Skip to content

{ai}[foss/2023b] DeepSpeed v0.14.5 w/ CUDA 12.4.0#25459

Open
Flamefire wants to merge 12 commits into
easybuilders:developfrom
Flamefire:20260227155302_new_pr_DeepSpeed0145
Open

{ai}[foss/2023b] DeepSpeed v0.14.5 w/ CUDA 12.4.0#25459
Flamefire wants to merge 12 commits into
easybuilders:developfrom
Flamefire:20260227155302_new_pr_DeepSpeed0145

Conversation

@Flamefire
Copy link
Copy Markdown
Contributor

@Flamefire Flamefire commented Feb 27, 2026

…tches: DeepSpeed-0.14.2_no-ninja-dep.patch, DeepSpeed-0.14.5_avoid-access-to-home.patch, DeepSpeed-0.14.5_pdsh-env-vars.patch, DeepSpeed-0.14.5_pic-compile.patch, DeepSpeed-0.14.5_test-nvme-offload.patch, DeepSpeed-0.14.5_use-eb-cutlass.patch
@github-actions
Copy link
Copy Markdown

github-actions Bot commented Feb 27, 2026

Updated software accelerate-1.10.0-foss-2023b-CUDA-12.4.0.eb

Diff against accelerate-1.11.0-foss-2024a-CUDA-12.6.0.eb

easybuild/easyconfigs/a/accelerate/accelerate-1.11.0-foss-2024a-CUDA-12.6.0.eb

diff --git a/easybuild/easyconfigs/a/accelerate/accelerate-1.11.0-foss-2024a-CUDA-12.6.0.eb b/easybuild/easyconfigs/a/accelerate/accelerate-1.10.0-foss-2023b-CUDA-12.4.0.eb
index 6a7fe9a893..ca32d9c698 100644
--- a/easybuild/easyconfigs/a/accelerate/accelerate-1.11.0-foss-2024a-CUDA-12.6.0.eb
+++ b/easybuild/easyconfigs/a/accelerate/accelerate-1.10.0-foss-2023b-CUDA-12.4.0.eb
@@ -1,7 +1,7 @@
 easyblock = 'PythonBundle'
 
 name = 'accelerate'
-version = '1.11.0'
+version = '1.10.0'
 versionsuffix = '-CUDA-%(cudaver)s'
 
 homepage = 'https://github.com/huggingface/accelerate'
@@ -9,21 +9,25 @@ description = """A simple way to launch, train, and use PyTorch models on almost
 distributed configuration, automatic mixed precision (including fp8),
 and easy-to-configure FSDP and DeepSpeed support."""
 
-toolchain = {'name': 'foss', 'version': '2024a'}
+toolchain = {'name': 'foss', 'version': '2023b'}
 
 dependencies = [
-    ('Python', '3.12.3'),
-    ('SciPy-bundle', '2024.05'),
-    ('CUDA', '12.6.0', '', SYSTEM),
-    ('PyTorch', '2.7.1', '-CUDA-%(cudaver)s'),
-    ('PyYAML', '6.0.2'),
-    ('Safetensors', '0.6.2'),
-    ('huggingface-hub', '0.34.4'),
+    ('Python', '3.11.5'),
+    ('Python-bundle-PyPI', '2023.10'),
+    ('SciPy-bundle', '2023.11'),
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('PyTorch-bundle', '2.3.0', versionsuffix),
+    ('PyYAML', '6.0.1'),
+    ('Safetensors', '0.4.4'),
 ]
 
 exts_list = [
+    ('huggingface-hub', '0.30.2', {
+        'sources': ['huggingface_hub-%(version)s.tar.gz'],
+        'checksums': ['9a7897c5b6fd9dad3168a794a8998d6378210f5b9688d0dfc180b1a228dc2466'],
+    }),
     (name, version, {
-        'checksums': ['bb1caf2597b4cd632b917b5000c591d10730bb024a79746f1ee205bba80bd229'],
+        'checksums': ['8270568fda9036b5cccdc09703fef47872abccd56eb5f6d53b54ea5fb7581496'],
     }),
 ]
 
Diff against accelerate-1.10.0-foss-2024a.eb

easybuild/easyconfigs/a/accelerate/accelerate-1.10.0-foss-2024a.eb

diff --git a/easybuild/easyconfigs/a/accelerate/accelerate-1.10.0-foss-2024a.eb b/easybuild/easyconfigs/a/accelerate/accelerate-1.10.0-foss-2023b-CUDA-12.4.0.eb
index 2e405316e4..ca32d9c698 100644
--- a/easybuild/easyconfigs/a/accelerate/accelerate-1.10.0-foss-2024a.eb
+++ b/easybuild/easyconfigs/a/accelerate/accelerate-1.10.0-foss-2023b-CUDA-12.4.0.eb
@@ -2,24 +2,30 @@ easyblock = 'PythonBundle'
 
 name = 'accelerate'
 version = '1.10.0'
+versionsuffix = '-CUDA-%(cudaver)s'
 
 homepage = 'https://github.com/huggingface/accelerate'
 description = """A simple way to launch, train, and use PyTorch models on almost any device and
 distributed configuration, automatic mixed precision (including fp8),
 and easy-to-configure FSDP and DeepSpeed support."""
 
