Improve list and collection materializers performance #3375
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
This commit optimizes the way collections are materialized by: 1. Grouping elements by type to reduce overhead 2. Adding version metadata for backward compatibility 3. Optimizing the load process to handle both old and new formats Fixes #3371 🤖 Generated with Claude Code Co-Authored-By: Claude <[email protected]>
strickvl
added
enhancement
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the You can disable this status message by setting the 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? 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
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)
Other keywords and placeholders
Documentation and Community
|
This implements a much more efficient approach to materializing collections by: 1. Grouping elements by type and serializing them in batches 2. Using compression (gzip + pickle) to reduce storage requirements 3. Implementing chunking for handling very large collections 4. Adding efficient chunk-based loading with caching The v3 format significantly reduces I/O operations to artifact stores, dramatically improving performance especially for cloud storage backends. 🤖 Generated with Claude Code Co-Authored-By: Claude <[email protected]>
This fixes an issue where the batch compression implementation was using Python's built-in gzip.open() directly with cloud storage URIs (like S3), which fails because the built-in functions only support local file paths. The solution: 1. Use ZenML's fileio utilities which support various backends 2. Implement in-memory compression/decompression using BytesIO buffers 3. Properly handle reading and writing to cloud storage The changes ensure our optimized collection materialization works properly with all storage backends, including S3, GCS, and Azure Blob Storage. 🤖 Generated with Claude Code Co-Authored-By: Claude <[email protected]>
- Removed unused variables (save_time, load_time) in test functions - Replaced with inline calculations that don't store results - Fixed ruff F841 (local variable assigned but never used) errors - Ran format.sh to ensure code follows project style 🤖 Generated with Claude Code Co-Authored-By: Claude <[email protected]>
…mentation - Add configurable chunk size via ZENML_MATERIALIZER_COLLECTION_CHUNK_SIZE env var - Implement adaptive sizing logic for very large elements - Update documentation to highlight optimization features - Add tests for both default and adaptive sizing - Add information to environment variables and big data docs 🤖 Generated with Claude Code Co-Authored-By: Claude <[email protected]>
schustmi
requested changes
Feb 25, 2025
| with gzip.GzipFile( | ||
| fileobj=f_buffer, mode="wb", compresslevel=6 | ||
| ) as f_gzip: | ||
| pickle.dump( |
There was a problem hiding this comment.
I think this needs a major rework: This code is not using ZenML materializers anymore and instead simply uses pickle for everything. If that was a good solution, instead of doing all this chunking we could just tell people to pickle the whole collection.
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.
Optimized Collection Materializers with Batch Compression
This PR significantly improves the performance of list and collection materializers, addressing a critical bottleneck in ZenML's artifact handling system.
Initial Improvements
Major Batch Compression Enhancement
Technical Implementation
Performance Impact
The impact on performance is substantial:
Configuration Options
Compatibility
This change significantly improves user experience when working with large collections, especially in cloud environments where storage operations have higher latency.
Fixes #3371
🤖 Generated with Claude Code
Co-Authored-By: Claude [email protected]