/serve/request_params.py
+1 -1
View File
@@ -1,5 +1,5 @@
import time
from ray.experimental.serve.constants import DEFAULT_LATENCY_SLO_MS
from ray.serve.constants import DEFAULT_LATENCY_SLO_MS
class RequestMetadata:
python/ray/experimental/serve/scripts.py → python/ray/serve/scripts.py
+1 -1
View File
@@ -3,7 +3,7 @@ import json
import click
import ray
import ray.experimental.serve as serve
import ray.serve as serve
@click.group("serve", help="Commands working with ray serve")
python/ray/experimental/serve/server.py → python/ray/serve/server.py
+5 -5
View File
@@ -5,10 +5,10 @@ import uvicorn
import ray
from ray.experimental.async_api import _async_init
from ray.experimental.serve.constants import HTTP_ROUTER_CHECKER_INTERVAL_S
from ray.experimental.serve.context import TaskContext
from ray.experimental.serve.utils import BytesEncoder
from ray.experimental.serve.request_params import RequestMetadata
from ray.serve.constants import HTTP_ROUTER_CHECKER_INTERVAL_S
from ray.serve.context import TaskContext
from ray.serve.utils import BytesEncoder
from ray.serve.request_params import RequestMetadata
from urllib.parse import parse_qs
@@ -62,7 +62,7 @@ class HTTPProxy:
assert ray.is_initialized()
# Delay import due to GlobalState depends on HTTP actor
from ray.experimental.serve.global_state import GlobalState
from ray.serve.global_state import GlobalState
self.serve_global_state = GlobalState()
self.route_table_cache = dict()
python/ray/experimental/serve/task_runner.py → python/ray/serve/task_runner.py
+4 -4
View File
@@ -2,11 +2,11 @@ import time
import traceback
import ray
from ray.experimental.serve import context as serve_context
from ray.experimental.serve.context import FakeFlaskRequest
from ray.serve import context as serve_context
from ray.serve.context import FakeFlaskRequest
from collections import defaultdict
from ray.experimental.serve.utils import parse_request_item
from ray.experimental.serve.exceptions import RayServeException
from ray.serve.utils import parse_request_item
from ray.serve.exceptions import RayServeException
class TaskRunner:
python/ray/experimental/serve/tests/conftest.py → python/ray/serve/tests/conftest.py
+1 -1
View File
@@ -4,7 +4,7 @@ import tempfile
import pytest
import ray
from ray.experimental import serve
from ray import serve
@pytest.fixture(scope="session")
python/ray/experimental/serve/tests/test_api.py → python/ray/serve/tests/test_api.py
+3 -3
View File
@@ -2,10 +2,10 @@ import time
import pytest
import requests
from ray.experimental import serve
from ray.experimental.serve import BackendConfig
from ray import serve
from ray.serve import BackendConfig
import ray
from ray.experimental.serve.constants import NO_ROUTE_KEY
from ray.serve.constants import NO_ROUTE_KEY
def test_e2e(serve_instance):
python/ray/experimental/serve/tests/test_metric.py → python/ray/serve/tests/test_metric.py
+1 -1
View File
@@ -2,7 +2,7 @@ import numpy as np
import pytest
import ray
from ray.experimental.serve.metric import MetricMonitor
from ray.serve.metric import MetricMonitor
@pytest.fixture(scope="session")
python/ray/experimental/serve/tests/test_persistence.py → python/ray/serve/tests/test_persistence.py
+2 -2
View File
@@ -3,7 +3,7 @@ import subprocess
import tempfile
import ray
from ray.experimental import serve
from ray import serve