-toolchain = {'name': 'foss', 'version': '2024a'}
+toolchain = {'name': 'foss', 'version': '2023b'}
 
 dependencies = [
-    ('Python', '3.12.3'),
-    ('SciPy-bundle', '2024.05'),
-    ('PyTorch', '2.6.0'),
-    ('PyYAML', '6.0.2'),
-    ('Safetensors', '0.6.2'),
-    ('huggingface-hub', '0.34.4'),
+    ('Python', '3.11.5'),
+    ('Python-bundle-PyPI', '2023.10'),
+    ('SciPy-bundle', '2023.11'),
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('PyTorch-bundle', '2.3.0', versionsuffix),
+    ('PyYAML', '6.0.1'),
+    ('Safetensors', '0.4.4'),
 ]
 
 exts_list = [
+    ('huggingface-hub', '0.30.2', {
+        'sources': ['huggingface_hub-%(version)s.tar.gz'],
+        'checksums': ['9a7897c5b6fd9dad3168a794a8998d6378210f5b9688d0dfc180b1a228dc2466'],
+    }),
     (name, version, {
         'checksums': ['8270568fda9036b5cccdc09703fef47872abccd56eb5f6d53b54ea5fb7581496'],
     }),

Updated software CUTLASS-4.1.0-foss-2023b-CUDA-12.4.0.eb

Diff against CUTLASS-4.3.5-GCCcore-14.2.0-CUDA-12.8.0.eb

easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.3.5-GCCcore-14.2.0-CUDA-12.8.0.eb

diff --git a/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.3.5-GCCcore-14.2.0-CUDA-12.8.0.eb b/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.1.0-foss-2023b-CUDA-12.4.0.eb
index 2500ce05ea..ee1c2c569e 100644
--- a/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.3.5-GCCcore-14.2.0-CUDA-12.8.0.eb
+++ b/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.1.0-foss-2023b-CUDA-12.4.0.eb
@@ -1,7 +1,7 @@
-easyblock = 'CMakeNinja'
+easyblock = 'CMakeMake'
 
 name = 'CUTLASS'
-version = '4.3.5'
+version = '4.1.0'
 versionsuffix = '-CUDA-%(cudaver)s'
 
 homepage = 'https://github.com/NVIDIA/cutlass'
@@ -17,37 +17,38 @@ tuned via custom tiling sizes, data types, and other algorithmic policy.
 The resulting flexibility simplifies their use as building blocks within
 custom kernels and applications."""
 
-toolchain = {'name': 'GCCcore', 'version': '14.2.0'}
-toolchainopts = {'cstd': 'c++20'}
+toolchain = {'name': 'foss', 'version': '2023b'}
 
 github_account = 'NVIDIA'
 source_urls = [GITHUB_LOWER_SOURCE]
-sources = [{'download_filename': 'v%(version)s.tar.gz', 'filename': SOURCE_TAR_GZ}]
-checksums = ['73d8c3914a6049ff5c43b7dfb9d70f26e44dc9e10e36049db5a999b9faf6dbf0']
+sources = [{'download_filename': V_VERSION_TAR_GZ, 'filename': SOURCE_TAR_GZ}]
+patches = [
+    'CUTLASS-4.1.0_fix-version.patch',
+    'CUTLASS-4.1.0_add-arch-guards-to-tests.patch',
+]
+checksums = [
+    {'CUTLASS-4.1.0.tar.gz': '8d4675b11e9e5207e3940eaac0f46db934ada371cbb3627c9fda642d912b6230'},
+    {'CUTLASS-4.1.0_fix-version.patch': 'e2c7f66e6fd298b3af5339e17c0c75ded7d726cdf6cde003f60263e27ae46495'},
+    {'CUTLASS-4.1.0_add-arch-guards-to-tests.patch':
+     '81cd18d83bdedf3ed1f7add68bbff1635cf9d76bb9e184efbc62cd95caee4275'},
+]
 
 builddependencies = [
-    ('binutils', '2.42'),
-    ('CMake', '3.31.3'),
-    ('Ninja', '1.12.1'),
-    ('Python', '3.13.1'),
+    ('CMake', '3.27.6'),
+    ('Python', '3.11.5'),
 ]
 
 dependencies = [
-    ('CUDA', '12.8.0', '', SYSTEM),
-    ('cuDNN', '9.10.1.4', versionsuffix, SYSTEM),
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('cuDNN', '9.1.1.17', versionsuffix, SYSTEM),
 ]
 
-# default CUDA compute capabilities to use (override via --cuda-compute-capabilities)
-cuda_compute_capabilities = ['6.0', '7.0', '7.5', '8.0', '8.6', '9.0']
-
-# fix C++ standard for nvcc
-preconfigopts = "sed -i 's/CUDA_STANDARD 17/CUDA_STANDARD 20/' ../%(namelower)s-%(version)s/CMakeLists.txt && "
-
-configopts = ' '.join([
+_copts = [
     '-DCUTLASS_NVCC_ARCHS="%(cuda_cc_cmake)s"',
     '-DCUTLASS_ENABLE_CUBLAS=1',
     '-DCUTLASS_ENABLE_CUDNN=1',
-])
+]
+configopts = ' '.join(_copts)
 
 sanity_check_paths = {
     'files': ['include/cutlass/cutlass.h', 'lib/libcutlass.%s' % SHLIB_EXT],
Diff against CUTLASS-4.1.0-foss-2024a-CUDA-12.6.0.eb

easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.1.0-foss-2024a-CUDA-12.6.0.eb

diff --git a/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.1.0-foss-2024a-CUDA-12.6.0.eb b/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.1.0-foss-2023b-CUDA-12.4.0.eb
index 44812ea174..ee1c2c569e 100644
--- a/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.1.0-foss-2024a-CUDA-12.6.0.eb
+++ b/easybuild/easyconfigs/c/CUTLASS/CUTLASS-4.1.0-foss-2023b-CUDA-12.4.0.eb
@@ -1,4 +1,4 @@
-easyblock = 'CMakeNinja'
+easyblock = 'CMakeMake'
 
 name = 'CUTLASS'
 version = '4.1.0'
@@ -17,11 +17,11 @@ tuned via custom tiling sizes, data types, and other algorithmic policy.
 The resulting flexibility simplifies their use as building blocks within
 custom kernels and applications."""
 
-toolchain = {'name': 'foss', 'version': '2024a'}
+toolchain = {'name': 'foss', 'version': '2023b'}
 
 github_account = 'NVIDIA'
 source_urls = [GITHUB_LOWER_SOURCE]
-sources = [{'download_filename': 'v%(version)s.tar.gz', 'filename': SOURCE_TAR_GZ}]
+sources = [{'download_filename': V_VERSION_TAR_GZ, 'filename': SOURCE_TAR_GZ}]
 patches = [
     'CUTLASS-4.1.0_fix-version.patch',
     'CUTLASS-4.1.0_add-arch-guards-to-tests.patch',
@@ -34,21 +34,21 @@ checksums = [
 ]
 
 builddependencies = [
-    ('CMake', '3.29.3'),
-    ('Ninja', '1.12.1'),
-    ('Python', '3.12.3'),
+    ('CMake', '3.27.6'),
+    ('Python', '3.11.5'),
 ]
 
 dependencies = [
-    ('CUDA', '12.6.0', '', SYSTEM),
-    ('cuDNN', '9.5.0.50', versionsuffix, SYSTEM),
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('cuDNN', '9.1.1.17', versionsuffix, SYSTEM),
 ]
 
-configopts = ' '.join([
+_copts = [
     '-DCUTLASS_NVCC_ARCHS="%(cuda_cc_cmake)s"',
     '-DCUTLASS_ENABLE_CUBLAS=1',
     '-DCUTLASS_ENABLE_CUDNN=1',
-])
+]
+configopts = ' '.join(_copts)
 
 sanity_check_paths = {
     'files': ['include/cutlass/cutlass.h', 'lib/libcutlass.%s' % SHLIB_EXT],

Updated software CuPy-13.6.0-foss-2023b-CUDA-12.4.0.eb

Diff against CuPy-13.6.0-foss-2024a-CUDA-12.6.0.eb

easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2024a-CUDA-12.6.0.eb

diff --git a/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2024a-CUDA-12.6.0.eb b/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2023b-CUDA-12.4.0.eb
index 72dac5f5cf..fe488db1d3 100644
--- a/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2024a-CUDA-12.6.0.eb
+++ b/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2023b-CUDA-12.4.0.eb
@@ -7,19 +7,19 @@ versionsuffix = '-CUDA-%(cudaver)s'
 homepage = 'https://cupy.dev'
 description = "CuPy is an open-source array library accelerated with NVIDIA CUDA."
 
