Update UMAP notebook, test precomputed_knn, write data to parquet #813
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
Collaborator
rishic3
commented
Dec 25, 2024
rishic3
commented
- Updated UMAP notebook run with Pyspark+Jupyter, aligning with instructions.
- Added test/functionality for precomputed_knn.
- Updated model writer/reader to save array attributes to parquet via Spark DF.
- Lowered error thresholds to track perf regressions based on empirical tests (bumped to +3 stdevs above mean error)
Signed-off-by: Rishi Chandra <[email protected]>
|
build |
|
build |
eordentlich
reviewed
Dec 30, 2024
python/src/spark_rapids_ml/umap.py
Outdated
| StructField("shape", ArrayType(IntegerType(), False), False), | ||
| ] | ||
| ) | ||
| data_df = spark.createDataFrame( |
There was a problem hiding this comment.
By putting all data in a single row, what are the limits on data sizes? Here and below for dense? Should add these limits to documentation.
Alternative is to create regular possibly multiple dataframes, save in multiple files/subfolders and add a column to order/sort by to restore original order.
| num_workers=gpu_number, | ||
| metric="sqeuclidean", | ||
| random_state=random_state, | ||
| precomputed_knn=precomputed_knn, |
There was a problem hiding this comment.
Can we convert back to numpy to emphasize that numpy ok here in keeping with our no need for gpu on driver?
|
build |
eordentlich
reviewed
Dec 31, 2024
python/src/spark_rapids_ml/umap.py
Outdated
| def write_sparse_array(array: scipy.sparse.spmatrix, df_dir: str) -> None: | ||
| indptr_df = spark.createDataFrame(array.indptr, schema=["indptr"]) | ||
| indices_data_df = spark.createDataFrame( | ||
| pd.DataFrame({"indices": array.indices, "data": array.data}) |
There was a problem hiding this comment.
Better to add the row_id here in the pd dataframe in case it is shuffled when creating spark dataframe.
|
build |
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.