[REP] Execution Optimizer for Ray Datasets #19
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: Cheng Su <scnju13@gmail.com>
Signed-off-by: Cheng Su <scnju13@gmail.com>
Signed-off-by: Cheng Su <scnju13@gmail.com>
ericl
reviewed
Dec 15, 2022
reps/2022-12-15-optimizer-data.md
Outdated
|
|
||
| Architecture after REP: | ||
|
|
||
| <img width="945" alt="new-architecture" src="https://user-images.githubusercontent.com/4629931/207807703-bb65db63-41a0-41d9-8e7b-154e1a0ed565.png"> |
There was a problem hiding this comment.
Since we are considering re-optimization outside the scope of this REP, can we also remove that from the diagram?
reps/2022-12-15-optimizer-data.md
Outdated
|
|
||
| #### 3.2.1. Interfaces | ||
|
|
||
| NOTE: `OneToOneOperator` used here is the same as `OneToOneOperator` in "Native pipelining support in Ray Datasets" REP. |
There was a problem hiding this comment.
This section needs to be updated, since the other REP now only proposes PhysicalOperator.
There was a problem hiding this comment.
@ericl - yeah updated. I need more thought to hook up BatchedOperator.process_batches with PhysicalOperator.add_input/inputs_done/has_next/get_next. But I think it should be implementation detail that we can figure it out later.
stephanie-wang
requested changes
Dec 16, 2022
reps/2022-12-15-optimizer-data.md
Outdated
|
|
||
| ## Summary | ||
|
|
||
| Build the breakthrough foundation to tackle a series of fundamental issues around Ray Data. The foundation is (1) **lazy execution**, (2) **optimizer**, and (3) **vectorized execution with data batch**. |
There was a problem hiding this comment.
The summary is a bit low-level right now and solution-heavy. It might be good to focus more on the problems (expensive and unnecessary materialization, current design lacks an optimizer which makes materialization impossible to elide).
There was a problem hiding this comment.
Moved this under General Motivation to make it more coherent, as the top-level summary seem not strictly needed (not see in other REPs).
13 tasks
ericl
pushed a commit
to ray-project/ray
that referenced
this pull request
Jan 6, 2023
This PR is to enable lazy execution by default. See ray-project/enhancements#19 for motivation. The change includes: * Change `Dataset` constructor: `Dataset.__init__(lazy: bool = True)`. Also remove `defer_execution` field, as it's no longer needed. * `read_api.py:read_datasource()` returns a lazy `Dataset` with computing the first input block. * Add `ds.fully_executed()` calls to required unit tests, to make sure they are passing. TODO: - [x] Fix all unit tests - [x] #31459 - [x] #31460 - [ ] Remove the behavior to eagerly compute first block for read - [ ] #31417 - [ ] Update documentation
Signed-off-by: Cheng Su <scnju13@gmail.com>
stephanie-wang
approved these changes
Jan 9, 2023
Signed-off-by: Cheng Su <scnju13@gmail.com>
AmeerHajAli
pushed a commit
to ray-project/ray
that referenced
this pull request
Jan 12, 2023
This PR is to enable lazy execution by default. See ray-project/enhancements#19 for motivation. The change includes: * Change `Dataset` constructor: `Dataset.__init__(lazy: bool = True)`. Also remove `defer_execution` field, as it's no longer needed. * `read_api.py:read_datasource()` returns a lazy `Dataset` with computing the first input block. * Add `ds.fully_executed()` calls to required unit tests, to make sure they are passing. TODO: - [x] Fix all unit tests - [x] #31459 - [x] #31460 - [ ] Remove the behavior to eagerly compute first block for read - [ ] #31417 - [ ] Update documentation
tamohannes
pushed a commit
to ju2ez/ray
that referenced
this pull request
Jan 25, 2023
This PR is to enable lazy execution by default. See ray-project/enhancements#19 for motivation. The change includes: * Change `Dataset` constructor: `Dataset.__init__(lazy: bool = True)`. Also remove `defer_execution` field, as it's no longer needed. * `read_api.py:read_datasource()` returns a lazy `Dataset` with computing the first input block. * Add `ds.fully_executed()` calls to required unit tests, to make sure they are passing. TODO: - [x] Fix all unit tests - [x] ray-project#31459 - [x] ray-project#31460 - [ ] Remove the behavior to eagerly compute first block for read - [ ] ray-project#31417 - [ ] Update documentation Signed-off-by: tmynn <hovhannes.tamoyan@gmail.com>
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.
This REP introduces (1) lazy execution, (2) optimizer, and (3) vectorized execution with data batch, to improve user experience and performance for Ray Datasets.