def test_new_driver(serve_instance):
@@ -11,7 +11,7 @@ def test_new_driver(serve_instance):
import ray
ray.init(address="auto")
from ray.experimental import serve
from ray import serve
serve.init()
@serve.route("/driver")
python/ray/experimental/serve/tests/test_queue.py → python/ray/serve/tests/test_queue.py
+2 -2
View File
@@ -3,10 +3,10 @@ import asyncio
import pytest
import ray
from ray.experimental.serve.policy import (
from ray.serve.policy import (
RandomPolicyQueue, RandomPolicyQueueActor, RoundRobinPolicyQueueActor,
PowerOfTwoPolicyQueueActor, FixedPackingPolicyQueueActor)
from ray.experimental.serve.request_params import RequestMetadata
from ray.serve.request_params import RequestMetadata
pytestmark = pytest.mark.asyncio
python/ray/experimental/serve/tests/test_routing.py → python/ray/serve/tests/test_routing.py
+2 -2
View File
@@ -1,8 +1,8 @@
import os
import tempfile
from ray.experimental.serve.kv_store_service import (
InMemoryKVStore, RayInternalKVStore, SQLiteKVStore)
from ray.serve.kv_store_service import (InMemoryKVStore, RayInternalKVStore,
SQLiteKVStore)
def test_default_in_memory_kv():
python/ray/experimental/serve/tests/test_serve.py → python/ray/serve/tests/test_serve.py
View File
python/ray/experimental/serve/tests/test_task_runner.py → python/ray/serve/tests/test_task_runner.py
+5 -5
View File
@@ -1,11 +1,11 @@
import pytest
import ray
import ray.experimental.serve.context as context
from ray.experimental.serve.policy import RoundRobinPolicyQueueActor
from ray.experimental.serve.task_runner import (
RayServeMixin, TaskRunner, TaskRunnerActor, wrap_to_ray_error)
from ray.experimental.serve.request_params import RequestMetadata
import ray.serve.context as context
from ray.serve.policy import RoundRobinPolicyQueueActor
from ray.serve.task_runner import (RayServeMixin, TaskRunner, TaskRunnerActor,
wrap_to_ray_error)
from ray.serve.request_params import RequestMetadata
pytestmark = pytest.mark.asyncio
python/ray/experimental/serve/tests/test_util.py → python/ray/serve/tests/test_util.py
+1 -1
View File
@@ -1,6 +1,6 @@
import json
from ray.experimental.serve.utils import BytesEncoder
from ray.serve.utils import BytesEncoder
def test_bytes_encoder():
python/ray/experimental/serve/utils.py → python/ray/serve/utils.py
+2 -2
View File
@@ -8,8 +8,8 @@ import os
import requests
from pygments import formatters, highlight, lexers
from ray.experimental.serve.context import FakeFlaskRequest, TaskContext
from ray.experimental.serve.http_util import build_flask_request
from ray.serve.context import FakeFlaskRequest, TaskContext
from ray.serve.http_util import build_flask_request
import itertools
python/ray/tests/test_actor_advanced.py
+6 -6
View File
@@ -628,22 +628,22 @@ def test_register_and_get_named_actors(ray_start_regular):
f1 = Foo.remote()
# Test saving f.
ray.experimental.register_actor("f1", f1)
ray.util.register_actor("f1", f1)
# Test getting f.
f2 = ray.experimental.get_actor("f1")
f2 = ray.util.get_actor("f1")
assert f1._actor_id == f2._actor_id
# Test same name register shall raise error.
with pytest.raises(ValueError):
ray.experimental.register_actor("f1", f2)
ray.util.register_actor("f1", f2)
# Test register with wrong object type.
with pytest.raises(TypeError):
ray.experimental.register_actor("f3", 1)
ray.util.register_actor("f3", 1)
