Added support for batch size and multi processing to evaluate_detections #5376

minhtuev · 2025-01-11T01:09:45Z

What changes are proposed in this pull request?

This PR introduces parallel processing capabilities to the detection evaluation functionality in FiftyOne. Key changes include:

Implemented parallel processing of samples using ProcessPoolExecutor or ThreadPoolExecutor
Added configurable number of workers (defaults to CPU count - 1)
Maintained proper batch-based sample downloading through FiftyOne's iter_samples(), with batch_size exposed as an argument

The parallelization significantly improves performance for large datasets while maintaining the same accuracy and output format as the original

How is this patch tested? If it is not, please explain why.

Local testing

Release Notes

Is this a user-facing change that should be mentioned in the release notes?

No. You can skip the rest of this section.
Yes. Give a description of this change to be included in the release
notes for FiftyOne users.

(Details in 1-2 sentences. You can just refer to another PR with a description
if this PR is part of a larger change.)

What areas of FiftyOne does this PR affect?

App: FiftyOne application changes
Build: Build and test infrastructure changes
Core: Core fiftyone Python library changes
Documentation: FiftyOne documentation changes
Other

coderabbitai · 2025-01-11T01:09:53Z

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Thank you for using CodeRabbit. We offer it for free to the OSS community and would appreciate your support in helping us grow. If you find it useful, would you consider giving us a shout-out on your favorite social media?

❤️ Share

🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

Review comments: Directly reply to a review comment made by CodeRabbit. Example:
- I pushed a fix in commit <commit_id>, please review it.
- Generate unit testing code for this file.
- Open a follow-up GitHub issue for this discussion.
Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
- @coderabbitai generate unit testing code for this file.
- @coderabbitai modularize this function.
PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
- @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
- @coderabbitai read src/utils.ts and generate unit testing code.
- @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
- @coderabbitai help me debug CodeRabbit configuration file.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

@coderabbitai pause to pause the reviews on a PR.
@coderabbitai resume to resume the paused reviews.
@coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
@coderabbitai full review to do a full review from scratch and review all the files again.
@coderabbitai summary to regenerate the summary of the PR.
@coderabbitai generate docstrings to generate docstrings for this PR. (Beta)
@coderabbitai resolve resolve all the CodeRabbit review comments.
@coderabbitai configuration to show the current CodeRabbit configuration for the repository.
@coderabbitai help to get help.

Other keywords and placeholders

Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
Add @coderabbitai anywhere in the PR title to generate the title automatically.

Documentation and Community

Visit our Documentation for detailed information on how to use CodeRabbit.
Join our Discord Community to get help, request features, and share feedback.
Follow us on X/Twitter for updates and announcements.

fiftyone/utils/eval/detection.py

kaixi-wang · 2025-01-11T15:11:42Z

How did you test this/on what platform? Pymongo is not fork safe and since fiftyone uses a single global mongo client, iterating over samples could result in deadlock. https://www.mongodb.com/docs/languages/python/pymongo-driver/current/faq/#can-i-use-pymongo-with-multiprocessing-

swheaton · 2025-01-12T00:30:22Z

Also we are pausing this for now as discussed the other day 😊

minhtuev · 2025-01-13T18:50:33Z

@kaixi-wang : I am running this locally on my mac computer against local & remote DB. The samples are fetched on the main thread and the actual IOU computation is parallelized, so I don't think there is a danger to Pymongo.

minhtuev · 2025-01-13T18:53:23Z

fiftyone/utils/eval/detection.py

+            futures.append((future, sample))
+
+        # Collect results
+        for future, sample in futures:


We probably should define a constant somewhere so that we can collect the results earlier without having to maintain a large list

minhtuev · 2025-01-15T23:28:01Z

Related PR:
https://github.com/voxel51/fiftyone/pull/5362/files

brimoor · 2025-01-24T04:14:09Z

fiftyone/utils/eval/detection.py

@@ -135,6 +139,9 @@ def evaluate_detections(
        progress (None): whether to render a progress bar (True/False), use the
            default value ``fiftyone.config.show_progress_bars`` (None), or a
            progress callback function to invoke instead
+        batch_size (None): the batch size at which to process samples. By
+            default, all samples are processed in a single (1) batch


It's not true that the default is a batch size of 1. iter_samples(autosave=True) uses a batching strategy by default:

fiftyone/fiftyone/core/view.py

Lines 500 to 516 in 67bdf7e

batch_size (None): the batch size to use when autosaving samples.

