Branch 25.12 merge branch 25.10

benchmarks/linear_programming/cuopt/run_mip.cpp

@@ -80,10 +80,11 @@ void merge_result_files(const std::string& out_dir,
 void write_to_output_file(const std::string& out_dir,
                           const std::string& base_filename,
                           int gpu_id,
+                          int n_gpus,
                           int batch_id,
                           const std::string& data)
 {
-  int output_id        = batch_id * 8 + gpu_id;
+  int output_id        = batch_id * n_gpus + gpu_id;
   std::string filename = out_dir + "/result_" + std::to_string(output_id) + ".txt";
   std::ofstream outfile(filename, std::ios_base::app);
   if (outfile.is_open()) {
@@ -149,6 +150,7 @@ std::vector<std::vector<double>> read_solution_from_dir(const std::string file_p
 int run_single_file(std::string file_path,
                     int device,
                     int batch_id,
+                    int n_gpus,
                     std::string out_dir,
                     std::optional<std::string> initial_solution_dir,
                     bool heuristics_only,
@@ -243,14 +245,15 @@ int run_single_file(std::string file_path,
      << obj_val << "," << benchmark_info.objective_of_initial_population << ","
      << benchmark_info.last_improvement_of_best_feasible << ","
      << benchmark_info.last_improvement_after_recombination << "\n";
-  write_to_output_file(out_dir, base_filename, device, batch_id, ss.str());
+  write_to_output_file(out_dir, base_filename, device, n_gpus, batch_id, ss.str());
   CUOPT_LOG_INFO("Results written to the file %s", base_filename.c_str());
   return sol_found;
 }
 
 void run_single_file_mp(std::string file_path,
                         int device,
                         int batch_id,
+                        int n_gpus,
                         std::string out_dir,
                         std::optional<std::string> input_file_dir,
                         bool heuristics_only,
@@ -265,6 +268,7 @@ void run_single_file_mp(std::string file_path,
   int sol_found = run_single_file(file_path,
                                   device,
                                   batch_id,
+                                  n_gpus,
                                   out_dir,
                                   input_file_dir,
                                   heuristics_only,
@@ -462,6 +466,7 @@ int main(int argc, char* argv[])
             run_single_file_mp(file_name,
                                gpu_id,
                                batch_num,
+                               n_gpus,
                                out_dir,
                                initial_solution_file,
                                heuristics_only,
@@ -501,6 +506,7 @@ int main(int argc, char* argv[])
     run_single_file(path,
                     0,
                     0,
+                    n_gpus,
                     out_dir,
                     initial_solution_file,
                     heuristics_only,

ci/test_python.sh

@@ -60,6 +60,9 @@ EXITCODE=0
 trap "EXITCODE=1" ERR
 set +e
 
+# Due to race condition in certain cases UCX might not be able to cleanup properly, so we set the number of threads to 1
+export OMP_NUM_THREADS=1
+
 rapids-logger "Test cuopt_cli"
 timeout 10m bash ./python/libcuopt/libcuopt/tests/test_cli.sh
 

ci/test_wheel_cuopt.sh

@@ -66,11 +66,14 @@ cd -
 RAPIDS_DATASET_ROOT_DIR="$(realpath datasets)"
 export RAPIDS_DATASET_ROOT_DIR
 
-# Please enable this once ISSUE https://github.com/NVIDIA/cuopt/issues/94 is fixed
 # Run CLI tests
 timeout 10m bash ./python/libcuopt/libcuopt/tests/test_cli.sh
 
 # Run Python tests
+
+# Due to race condition in certain cases UCX might not be able to cleanup properly, so we set the number of threads to 1
+export OMP_NUM_THREADS=1
+
 RAPIDS_DATASET_ROOT_DIR=./datasets timeout 30m python -m pytest --verbose --capture=no ./python/cuopt/cuopt/tests/
 
 # run jump tests and cvxpy integration tests for only nightly builds

ci/test_wheel_cuopt_server.sh

@@ -37,4 +37,7 @@ rapids-pip-retry install \
 ./datasets/linear_programming/download_pdlp_test_dataset.sh
 ./datasets/mip/download_miplib_test_dataset.sh
 
