diff --git a/conf/common/test/dse_nccl_all_gather.toml b/conf/common/test/dse_nccl_all_gather.toml index f36671e62..94cb2579e 100644 --- a/conf/common/test/dse_nccl_all_gather.toml +++ b/conf/common/test/dse_nccl_all_gather.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -19,7 +19,7 @@ description = "all_gather" test_template_name = "NcclTest" [cmd_args] -"docker_image_url" = "nvcr.io#nvidia/pytorch:25.06-py3" +"docker_image_url" = "nvcr.io#nvidia/pytorch:25.12-py3" "subtest_name" = "all_gather_perf_mpi" "ngpus" = "1" "minbytes" = "128" diff --git a/conf/common/test/nccl_test.toml b/conf/common/test/nccl_test.toml index ea749c386..ecb1216e1 100644 --- a/conf/common/test/nccl_test.toml +++ b/conf/common/test/nccl_test.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -19,7 +19,7 @@ description = "NCCL base test configuration" test_template_name = "NcclTest" [cmd_args] -docker_image_url = "nvcr.io#nvidia/pytorch:25.06-py3" +docker_image_url = "nvcr.io#nvidia/pytorch:25.12-py3" ngpus = 1 minbytes = "128" maxbytes = "4G" diff --git a/conf/common/test/nccl_test_all_gather.toml b/conf/common/test/nccl_test_all_gather.toml index cf68098b5..b6bc19329 100644 --- a/conf/common/test/nccl_test_all_gather.toml +++ b/conf/common/test/nccl_test_all_gather.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -19,7 +19,7 @@ description = "all_gather" test_template_name = "NcclTest" [cmd_args] -"docker_image_url" = "nvcr.io#nvidia/pytorch:25.06-py3" +"docker_image_url" = "nvcr.io#nvidia/pytorch:25.12-py3" "subtest_name" = "all_gather_perf_mpi" "ngpus" = "1" "minbytes" = "128" diff --git a/conf/common/test/nemo_run_llama3_8b.toml b/conf/common/test/nemo_run_llama3_8b.toml index b545a231a..f3ddec1d9 100644 --- a/conf/common/test/nemo_run_llama3_8b.toml +++ b/conf/common/test/nemo_run_llama3_8b.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -19,7 +19,7 @@ description = "nemo_run_llama3_8b" test_template_name = "NeMoRun" [cmd_args] -docker_image_url = "nvcr.io#nvidia/nemo:25.09.00" +docker_image_url = "nvcr.io#nvidia/nemo:25.11.01" task = "pretrain" recipe_name = "cloudai_llama3_8b_recipe" diff --git a/conf/common/test/osu_test.toml b/conf/common/test/osu_test.toml index 5e6285266..c279a9be7 100644 --- a/conf/common/test/osu_test.toml +++ b/conf/common/test/osu_test.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -19,7 +19,7 @@ test_template_name = "OSUBench" description = "OSU Benchmark example" [cmd_args] -"docker_image_url" = "nvcr.io#nvidia/pytorch:25.06-py3" +"docker_image_url" = "nvcr.io#nvidia/pytorch:25.12-py3" "benchmarks_dir" = "/opt/hpcx/ompi/tests/osu-micro-benchmarks" "benchmark" = "osu_allreduce" "iterations" = 10 diff --git a/conf/common/test/ucc_test.toml b/conf/common/test/ucc_test.toml index ed47cd0f2..ca308d064 100644 --- a/conf/common/test/ucc_test.toml +++ b/conf/common/test/ucc_test.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -19,4 +19,4 @@ description = "Base config for UCC tests" test_template_name = "UCCTest" [cmd_args] -docker_image_url = "nvcr.io#nvidia/pytorch:25.06-py3" +docker_image_url = "nvcr.io#nvidia/pytorch:25.12-py3" diff --git a/conf/common/test_scenario/dse_nemo_run_llama3_8b.toml b/conf/common/test_scenario/dse_nemo_run_llama3_8b.toml index d1887fd49..7c4b79757 100644 --- a/conf/common/test_scenario/dse_nemo_run_llama3_8b.toml +++ b/conf/common/test_scenario/dse_nemo_run_llama3_8b.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -26,7 +26,7 @@ description = "dse_nemo_run_llama3_8b" test_template_name = "NeMoRun" [Tests.cmd_args] - docker_image_url = "nvcr.io#nvidia/nemo:25.09.00" + docker_image_url = "nvcr.io#nvidia/nemo:25.11.01" task = "pretrain" recipe_name = "cloudai_llama3_8b_recipe" num_layers = 32 diff --git a/conf/common/test_scenario/nccl_test.toml b/conf/common/test_scenario/nccl_test.toml index c9f3cf1be..9f850ca06 100644 --- a/conf/common/test_scenario/nccl_test.toml +++ b/conf/common/test_scenario/nccl_test.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -68,7 +68,7 @@ description = "scatter_perf" test_template_name = "NcclTest" [Tests.cmd_args] - docker_image_url = "nvcr.io#nvidia/pytorch:25.06-py3" + docker_image_url = "nvcr.io#nvidia/pytorch:25.12-py3" subtest_name = "scatter_perf_mpi" ngpus = 1 minbytes = "128" diff --git a/conf/common/test_scenario/slurm_container.toml b/conf/common/test_scenario/slurm_container.toml index 4c92e8582..73487a21b 100644 --- a/conf/common/test_scenario/slurm_container.toml +++ b/conf/common/test_scenario/slurm_container.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -26,5 +26,5 @@ description = "NCCL alltoall via SlurmContainer" test_template_name = "SlurmContainer" [Tests.cmd_args] - docker_image_url = "nvcr.io#nvidia/pytorch:25.06-py3" + docker_image_url = "nvcr.io#nvidia/pytorch:25.12-py3" cmd = "alltoall_perf_mpi --nthreads 1 --ngpus 1 --minbytes 128 --maxbytes 4G --stepbytes 1M --op sum --datatype float --root 0 --iters 100 --warmup_iters 50 --agg_iters 1 --average 1 --parallel_init 0 --check 1 --blocking 0 --cudagraph 0 --stepfactor 2" diff --git a/conf/common/test_scenario/ucc_generator_test.toml b/conf/common/test_scenario/ucc_generator_test.toml index ddd180032..91ff0a04e 100644 --- a/conf/common/test_scenario/ucc_generator_test.toml +++ b/conf/common/test_scenario/ucc_generator_test.toml @@ -1,5 +1,5 @@ # SPDX-FileCopyrightText: NVIDIA CORPORATION & AFFILIATES -# Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. # SPDX-License-Identifier: Apache-2.0 # # Licensed under the Apache License, Version 2.0 (the "License"); @@ -30,6 +30,6 @@ extra_container_mounts = [ ] [Tests.cmd_args] - docker_image_url = "nvcr.io/nvidia/pytorch:25.09-py3" + docker_image_url = "nvcr.io/nvidia/pytorch:25.12-py3" collective = "alltoallv" gen = "file:name=/opt/hpcx/ucc/tools/perf/generator/input_matrices.txt"