Skip to content

[DOC] update cuda13 related jars in download doc [skip ci]#14027

Closed
nvliyuan wants to merge 83 commits into
NVIDIA:release/25.12from
nvliyuan:download-doc-cuda13
Closed

[DOC] update cuda13 related jars in download doc [skip ci]#14027
nvliyuan wants to merge 83 commits into
NVIDIA:release/25.12from
nvliyuan:download-doc-cuda13

Conversation

@nvliyuan

Copy link
Copy Markdown
Collaborator

Fix #14019
Add cuda13 jars on download page.

wjxiz1992 and others added 30 commits November 17, 2025 10:27
…he Spark Suites (NVIDIA#13790)

Fixes NVIDIA#13789 

## Description

### Problem
When multiple developers simultaneously add new test suites to the
Spark-RAPIDS test framework, they often encounter merge conflicts in the
`RapidsTestSettings.scala` file, specifically in the import section.
With 30+ explicit imports for test suites, every new suite addition
requires modifying the same file region, leading to frequent conflicts
and slowing down the development workflow.

### Solution
This PR refactors the import section in `RapidsTestSettings.scala` by
replacing 30 explicit imports with a single wildcard import:

// Before (30 lines)
import org.apache.spark.sql.rapids.suites.RapidsCastSuite
import org.apache.spark.sql.rapids.suites.RapidsDataFrameAggregateSuite
// ... 28 more imports

// After (2 lines)
// Import all Rapids test suites to avoid merge conflicts when multiple
people add new suites
import org.apache.spark.sql.rapids.suites._### Benefits
- **Reduces merge conflicts**: When adding new test suites, developers
only need to modify the `enableSuite` configuration section, not the
import section
- **Simplifies maintenance**: No need to update imports when adding new
test suites
- **Cleaner code**: Reduces code from 30 lines to 2 lines (including
comment)
- **No behavioral changes**: All existing functionality remains
unchanged

### User Experience
After this change, when developers add new test suites:
1. Create the test suite class in `org.apache.spark.sql.rapids.suites`
package
2. Add the `enableSuite[YourNewSuite]` configuration (no import
statement needed)
3. Reduced likelihood of merge conflicts during PR review

### Checklists
- [ ] This PR has added documentation for new or modified features or
behaviors.
- [ ] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

Signed-off-by: Allen Xu <allxu@nvidia.com>
Keep the rapids JNI, private and hybrid dependency version at 25.12
until the nightly CI for 26.02 is complete. Track the dependency update
process at: NVIDIA#13794

Signed-off-by: nvauto <70000568+nvauto@users.noreply.github.com>
Add RapidsDateExpressionsSuite.
Excluded 6 test cases.
- Filed 4 plugin issues.
- Filed 1 UT framework issue.
- one case is blocked by Gpu UT limitation, not able to intercept
exception. I disabled the original test case and created a new one which
is the subset of the original one.

---------

Signed-off-by: Gary Shen <gashen@nvidia.com>
…IA#13767)

## Description

This PR migrates `DataFrameComplexTypeSuite` tests from Apache Spark to
RAPIDS for GPU execution testing.

## Changes

- Created `RapidsDataFrameComplexTypeSuite` extending
`DataFrameComplexTypeSuite` with `RapidsSQLTestsBaseTrait`
- Registered test suite in `RapidsTestSettings.scala`

## Test Results

- **Total tests**: 7
- **Passing**: 7 (100%)
- **Excluded**: 0
- **Perfect GPU compatibility**: All tests pass without modification

## Test Coverage

All complex data type operations work correctly on GPU:

1. **UDF on struct** - User-defined functions on struct columns
2. **UDF on named_struct** - UDF on named_struct expressions  
3. **UDF on array** - User-defined functions on array columns
4. **UDF on map** - User-defined functions on map columns
5. **SPARK-12477: accessing null element in array field** - Null
handling in nested arrays
6. **SPARK-15285: Generated SpecificSafeProjection.apply method grows
beyond 64KB** - Large projection handling
7. **SPARK-29503: nest unsafe struct inside safe array** - Mixed
safe/unsafe nested structures

## Key Observations

- ✅ **100% pass rate** - All tests pass without modification
- ✅ **No exclusions needed** - Perfect compatibility
- ✅ **Complex nested types** work correctly on GPU (arrays, structs,
maps)
- ✅ **UDFs on complex types** execute successfully on GPU
- ✅ **Edge cases** (null handling, large projections, mixed safe/unsafe)
all work