+# Due to race condition in certain cases UCX might not be able to cleanup properly, so we set the number of threads to 1
+export OMP_NUM_THREADS=1
+
 RAPIDS_DATASET_ROOT_DIR=./datasets timeout 30m python -m pytest --verbose --capture=no ./python/cuopt_server/cuopt_server/tests/

conda/environments/all_cuda-129_arch-aarch64.yaml

@@ -59,10 +59,9 @@ dependencies:
 - pytest-cov
 - pytest<8
 - python>=3.10,<3.14
-- raft-dask==25.12.*,>=0.0.0a0
 - rapids-build-backend>=0.4.0,<0.5.0.dev0
 - rapids-dask-dependency==25.12.*,>=0.0.0a0
-- rapids-logger==0.1.*,>=0.0.0a0
+- rapids-logger==0.2.*,>=0.0.0a0
 - requests
 - rmm==25.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0

conda/environments/all_cuda-129_arch-x86_64.yaml

@@ -59,10 +59,9 @@ dependencies:
 - pytest-cov
 - pytest<8
 - python>=3.10,<3.14
-- raft-dask==25.12.*,>=0.0.0a0
 - rapids-build-backend>=0.4.0,<0.5.0.dev0
 - rapids-dask-dependency==25.12.*,>=0.0.0a0
-- rapids-logger==0.1.*,>=0.0.0a0
+- rapids-logger==0.2.*,>=0.0.0a0
 - requests
 - rmm==25.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0

conda/environments/all_cuda-130_arch-aarch64.yaml

@@ -59,10 +59,9 @@ dependencies:
 - pytest-cov
 - pytest<8
 - python>=3.10,<3.14
-- raft-dask==25.12.*,>=0.0.0a0
 - rapids-build-backend>=0.4.0,<0.5.0.dev0
 - rapids-dask-dependency==25.12.*,>=0.0.0a0
-- rapids-logger==0.1.*,>=0.0.0a0
+- rapids-logger==0.2.*,>=0.0.0a0
 - requests
 - rmm==25.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0

conda/environments/all_cuda-130_arch-x86_64.yaml

@@ -59,10 +59,9 @@ dependencies:
 - pytest-cov
 - pytest<8
 - python>=3.10,<3.14
-- raft-dask==25.12.*,>=0.0.0a0
 - rapids-build-backend>=0.4.0,<0.5.0.dev0
 - rapids-dask-dependency==25.12.*,>=0.0.0a0
-- rapids-logger==0.1.*,>=0.0.0a0
+- rapids-logger==0.2.*,>=0.0.0a0
 - requests
 - rmm==25.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0

conda/recipes/libcuopt/recipe.yaml

@@ -67,7 +67,7 @@ cache:
       - gtest ${{ gtest_version }}
       - libraft-headers =${{ dep_minor_version }}
       - librmm =${{ dep_minor_version }}
-      - rapids-logger =0.1
+      - rapids-logger =0.2
       - cuda-nvtx-dev
       - libcudss-dev >=0.7
       - libcurand-dev
@@ -148,7 +148,7 @@ outputs:
         - ${{ pin_subpackage("libmps-parser", exact=True) }}
         - boost
         - cuda-version =${{ cuda_version }}
-        - rapids-logger =0.1
+        - rapids-logger =0.2
         - librmm =${{ dep_minor_version }}
         - cuda-cudart-dev
         - libcublas

cpp/CMakeLists.txt

@@ -190,9 +190,9 @@ FetchContent_Declare(
   # does not have some of the presolvers and settings that we need
   # Mainly, probing and clique merging.
   # This is the reason we are using the development branch
-  # commit from Oct 8, 2025. Once these changes are merged into the main branch,
+  # commit from cliquemergebug branch. Once these changes are merged into the main branch,
   #we can switch to the main branch.
-  GIT_TAG "24ccf5752656df0f15dd9aabe5b97feae829b9ec"
+  GIT_TAG "8f710e33d352bf319d30b9c57e70516222f3f5ca"
   GIT_PROGRESS TRUE
   SYSTEM
 )
@@ -201,8 +201,6 @@ find_package(TBB REQUIRED)
 set(BUILD_TESTING OFF CACHE BOOL "Disable test build for papilo")
 set(PAPILO_NO_BINARIES ON)
 option(LUSOL "Disable LUSOL" OFF)
-# Disable TBB because of a bug in CliqueMerging parallel version
-set(TBB OFF CACHE BOOL "Disable TBB for papilo")
 
 FetchContent_MakeAvailable(papilo)