[core][experimental] Add experimental compiled DAG #41769
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
3 tasks
ericl
approved these changes
Dec 10, 2023
python/ray/dag/class_node.py
Outdated
| def __init__(self, actor: ClassNode, method_name: str, options: dict): | ||
| # TODO(sang): Theoretically, We should use weakref cuz it is | ||
| # a circular dependency but when I used weakref, it fails | ||
| # because we cannot serialize the weakref. |
There was a problem hiding this comment.
Explain the concern about circular dependencies here?
There was a problem hiding this comment.
you can add a simple wrapper for weakref to address the serialization issue if it's needed. I did the same in a patch to the pydantic repo: pydantic/pydantic#7780
ericl
added
the
@author-action-required
rkooo567
approved these changes
Dec 11, 2023
rkooo567
approved these changes
Dec 11, 2023
edoakes
reviewed
Dec 11, 2023
| logger = logging.getLogger(__name__) | ||
|
|
||
|
|
||
| @ray.remote(concurrency_groups={"_ray_system": 1}) |
There was a problem hiding this comment.
What is the _ray_system concurrency group and why's it required here?
There was a problem hiding this comment.
We shouldn't need to declare this group, it's automatically created per #41605
ericl
reviewed
Dec 12, 2023
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
8 tasks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Why are these changes needed?
This adds an experimental interface that allows you to compile a normal ray.dag to use a faster execution path. The execution path uses ray.experimental.channel to pass data directly between actors instead of going through the normal Ray control plane. See the REP for more information.
Several limitations will be addressed in a follow-up PR:
dag.executereturns a Channel instead of an ObjectRef. Reading from a Channel requires the caller to release the value as soon as they are done with it, to avoid blocking the next DAG execution.Initial results show about 35x speedup compared to normal Ray core on simple single-node DAGs: