Repository navigation
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
…wo relatively clean approaches to choose from.
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## main #10 +/- ##
============================================
- Coverage 100.00% 26.92% -73.08%
============================================
Files 2 5 +3
Lines 7 182 +175
============================================
+ Hits 7 49 +42
- Misses 0 133 +133 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
…naries of values. Here we expect a model that returns a dict to have keys that match the columns in PRED_META.
There was a problem hiding this comment.
🟡 Changes recommended
Nested-input handling, CPU initialization, and shared or stale dataset state can prevent inference or produce incorrect predictions.
4 open issues
What changed in this PR
This PR moves Hyrax inference boilerplate from a proof-of-concept notebook into reusable LSDB integration APIs, supporting both managed clients and existing Dask clients.
Changes:
- Adds
ml_client, worker-plugin registration, and partition-level prediction. - Introduces a dataframe dataset adapter with nested-field getters and padded batching.
- Adds usage notebooks and example model assets, updates dependencies, and removes template tests.
| File | Description |
|---|---|
tests/lsdb_ml/test_example_module.py |
Removes template example tests. |
src/lsdb_ml/hyrax/worker_plugin.py |
Adds model loading, plugin registration, and partition inference. |
src/lsdb_ml/hyrax/datasets/hyrax_dataframe_dataset.py |
Adapts dataframe fields and nested arrays for Hyrax. |
src/lsdb_ml/hyrax/client.py |
Adds managed Dask client creation and cleanup. |
src/lsdb_ml/__init__.py |
Exports ml_client and predict. |
pyproject.toml |
Makes LSDB and Hyrax required dependencies. |
docs/pre_executed/original_lsdb_style_api_infer.ipynb |
Preserves the original inference prototype. |
docs/pre_executed/loopback_weights/prepare_inputs.py |
Supplies input preparation for the loopback example. |
docs/pre_executed/loopback_weights/fake_weights.pth |
Adds an example weights asset. |
docs/pre_executed/goal_lsdb_style_api_infer.ipynb |
Demonstrates managed and explicit plugin-registration workflows. |
Review effort: Balanced
|
|
||
| for col in nested_cols: | ||
| meta_dict[col] = {} | ||
| for ncols in nested_frame[col].columns: |
Comment on lines
+71
to
+77
| data_request = {"infer": {"data": data}} | ||
| self.config["data_request"] = data_request | ||
|
|
||
| if not self.dataset_prepped: | ||
| datasets = setup_dataset(self.config, shuffle=False) | ||
| self.dataset = datasets["infer"] | ||
| self.dataset_prepped = True |
| # for inference. | ||
| # NOTE: `dataset` is a Hyrax DataProvider passed to `dist_data_loader`, | ||
| # not the bare dataset class. | ||
| dataset.prepped_datasets["data"].data = df |
Comment on lines
+50
to
+51
| device = torch.accelerator.current_accelerator() | ||
| self.process_fn = create_process_func("infer_batch", device.type, self.model, self.config) |
This file contains hidden or 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
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.
The goal of the PR is to start with the POC notebook that showed use of a WorkerPlugin to create a model instance on each dask worker for running inference and abstract away as much of the boilerplate code as possible.
The notebook
goal_lsdb_style_api_infer.ipynbrepresents (what I think is) the best case API.From that notebook there are two example paths shows:
ml_clientcreate and add the Hyrax WorkerPlugin to the clients. Then the user callscatalog.map_partitions(predict, ...).map_partitions(predict, ...)call - register the WorkerPlugin with clients that are already instantiate, then call.map_partitions(predict, ...). This approach might allow for easier subsequent UDFs to be run after data has been processed through the model.