If a ``batching_strategy`` is provided, this parameter

configures the strategy as described below. If no

``batching_strategy`` is provided, this can either be an

integer specifying the number of samples to save in a batch

(in which case ``batching_strategy`` is implicitly set to

``"static"``) or a float number of seconds between batched

saves (in which case ``batching_strategy`` is implicitly set to

``"latency"``)

batching_strategy (None): the batching strategy to use for each

save operation when autosaving samples. Supported values are:

- ``"static"``: a fixed sample batch size for each save

- ``"size"``: a target batch size, in bytes, for each save

- ``"latency"``: a target latency, in seconds, between saves

By default, ``fo.config.default_batcher`` is used

The default batcher sends requests every 0.2 seconds:

fiftyone/fiftyone/core/config.py

Lines 159 to 182 in 67bdf7e

self.default_batcher = self.parse_string(

d,

"default_batcher",

env_var="FIFTYONE_DEFAULT_BATCHER",

default="latency",

)

self.batcher_static_size = self.parse_int(

d,

"batcher_static_size",

env_var="FIFTYONE_BATCHER_STATIC_SIZE",

default=100,

)

self.batcher_target_size_bytes = self.parse_int(

d,

"batcher_target_size_bytes",

env_var="FIFTYONE_BATCHER_TARGET_SIZE_BYTES",

default=2**20,

)

self.batcher_target_latency = self.parse_number(

d,

"batcher_target_latency",

env_var="FIFTYONE_BATCHER_TARGET_LATENCY",

default=0.2,

)

Ah, so if batch_size is not passed, we use latency strategy by default 👍

Added support for batch size and multi processing

21c0fb9

minhtuev requested a review from swheaton January 11, 2025 01:09

minhtuev changed the title ~~Added support for batch size and multi processing~~ Added support for batch size and multi processing to evaluate_detections Jan 11, 2025

minhtuev commented Jan 11, 2025

View reviewed changes

fiftyone/utils/eval/detection.py Outdated Show resolved Hide resolved

minhtuevo added 2 commits January 10, 2025 18:05

Cleaned up

f9abb61

Using threadpool executor to avoid pickling mongo obj

7ba6a45

minhtuev force-pushed the minh/batch-size-mp branch from e49f1ea to 7ba6a45 Compare January 11, 2025 02:44

minhtuev commented Jan 13, 2025

View reviewed changes

minhtuev mentioned this pull request Jan 15, 2025

[WIP] Batching evaluate_detections #5361

Closed

7 tasks

brimoor reviewed Jan 24, 2025

View reviewed changes

Update docs

3a73c52

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Added support for batch size and multi processing to evaluate_detections #5376

Added support for batch size and multi processing to evaluate_detections #5376

minhtuev commented Jan 11, 2025

coderabbitai bot commented Jan 11, 2025

Review skipped

Chat

CodeRabbit Commands (Invoked using PR comments)

Other keywords and placeholders

Documentation and Community

kaixi-wang commented Jan 11, 2025

swheaton commented Jan 12, 2025

minhtuev commented Jan 13, 2025

minhtuev Jan 13, 2025 •

edited

Loading

minhtuev commented Jan 15, 2025

brimoor Jan 24, 2025

minhtuev Jan 24, 2025

	batch_size (None): the batch size to use when autosaving samples.
	If a ``batching_strategy`` is provided, this parameter
	configures the strategy as described below. If no
	``batching_strategy`` is provided, this can either be an
	integer specifying the number of samples to save in a batch
	(in which case ``batching_strategy`` is implicitly set to
	``"static"``) or a float number of seconds between batched
	saves (in which case ``batching_strategy`` is implicitly set to
	``"latency"``)
	batching_strategy (None): the batching strategy to use for each
	save operation when autosaving samples. Supported values are:

	- ``"static"``: a fixed sample batch size for each save
	- ``"size"``: a target batch size, in bytes, for each save
	- ``"latency"``: a target latency, in seconds, between saves

	By default, ``fo.config.default_batcher`` is used

	self.default_batcher = self.parse_string(
	d,
	"default_batcher",
	env_var="FIFTYONE_DEFAULT_BATCHER",
	default="latency",
	)
	self.batcher_static_size = self.parse_int(
	d,
	"batcher_static_size",
	env_var="FIFTYONE_BATCHER_STATIC_SIZE",
	default=100,
	)
	self.batcher_target_size_bytes = self.parse_int(
	d,
	"batcher_target_size_bytes",
	env_var="FIFTYONE_BATCHER_TARGET_SIZE_BYTES",
	default=2**20,
	)
	self.batcher_target_latency = self.parse_number(
	d,
	"batcher_target_latency",
	env_var="FIFTYONE_BATCHER_TARGET_LATENCY",
	default=0.2,
	)

Added support for batch size and multi processing to evaluate_detections #5376

Are you sure you want to change the base?

Added support for batch size and multi processing to evaluate_detections #5376

Conversation

minhtuev commented Jan 11, 2025

What changes are proposed in this pull request?

How is this patch tested? If it is not, please explain why.

Release Notes

Is this a user-facing change that should be mentioned in the release notes?

What areas of FiftyOne does this PR affect?

coderabbitai bot commented Jan 11, 2025

Review skipped

Chat

CodeRabbit Commands (Invoked using PR comments)

Other keywords and placeholders

Documentation and Community

kaixi-wang commented Jan 11, 2025

swheaton commented Jan 12, 2025

minhtuev commented Jan 13, 2025

minhtuev Jan 13, 2025 • edited Loading

Choose a reason for hiding this comment

minhtuev commented Jan 15, 2025

brimoor Jan 24, 2025

Choose a reason for hiding this comment

minhtuev Jan 24, 2025

Choose a reason for hiding this comment

minhtuev Jan 13, 2025 •

edited

Loading