## Files Changed

-
`tests/src/test/spark330/scala/org/apache/spark/sql/rapids/suites/RapidsDataFrameComplexTypeSuite.scala`
(new)
-
`tests/src/test/spark330/scala/org/apache/spark/sql/rapids/utils/RapidsTestSettings.scala`
(updated)

## Related Issues

Closes part of NVIDIA#11297 (Spark UT migration tracking issue)

Signed-off-by: Allen Xu <allxu@nvidia.com>

---------

Signed-off-by: Allen Xu <allxu@nvidia.com>
…kip ci] (NVIDIA#13796)

To fix NVIDIA#13794

Wait for the pre-merge CI job to SUCCEED

Signed-off-by: nvauto <70000568+nvauto@users.noreply.github.com>
auto-merge triggered by github actions on `release/25.12` 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.
…IA#13777)

## Description

This PR migrates `DataFrameNaFunctionsSuite` tests from Apache Spark to
RAPIDS for GPU execution testing.

## Changes

- Created `RapidsDataFrameNaFunctionsSuite` extending
`DataFrameNaFunctionsSuite` with `RapidsSQLTestsBaseTrait`
- Registered test suite in `RapidsTestSettings.scala` with no exclusions

## Test Results

- **Total tests**: 27
- **Passing**: 27 (100%)
- **Excluded**: 0
- **Perfect GPU compatibility**: 100% pass rate ✨

## Test Coverage

### Drop Operations (5 tests)
- ✅ **drop()** - Drop rows with null values in specified columns
- ✅ **drop() with how** - Drop using "any" or "all" strategy
- ✅ **drop() with threshold** - Drop when minimum non-null count not met
- ✅ **drop() with col(*)** - Error handling for wildcard columns
- ✅ **drop() with nested columns** - Drop based on nested struct fields

### Fill Operations (8 tests)
- ✅ **fill() numeric** - Fill null values with numeric defaults
- ✅ **fill() string** - Fill null values with string defaults
- ✅ **fill() boolean** - Fill null values with boolean defaults
- ✅ **fill() with map** - Fill multiple columns with different values
- ✅ **fill() with subset columns** - Fill only specified columns
- ✅ **fill() with col(*)** - Wildcard column handling
- ✅ **fill() with nested columns** - Nested column fill behavior
- ✅ **fillMap() with dotted names** - SPARK-34417: Column names
containing dots

### Replace Operations (8 tests)
- ✅ **replace() numeric** - Replace specific numeric values
- ✅ **replace() with null** - Replace values with null
- ✅ **replace() NaN with float** - Replace NaN values in float columns
- ✅ **replace() NaN with double** - Replace NaN values in double columns
- ✅ **replace() float with NaN** - Replace float values with NaN
- ✅ **replace() double with NaN** - Replace double values with NaN
- ✅ **replace() with dotted names** - SPARK-34649: Column names
containing dots
- ✅ **replace() nested columns** - Expected
UnsupportedOperationException

### Ambiguity Handling (2 tests)
- ✅ **fill() ambiguous columns** - AnalysisException for ambiguous
columns after join
- ✅ **drop() ambiguous columns** - AnalysisException for ambiguous
columns after join

### Duplicate Columns (2 tests)
- ✅ **SPARK-29890** - fill() with duplicate column names (no column
specification)
- ✅ **SPARK-30065** - drop() with duplicate column names (no column
specification)

### Edge Cases (2 tests)
- ✅ **Column name validation** - Error for non-existent columns
- ✅ **Nested column operations** - Proper exception handling

## Key Observations

- ✅ **100% pass rate** - Perfect GPU compatibility
- ✅ **No modifications needed** - All tests pass as-is
- ✅ **Complete NA handling** - All null/NaN operations work correctly
- ✅ **Join operations** - Ambiguity detection works on GPU
- ✅ **Dotted column names** - SPARK-34417 & SPARK-34649 support
- ✅ **NaN handling** - Float and double NaN replacement works correctly
- ✅ **Type casting** - Automatic type casting in fill() operations
- ✅ **Error handling** - All expected exceptions properly thrown

## Files Changed

-
`tests/src/test/spark330/scala/org/apache/spark/sql/rapids/suites/RapidsDataFrameNaFunctionsSuite.scala`
(new)
-
`tests/src/test/spark330/scala/org/apache/spark/sql/rapids/utils/RapidsTestSettings.scala`
(updated)

## Related Issues

Closes part of NVIDIA#11297 (Spark UT migration tracking issue)

Signed-off-by: Allen Xu <allxu@nvidia.com>

---------

Signed-off-by: Allen Xu <allxu@nvidia.com>
auto-merge triggered by github actions on `release/25.12` 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.
…ilitate CI pipeline (NVIDIA#13811)

### Description

This is a combination of 3 exitsing PRs:

- NVIDIA#13766
- NVIDIA#13770
- NVIDIA#13776

There might be merge conflict if we merge them one by one as they are
modifying the exact same files and adjacent lines, so putting them
together in just one PR is a more friendly way to get tests in.