# Test getting a nonexistent actor.
with pytest.raises(ValueError):
ray.experimental.get_actor("nonexistent")
ray.util.get_actor("nonexistent")
# Test method
assert ray.get(f1.method.remote()) == 1
@@ -682,7 +682,7 @@ ray.get(actor.ping.remote())
""".format(redis_address, actor_name)
run_string_as_driver(driver_script)
detached_actor = ray.experimental.get_actor(actor_name)
detached_actor = ray.util.get_actor(actor_name)
assert ray.get(detached_actor.ping.remote()) == "pong"
python/ray/tests/test_actor_pool.py
+1 -1
View File
@@ -2,7 +2,7 @@ import time
import pytest
import ray
from ray.experimental import ActorPool
from ray.util import ActorPool
@pytest.fixture
python/ray/tests/test_iter.py
+1 -1
View File
@@ -2,7 +2,7 @@ import time
from collections import Counter
import ray
from ray.experimental.iter import from_items, from_iterators, from_range, \
from ray.util.iter import from_items, from_iterators, from_range, \
from_actors, ParallelIteratorWorker
python/ray/tests/test_joblib.py
+2 -2
View File
@@ -16,7 +16,7 @@ from sklearn.linear_model import LogisticRegression
from sklearn.neural_network import MLPClassifier
from sklearn.model_selection import cross_val_score
from ray.experimental.joblib import register_ray
from ray.util.joblib import register_ray
import ray
@@ -28,7 +28,7 @@ def test_register_ray():
def test_ray_backend(shutdown_only):
register_ray()
from ray.experimental.joblib.ray_backend import RayBackend
from ray.util.joblib.ray_backend import RayBackend
with joblib.parallel_backend("ray"):
assert type(joblib.parallel.get_active_backend()[0]) == RayBackend
python/ray/tests/test_multi_node.py
+2 -2
View File
@@ -188,7 +188,7 @@ class Counter:
self.count += 1
return self.count
counter = Counter.remote()
ray.experimental.register_actor("Counter", counter)
ray.util.register_actor("Counter", counter)
time.sleep(100)
""".format(address)
@@ -199,7 +199,7 @@ import time
ray.init(address="{}")
while True:
try:
counter = ray.experimental.get_actor("Counter")
counter = ray.util.get_actor("Counter")
break
except ValueError:
time.sleep(1)
python/ray/tests/test_multi_node_2.py
-1
View File
@@ -61,7 +61,6 @@ def setup_monitor(address):
monitor.subscribe(ray.gcs_utils.XRAY_HEARTBEAT_BATCH_CHANNEL)
monitor.subscribe(ray.gcs_utils.XRAY_JOB_CHANNEL) # TODO: Remove?
monitor.update_raylet_map(_append_port=True)
monitor._maybe_flush_gcs()
return monitor
python/ray/util/__init__.py
+10
View File
@@ -0,0 +1,10 @@
from .named_actors import get_actor, register_actor
from .actor_pool import ActorPool
from . import iter
__all__ = [
"ActorPool",
"iter",
"get_actor",
"register_actor",
]
python/ray/experimental/actor_pool.py → python/ray/util/actor_pool.py
View File
python/ray/experimental/iter.py → python/ray/util/iter.py
+4 -4
View File
@@ -112,7 +112,7 @@ class ParallelIterator(Generic[T]):
Examples:
>>> # Applying a function over items in parallel.
>>> it = ray.experimental.iter.from_items([1, 2, 3], num_shards=2)
>>> it = ray.util.iter.from_items([1, 2, 3], num_shards=2)
... <__main__.ParallelIterator object>
>>> it = it.for_each(lambda x: x * 2).gather_sync()
... <__main__.LocalIterator object>
@@ -120,13 +120,13 @@ class ParallelIterator(Generic[T]):
... [2, 4, 6]
>>> # Creating from generators.
>>> it = ray.experimental.iter.from_iterators([range(3), range(3)])
>>> it = ray.util.iter.from_iterators([range(3), range(3)])
... <__main__.ParallelIterator object>
>>> print(list(it.gather_sync()))
... [0, 0, 1, 1, 2, 2]
>>> # Accessing the individual shards of an iterator.
>>> it = ray.experimental.iter.from_range(10, num_shards=2)
>>> it = ray.util.iter.from_range(10, num_shards=2)
... <__main__.ParallelIterator object>
>>> it0 = it.get_shard(0)
... <__main__.LocalIterator object>
@@ -138,7 +138,7 @@ class ParallelIterator(Generic[T]):
... [5, 6, 7, 8, 9]
>>> # Gathering results from actors synchronously in parallel.
