Skip to content

Commit

Permalink
Merge branch 'branch-24.12' into nccl-clique-init
Browse files Browse the repository at this point in the history
  • Loading branch information
cjnolet authored Dec 12, 2024
2 parents 7f1961b + e8dca7c commit 5f1ce2a
Show file tree
Hide file tree
Showing 2 changed files with 64 additions and 0 deletions.
39 changes: 39 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
@@ -1,3 +1,42 @@
# raft 24.12.00 (11 Dec 2024)

## 🚨 Breaking Changes

- Do not initialize the pinned mdarray at construction time ([#2478](https://github.com/rapidsai/raft/pull/2478)) [@achirkin](https://github.com/achirkin)

## 🐛 Bug Fixes

- Skip gtests for new lanczos solver when CUDA version is 11.4 or below. ([#2520](https://github.com/rapidsai/raft/pull/2520)) [@cjnolet](https://github.com/cjnolet)
- Switch `assert` to `static_assert` ([#2510](https://github.com/rapidsai/raft/pull/2510)) [@divyegala](https://github.com/divyegala)
- Revert use of new Lanczos solver in spectral clustering ([#2507](https://github.com/rapidsai/raft/pull/2507)) [@lowener](https://github.com/lowener)
- Put a ceiling on cuda-python ([#2486](https://github.com/rapidsai/raft/pull/2486)) [@bdice](https://github.com/bdice)
- Don't presume pointers location infers usability. ([#2480](https://github.com/rapidsai/raft/pull/2480)) [@robertmaynard](https://github.com/robertmaynard)
- Use Python for sccache hit rate computation. ([#2474](https://github.com/rapidsai/raft/pull/2474)) [@bdice](https://github.com/bdice)
- Allow compilation with CUDA 12.6.1 ([#2469](https://github.com/rapidsai/raft/pull/2469)) [@robertmaynard](https://github.com/robertmaynard)

## 🚀 New Features

- [FEA] Lanczos solver v2 ([#2481](https://github.com/rapidsai/raft/pull/2481)) [@lowener](https://github.com/lowener)

## 🛠️ Improvements

- Skip gtests for Rmat Lanczos tests with cuda <= 11.4 ([#2525](https://github.com/rapidsai/raft/pull/2525)) [@benfred](https://github.com/benfred)
- Upgrade to latest cutlass version ([#2503](https://github.com/rapidsai/raft/pull/2503)) [@vyasr](https://github.com/vyasr)
- Removing some left over places where implicit instantiations were being ignored in headers ([#2501](https://github.com/rapidsai/raft/pull/2501)) [@cjnolet](https://github.com/cjnolet)
- Remove leftover template project code. ([#2500](https://github.com/rapidsai/raft/pull/2500)) [@bdice](https://github.com/bdice)
- 2412 remove libraft vss instantiations ([#2498](https://github.com/rapidsai/raft/pull/2498)) [@cjnolet](https://github.com/cjnolet)
- Remove raft-ann-bench ([#2497](https://github.com/rapidsai/raft/pull/2497)) [@cjnolet](https://github.com/cjnolet)
- Pin FAISS Version for raft-ann-bench ([#2496](https://github.com/rapidsai/raft/pull/2496)) [@tarang-jain](https://github.com/tarang-jain)
- enforce wheel size limits and README formatting in CI, put a ceiling on Cython dependency ([#2490](https://github.com/rapidsai/raft/pull/2490)) [@jameslamb](https://github.com/jameslamb)
- Do not initialize the pinned mdarray at construction time ([#2478](https://github.com/rapidsai/raft/pull/2478)) [@achirkin](https://github.com/achirkin)
- Use environment variables in cache hit rate computation. ([#2475](https://github.com/rapidsai/raft/pull/2475)) [@bdice](https://github.com/bdice)
- devcontainer: replace `VAULT_HOST` with `AWS_ROLE_ARN` ([#2472](https://github.com/rapidsai/raft/pull/2472)) [@jjacobelli](https://github.com/jjacobelli)
- print sccache stats in builds ([#2470](https://github.com/rapidsai/raft/pull/2470)) [@jameslamb](https://github.com/jameslamb)
- make package installations in CI stricter ([#2467](https://github.com/rapidsai/raft/pull/2467)) [@jameslamb](https://github.com/jameslamb)
- Prune workflows based on changed files ([#2466](https://github.com/rapidsai/raft/pull/2466)) [@KyleFromNVIDIA](https://github.com/KyleFromNVIDIA)
- Merge branch-24.10 into branch-24.12 ([#2461](https://github.com/rapidsai/raft/pull/2461)) [@jameslamb](https://github.com/jameslamb)
- Update all rmm imports to use pylibrmm/librmm ([#2451](https://github.com/rapidsai/raft/pull/2451)) [@Matt711](https://github.com/Matt711)

# raft 24.10.00 (9 Oct 2024)

## 🚨 Breaking Changes
Expand Down
25 changes: 25 additions & 0 deletions cpp/test/sparse/solver/lanczos.cu
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,7 @@
#include <raft/spectral/matrix_wrappers.hpp>
#include <raft/util/cudart_utils.hpp>

#include <cuda_runtime.h>
#include <driver_types.h>

#include <gtest/gtest.h>
Expand Down Expand Up @@ -108,6 +109,18 @@ class rmat_lanczos_tests

void Run()
{
int runtimeVersion;
cudaError_t result = cudaRuntimeGetVersion(&runtimeVersion);

if (result == cudaSuccess) {
int major = runtimeVersion / 1000;
int minor = (runtimeVersion % 1000) / 10;

// Skip gtests for CUDA 11.4.x and below because hard-coded results are causing issues.
// See https://github.com/rapidsai/raft/issues/2519 for more information.
if (major == 11 && minor <= 4) { GTEST_SKIP(); }
}

uint64_t n_edges = sparsity * ((long long)(1 << r_scale) * (long long)(1 << c_scale));
uint64_t n_nodes = 1 << std::max(r_scale, c_scale);
uint64_t theta_len = std::max(r_scale, c_scale) * 4;
Expand Down Expand Up @@ -265,6 +278,18 @@ class lanczos_tests : public ::testing::TestWithParam<lanczos_inputs<IndexType,

void Run()
{
int runtimeVersion;
cudaError_t result = cudaRuntimeGetVersion(&runtimeVersion);

if (result == cudaSuccess) {
int major = runtimeVersion / 1000;
int minor = (runtimeVersion % 1000) / 10;

// Skip gtests for CUDA 11.4.x and below because hard-coded results are causing issues.
// See https://github.com/rapidsai/raft/issues/2519 for more information.
if (major == 11 && minor <= 4) { GTEST_SKIP(); }
}

raft::random::uniform<ValueType>(handle, rng, v0.view(), 0, 1);
std::tuple<IndexType, ValueType, IndexType> stats;

Expand Down

0 comments on commit 5f1ce2a

Please sign in to comment.