### Checklists


- [ ] This PR has added documentation for new or modified features or
behaviors.
- [x] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

---------

Signed-off-by: Allen Xu <allxu@nvidia.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
Add below suites.
- RapidsApproximatePercentileSuite
- RapidsFirstLastTestSuite
- RapidsHistogramNumericSuite
- RapidsPercentileSuite
- RapidsProductAggSuite
- RapidsComplexTypesSuite
- RapidsCsvFunctionsSuite
- RapidsGeneratorFunctionSuite
- RapidsRandomSuite 
- RapidsDecimalExpressionSuite

---------

Signed-off-by: Gary Shen <gashen@nvidia.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
…de (NVIDIA#13817)

close NVIDIA#13731

As reported in the linked issue, the `GpuBroadcastNestedLoopJoinExec`
will complain errors when meeting the degenerate left-outer joins as
below. (aka there are no columns in either build or stream side)
```
Caused by: java.lang.ArrayIndexOutOfBoundsException: Index 0 out of bounds for length 0
        at ai.rapids.cudf.Table.<init>(Table.java:58)
        at com.nvidia.spark.rapids.GpuColumnVector.from(GpuColumnVector.java:591)
        at org.apache.spark.sql.rapids.execution.ConditionalNestedLoopJoinIterator.$anonfun$computeNumJoinRows$2(GpuBroadcastNestedLoopJoinExecBase.scala:227)
        at scala.runtime.java8.JFunction0$mcJ$sp.apply(JFunction0$mcJ$sp.scala:17)
```

This is because the cuDF `Table` requires at least one column as the
input, then the current code can NOT support the degenerate cases.
This PR adds in the support by performing the similar join logic by
leveraging the existing Java APIs as the Spark nested-loop left-outer
join for the degenerate cases.

---------

Signed-off-by: Firestarman <firestarmanllc@gmail.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
…NVIDIA#13819)

Fixes NVIDIA#13798.

### Description

This change was missing in my previous PR NVIDIA#13742. 

### Checklists

- [ ] This PR has added documentation for new or modified features or
behaviors.
- [x] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

Signed-off-by: Jihoon Son <ghoonson@gmail.com>
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
…A#13857)

Follow-up NVIDIA#13819

### Description

We don't support the `GpuDataWritingCommandExec` for LoRe since
`WriteFiles` was introduced in Spark 3.4.0. Currently, we check if it is
supported using a hardcoded list, which makes it easy to forget to
update it when a new shim layer is added.

This PR moves the check to the shim layer, so we no longer need to
remember to sync the list.

I thought this PR would fix
NVIDIA#13855, but it turns out
that it's a duplicate of
NVIDIA#13798 that has already
been fixed by @jihoonson in
NVIDIA#13819. I didn't upmerge my
code. Thanks!

### Checklists

- [x] This PR has added documentation for new or modified features or
behaviors.
- [x] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

Signed-off-by: Haoyang Li <haoyangl@nvidia.com>
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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 and others added 20 commits December 9, 2025 15:15
auto-merge triggered by github actions on `release/25.12` 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.
This is part 1 of splitting
NVIDIA#13368 into smaller parts

### Description

This adds in framework changes to allow an expression to update SQL
metrics. The metrics are passed in when the expressions are bound and
then can be updated as they run. There are no expressions added, yet
that would let us truly test it beyond seeing that it does not break
existing functionality.

### Checklists

- [ ] This PR has added documentation for new or modified features or
behaviors.
- [ ] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

This is an internal only feature so no docs or tests are really needed,
beyond verifying that we didn't break anything.

Most of the changes are very small, except for the ones in
GpuBoundAttribute and GpuMetrics

---------

Signed-off-by: Robert (Bobby) Evans <bobby@apache.org>
…IA#13987)

This is part 2 of splitting up
NVIDIA#13368

### Description

This adds in a layer of indirection when converting an expression to be
on the GPU. It is not used in the PR, but in the final PR it will allow
us to wrap CPU expressions in a GPU expression compatibility layer
(bridge).

