[auto-merge] release/26.06 to main [skip ci] [bot]

[nvauto]

auto-merge triggered by github actions on release/26.06 to create a PR keeping main up-to-date. If this PR is unable to be merged due to conflicts, it will remain open until manually fix.

[nvauto]

SUCCESS - auto-merge

[greptile-apps]

Greptile Summary

This PR merges release/26.06 into main via an automated bot merge. The only code change extends two conditional checks in umap.py to also trigger when cuda_system_mem_enabled is set, matching the same pattern already used earlier in the same function.

• The fix adds or cuda_system_mem_enabled to the two yield-site conditions in _train_udf (lines 1259 and 1284), ensuring .get() is called to copy CuPy GPU arrays back to host memory before they are wrapped into a pandas DataFrame — previously this copy only happened when cuda_managed_mem_enabled was set, leaving system-memory mode without a proper device-to-host transfer.
• The change is consistent with an identical guard at line 1203 in the same function.

Confidence Score: 5/5

Safe to merge — the two-line change closes a gap where CuPy arrays were not transferred to host memory before being wrapped in a pandas DataFrame when running in system-memory mode.

Both changed lines mirror an identical guard that already exists at line 1203 in the same function, making the fix well-precedented. The else branches (no .get()) are only reached when data is already on the host, so the logic remains correct in all four memory-mode combinations.

No files require special attention.

Important Files Changed

Filename	Overview
python/src/spark_rapids_ml/umap.py	Two conditions extended to include cuda_system_mem_enabled, ensuring .get() is called to transfer CuPy arrays from GPU to host before yielding pandas DataFrames; consistent with the existing guard at line 1203.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[_train_udf: iterate embedding chunks] --> B{use_sparse_array?}
    B -- yes --> C[slice csr_chunk, indices, indptr, data]
    B -- no --> D{cuda_managed_mem_enabled\nor cuda_system_mem_enabled?}
    C --> E{cuda_managed_mem_enabled\nor cuda_system_mem_enabled?}
    E -- yes --> F["yield DataFrame with .get()\n(GPU → host transfer)"]
    E -- no --> G["yield DataFrame without .get()\n(data already on host)"]
    D -- yes --> H["yield DataFrame with .get()\n(GPU → host transfer)"]
    D -- no --> I["yield DataFrame without .get()\n(data already on host)"]

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
flowchart TD
    A[_train_udf: iterate embedding chunks] --> B{use_sparse_array?}
    B -- yes --> C[slice csr_chunk, indices, indptr, data]
    B -- no --> D{cuda_managed_mem_enabled\nor cuda_system_mem_enabled?}
    C --> E{cuda_managed_mem_enabled\nor cuda_system_mem_enabled?}
    E -- yes --> F["yield DataFrame with .get()\n(GPU → host transfer)"]
    E -- no --> G["yield DataFrame without .get()\n(data already on host)"]
    D -- yes --> H["yield DataFrame with .get()\n(GPU → host transfer)"]
    D -- no --> I["yield DataFrame without .get()\n(data already on host)"]

Loading

_{Reviews (1): Last reviewed commit: "assume cupy for sam when returning embed..." | Re-trigger Greptile}