>>> it = ray.experimental.iter.from_actors(workers)
>>> it = ray.util.iter.from_actors(workers)
... <__main__.ParallelIterator object>
>>> it = it.batch_across_shards()
... <__main__.LocalIterator object>
python/ray/experimental/joblib/__init__.py → python/ray/util/joblib/__init__.py
+1 -1
View File
@@ -4,7 +4,7 @@ from joblib.parallel import register_parallel_backend
def register_ray():
""" Register Ray Backend to be called with parallel_backend("ray"). """
try:
from ray.experimental.joblib.ray_backend import RayBackend
from ray.util.joblib.ray_backend import RayBackend
register_parallel_backend("ray", RayBackend)
except ImportError:
msg = ("To use the ray backend you must install ray."
python/ray/experimental/joblib/ray_backend.py → python/ray/util/joblib/ray_backend.py
+2 -2
View File
@@ -2,7 +2,7 @@ from joblib._parallel_backends import MultiprocessingBackend
from joblib.pool import PicklingPool
import logging
from ray.experimental.multiprocessing.pool import Pool
from ray.util.multiprocessing.pool import Pool
import ray
RAY_ADDRESS_ENV = "RAY_ADDRESS"
@@ -24,7 +24,7 @@ class RayBackend(MultiprocessingBackend):
"""Make Ray Pool the father class of PicklingPool. PicklingPool is a
father class that inherits Pool from multiprocessing.pool. The next
line is a patch, which changes the inheritance of Pool to be from
ray.experimental.multiprocessing.pool.
ray.util.multiprocessing.pool.
"""
PicklingPool.__bases__ = (Pool, )
"""Use all available resources when n_jobs == -1. Must set RAY_ADDRESS
python/ray/experimental/named_actors.py → python/ray/util/named_actors.py
View File
python/ray/util/sgd/__init__.py
+4
View File
@@ -0,0 +1,4 @@
from ray.util.sgd.pytorch import PyTorchTrainer
from ray.util.sgd.tf import TFTrainer
__all__ = ["PyTorchTrainer", "TFTrainer"]
python/ray/experimental/sgd/pytorch/__init__.py → python/ray/util/sgd/pytorch/__init__.py
+2 -2
View File
@@ -7,8 +7,8 @@ PyTorchTrainable = None
try:
import torch # noqa: F401
from ray.experimental.sgd.pytorch.pytorch_trainer import (PyTorchTrainer,
PyTorchTrainable)
from ray.util.sgd.pytorch.pytorch_trainer import (PyTorchTrainer,
PyTorchTrainable)
__all__ = ["PyTorchTrainer", "PyTorchTrainable"]
except ImportError:
python/ray/experimental/sgd/pytorch/distributed_pytorch_runner.py → python/ray/util/sgd/pytorch/distributed_pytorch_runner.py
+1 -1
View File
@@ -7,7 +7,7 @@ import torch.distributed as dist
import torch.utils.data
from torch.nn.parallel import DistributedDataParallel
from ray.experimental.sgd.pytorch.pytorch_runner import PyTorchRunner
from ray.util.sgd.pytorch.pytorch_runner import PyTorchRunner
logger = logging.getLogger(__name__)
python/ray/experimental/sgd/pytorch/examples/cifar_pytorch_example.py → python/ray/util/sgd/pytorch/examples/cifar_pytorch_example.py
+3 -3
View File
@@ -8,9 +8,9 @@ import torchvision
import torchvision.transforms as transforms
import ray
from ray.experimental.sgd.pytorch import (PyTorchTrainer, PyTorchTrainable)
from ray.experimental.sgd.pytorch.resnet import ResNet18
from ray.experimental.sgd.pytorch.utils import TEST_MODE
from ray.util.sgd.pytorch import (PyTorchTrainer, PyTorchTrainable)
from ray.util.sgd.pytorch.resnet import ResNet18