### Checklists

- [ ] This PR has added documentation for new or modified features or
behaviors.
- [ ] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

No real need for new tests. If they still all pass and everything
compiles, then the refactoring is good.

---------

Signed-off-by: Robert (Bobby) Evans <bobby@apache.org>
Part of NVIDIA#13751 .

### Description

1. Update scala2.13 CI build + test to cover spark3.5.0+ only.
2. Remove also CDH profiles

Note: the actual cleanup of code and doc should be covered in other
changes.

### Checklists

- [ ] This PR has added documentation for new or modified features or
behaviors.
- [X] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

Signed-off-by: Peixin Li <pxLi@nyu.edu>
Adds myself to blossom-ci.yml to trigger builds.

### Checklists

- [ ] This PR has added documentation for new or modified features or
behaviors.
- [ ] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

---------

Signed-off-by: Paul Mattione <pmattione@nvidia.com>
auto-merge triggered by github actions on `release/25.12` 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.
…3994)

Some cases the dataframe failed to collect like the case in NVIDIA#13760 
Cast the column to string and compare with the expectation as String
type when collect() call throws exception.

Close NVIDIA#13760

---------

Signed-off-by: Gary Shen <gashen@nvidia.com>
Fixes #[13739](NVIDIA#13739).

### Description

This adds an integration test to test writing and reading parquet
columns with mixed encodings. We write a string column with a few pages
of repeated short strings, and a with a few pages of extremely long
unique strings. We test this with both the cpu and gpu parquet writers.

Both writers fall back to plain encoding for the pages with long unique
strings. The gpu writer puts the data with different encodings in
different row groups, whereas the cpu writer keeps them in the same row
group. This test then makes sure that both the cpu & gpu reads of both
the cpu- & gpu-generated data are identical.

### Checklists

- [ ] This PR has added documentation for new or modified features or
behaviors.
- [x] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

---------

Signed-off-by: Paul Mattione <pmattione@nvidia.com>
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
resolve auto-merge conflict at NVIDIA#14007  

There is no code change for 26.02 but only keep commit history

NOTE: this must be merged as `create a merge commit`
auto-merge triggered by github actions on `release/25.12` 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.
Contributes to NVIDIA#13726

### Description

In NVIDIA#13726, we see that many cases that are not wrapped with a
'withRetry' are from C2R. We have also seen OOM occur in C2R from
customers.

We wanted to wrap `RowToColumnarIterator.buildBatch` in `withRetry` to
prevent CPU/GPU OOMs. That didn’t work because `buildBatch` consumed
rows straight from the upstream iterator and threw them away. When the
retry framework requested a retry, there was nothing to replay.

~~This PR buffers each incoming row (copying UnsafeRows as needed),
builds batches from that buffer, and records how many rows were consumed
per attempt. When a retry happens, we can re-run `buildBatchAttempt`
against the buffered rows with a smaller target size.~~

This PR:
- Added a retryable wrapper for `convert` that restore one row when OOM
occurs, and roll back column builders to pre-conversion state on retry.
- Added snapshot/restore to `GpuColumnarBatchBuilder`.

### Checklists

- [ ] This PR has added documentation for new or modified features or
behaviors.
- [x] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

---------

Signed-off-by: Haoyang Li <haoyangl@nvidia.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
auto-merge triggered by github actions on `release/25.12` 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.
auto-merge triggered by github actions on `release/25.12` 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.
## Summary

Upgrade GitHub Actions to their latest versions to ensure compatibility
with Node 24, as Node 20 will reach end-of-life in April 2026.

## Changes

| Action | Old Version(s) | New Version | Release | Files |
|--------|---------------|-------------|---------|-------|
| `actions/setup-java` |
[`v4`](https://github.com/actions/setup-java/releases/tag/v4) |
[`v5`](https://github.com/actions/setup-java/releases/tag/v5) |
[Release](https://github.com/actions/setup-java/releases/tag/v5) |
blossom-ci.yml |

## Context

Per [GitHub's
announcement](https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/),
Node 20 is being deprecated and runners will begin using Node 24 by
default starting March 4th, 2026.

**NOTE:** Blossom-ci integrity check has been updated to cover this
change internally

Signed-off-by: Peixin Li <pxLi@nyu.edu>
Fixes NVIDIA#13750 
Part NVIDIA#13751 

### Description

1. Update scala2.12 CI build + test to cover spark3.3.0+ only.
2. Keep 321 profile for legacy scenarios

Note: the actual cleanup of code and doc should be covered in other
changes.