from ray.util.sgd.pytorch.utils import TEST_MODE
def initialization_hook(runner):
python/ray/experimental/sgd/pytorch/examples/dcgan.py → python/ray/util/sgd/pytorch/examples/dcgan.py
+3 -4
View File
@@ -15,8 +15,8 @@ from torch.nn import functional as F
from scipy.stats import entropy
import ray
from ray.experimental.sgd import PyTorchTrainer
from ray.experimental.sgd.pytorch.utils import TEST_MODE
from ray.util.sgd import PyTorchTrainer
from ray.util.sgd.pytorch.utils import TEST_MODE
# Training parameters
TRAIN_BATCHES = 5
@@ -256,8 +256,7 @@ if __name__ == "__main__":
ray.init(address=args.address)
path = os.path.dirname(ray.__file__)
model_path = os.path.join(
path, "experimental/sgd/pytorch/examples/mnist_cnn.pt")
model_path = os.path.join(path, "util/sgd/pytorch/examples/mnist_cnn.pt")
# load the pretrained mnist classification model for inception_score
trainer = train_example(
python/ray/experimental/sgd/pytorch/examples/example-sgd.yaml → python/ray/util/sgd/pytorch/examples/example-sgd.yaml
View File
python/ray/experimental/sgd/pytorch/examples/mnist_cnn.pt → python/ray/util/sgd/pytorch/examples/mnist_cnn.pt
View File
python/ray/experimental/sgd/pytorch/examples/train_example.py → python/ray/util/sgd/pytorch/examples/train_example.py
+1 -1
View File
@@ -13,7 +13,7 @@ import numpy as np
import torch
import torch.nn as nn
from ray.experimental.sgd import PyTorchTrainer
from ray.util.sgd import PyTorchTrainer
class LinearDataset(torch.utils.data.Dataset):
python/ray/experimental/sgd/pytorch/examples/tune_example.py → python/ray/util/sgd/pytorch/examples/tune_example.py
+1 -1
View File
@@ -14,7 +14,7 @@ import torch.nn as nn
import ray
from ray import tune
from ray.experimental.sgd.pytorch.pytorch_trainer import PyTorchTrainable
from ray.util.sgd.pytorch.pytorch_trainer import PyTorchTrainable
class LinearDataset(torch.utils.data.Dataset):
python/ray/experimental/sgd/pytorch/pytorch_runner.py → python/ray/util/sgd/pytorch/pytorch_runner.py
+2 -2
View File
@@ -8,8 +8,8 @@ import torch.utils.data
from torch.utils.data import Dataset
import ray
from ray.experimental.sgd.pytorch import utils as pytorch_utils
from ray.experimental.sgd import utils
from ray.util.sgd.pytorch import utils as pytorch_utils
from ray.util.sgd import utils
logger = logging.getLogger(__name__)
amp = None
python/ray/experimental/sgd/pytorch/pytorch_trainer.py → python/ray/util/sgd/pytorch/pytorch_trainer.py
+4 -4
View File
@@ -11,11 +11,11 @@ import ray
from ray.tune import Trainable
from ray.tune.trial import Resources
from ray.experimental.sgd.pytorch.distributed_pytorch_runner import (
from ray.util.sgd.pytorch.distributed_pytorch_runner import (
DistributedPyTorchRunner)
from ray.experimental.sgd import utils
from ray.experimental.sgd.pytorch.pytorch_runner import PyTorchRunner
from ray.experimental.sgd.pytorch import utils as pytorch_utils
from ray.util.sgd import utils
from ray.util.sgd.pytorch.pytorch_runner import PyTorchRunner
from ray.util.sgd.pytorch import utils as pytorch_utils
logger = logging.getLogger(__name__)
RESIZE_COOLDOWN_S = 10

Some files were not shown because too many files have changed in this diff Show More