### Checklists


- [ ] This PR has added documentation for new or modified features or
behaviors.
- [X] This PR has added new tests or modified existing tests to cover
new code paths.
(Please explain in the PR description how the new code paths are tested,
such as names of the new/existing tests that cover them.)
- [ ] Performance testing has been performed and its results are added
in the PR description. Or, an issue has been filed with a link in the PR
description.

---------

Signed-off-by: Peixin Li <pxLi@nyu.edu>
Signed-off-by: liyuan <yuali@nvidia.com>
@nvliyuan nvliyuan changed the title update cuda13 related jars in download doc [DOC] update cuda13 related jars in download doc [skip ci] Dec 17, 2025
@nvliyuan nvliyuan requested a review from NvTimLiu December 17, 2025 06:06
@nvliyuan

Copy link
Copy Markdown
Collaborator Author

Is this still correct? CC @NvTimLiu

This package is built against CUDA 12.9. It is tested on V100, T4, A10, A100, L4, H100 and GB100 GPUs with 
CUDA 12.9.  

@nvliyuan nvliyuan self-assigned this Dec 17, 2025
@nvliyuan nvliyuan added the documentation Improvements or additions to documentation label Dec 17, 2025
@greptile-apps

greptile-apps Bot commented Dec 17, 2025

Copy link
Copy Markdown
Contributor

Greptile Summary

Added CUDA 13 support documentation to the download page, including download links, signatures, and Maven dependencies for all processor architectures (x86_64 and arm64) and Scala versions (2.12 and 2.13).

Changes Made

  • Reorganized CUDA 12 downloads under a dedicated "CUDA 12" section header
  • Added new "CUDA 13" section with comprehensive jar download links following the same structure
  • Included all four combinations: x86_64 (Scala 2.12/2.13) and arm64 (Scala 2.12/2.13)
  • All links follow consistent naming pattern with cuda13 and cuda13-arm64 classifiers

Issues Found

  • The documentation note at the bottom still references only "CUDA 12.9", which is now incomplete given that CUDA 13 variants are also available. This needs to be updated to clarify the CUDA versions for each variant.

Confidence Score: 4/5

  • This PR is safe to merge with minimal risk - it's a documentation-only change adding CUDA 13 jar references
  • Score reflects that this is a straightforward documentation update with consistent formatting and complete coverage of all CUDA 13 variants. The only issue is an outdated note about CUDA versions that should be updated for clarity.
  • The note at lines 95-96 in docs/download.md needs updating to accurately reflect both CUDA 12 and CUDA 13 variants

Important Files Changed

Filename Overview
docs/download.md Added CUDA 13 section with jar download links for all processor/Scala combinations. Documentation note about CUDA versions needs updating to reflect both CUDA 12 and 13 variants.

Sequence Diagram

sequenceDiagram
    participant Author as PR Author
    participant Docs as download.md
    participant Maven as Maven Repository
    participant Users as End Users
    
    Author->>Docs: Add "CUDA 12" section header
    Author->>Docs: Add "CUDA 13" section with table
    Author->>Docs: Include x86_64 Scala 2.12 link
    Author->>Docs: Include x86_64 Scala 2.13 link
    Author->>Docs: Include arm64 Scala 2.12 link
    Author->>Docs: Include arm64 Scala 2.13 link
    
    Note over Docs: Each entry includes:<br/>- Download jar link<br/>- Signature link<br/>- Maven dependency XML
    
    Users->>Docs: Read documentation
    Users->>Maven: Download CUDA 13 jars
    
    Note over Docs,Users: Issue: CUDA version note<br/>still references only 12.9
Loading

@greptile-apps greptile-apps Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Additional Comments (1)

  1. docs/download.md, line 95-96 (link)

    logic: Documentation now covers both CUDA 12 and CUDA 13 variants, but this note only mentions CUDA 12.9. Update to clarify which CUDA versions the different jar variants are built against.

1 file reviewed, 1 comment

Edit Code Review Agent Settings | Greptile

@nvliyuan nvliyuan changed the base branch from main to release/25.12 December 17, 2025 06:14
@nvliyuan nvliyuan requested a review from a team as a code owner December 17, 2025 06:14
@nvliyuan nvliyuan closed this Dec 17, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[DOC] Need to update cuda13 related jars in download page