-toolchain = {'name': 'foss', 'version': '2024a'}
+toolchain = {'name': 'foss', 'version': '2023b'}
 
 builddependencies = [
-    ('hypothesis', '6.103.1'),
+    ('hypothesis', '6.90.0'),
     ('Cython', '3.0.10'),
     ('setuptools', '80.9.0'),
 ]
 
 dependencies = [
-    ('Python', '3.12.3'),
-    ('SciPy-bundle', '2024.05'),
-    ('CUDA', '12.6.0', '', SYSTEM),
-    ('NCCL', '2.22.3', versionsuffix),
+    ('Python', '3.11.5'),
+    ('SciPy-bundle', '2023.11'),
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('NCCL', '2.20.5', versionsuffix),
     ('cuTENSOR', '2.0.2.5', versionsuffix, SYSTEM),
     ('cuSPARSELt', '0.8.0.4', versionsuffix, SYSTEM),  # docs say 0.7.0 or 0.7.1
 ]
Diff against CuPy-13.6.0-foss-2025a-CUDA-12.8.0.eb

easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2025a-CUDA-12.8.0.eb

diff --git a/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2025a-CUDA-12.8.0.eb b/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2023b-CUDA-12.4.0.eb
index 591cc5bb8d..fe488db1d3 100644
--- a/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2025a-CUDA-12.8.0.eb
+++ b/easybuild/easyconfigs/c/CuPy/CuPy-13.6.0-foss-2023b-CUDA-12.4.0.eb
@@ -7,20 +7,20 @@ versionsuffix = '-CUDA-%(cudaver)s'
 homepage = 'https://cupy.dev'
 description = "CuPy is an open-source array library accelerated with NVIDIA CUDA."
 
-toolchain = {'name': 'foss', 'version': '2025a'}
+toolchain = {'name': 'foss', 'version': '2023b'}
 
 builddependencies = [
-    ('hypothesis', '6.133.2'),
+    ('hypothesis', '6.90.0'),
+    ('Cython', '3.0.10'),
     ('setuptools', '80.9.0'),
-    ('Cython', '3.1.1'),
 ]
 
 dependencies = [
-    ('Python', '3.13.1'),
-    ('SciPy-bundle', '2025.06'),
-    ('CUDA', '12.8.0', '', SYSTEM),
-    ('NCCL', '2.27.7', versionsuffix),
-    ('cuTENSOR', '2.3.0.6', versionsuffix, SYSTEM),
+    ('Python', '3.11.5'),
+    ('SciPy-bundle', '2023.11'),
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('NCCL', '2.20.5', versionsuffix),
+    ('cuTENSOR', '2.0.2.5', versionsuffix, SYSTEM),
     ('cuSPARSELt', '0.8.0.4', versionsuffix, SYSTEM),  # docs say 0.7.0 or 0.7.1
 ]
 

Updated software cuSPARSELt-0.8.0.4-CUDA-12.4.0.eb

Diff against cuSPARSELt-0.6.3.2-CUDA-12.6.0.eb

easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.6.3.2-CUDA-12.6.0.eb

diff --git a/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.6.3.2-CUDA-12.6.0.eb b/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.8.0.4-CUDA-12.4.0.eb
index 865a0d7771..e59c12b3bf 100644
--- a/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.6.3.2-CUDA-12.6.0.eb
+++ b/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.8.0.4-CUDA-12.4.0.eb
@@ -1,7 +1,7 @@
 easyblock = 'Tarball'
 
 name = 'cuSPARSELt'
-version = '0.6.3.2'
+version = '0.8.0.4'
 versionsuffix = '-CUDA-%(cudaver)s'
 
 homepage = 'https://docs.nvidia.com/cuda/cusparselt/index.html'
@@ -10,17 +10,20 @@ which at least one operand is a sparse matrix"""
 
 toolchain = SYSTEM
 
+local_cudamajver = '12'
+
 local_arch = {'arm64': 'sbsa', 'aarch64': 'sbsa'}.get(ARCH, ARCH)
 source_urls = ['https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-%s/' % local_arch]
-sources = ['libcusparse_lt-linux-%s-%%(version)s-archive.tar.xz' % local_arch]
+sources = ['libcusparse_lt-linux-%s-%%(version)s_cuda%s-archive.tar.xz' %
+           (local_arch, local_cudamajver)]
 checksums = [{
-    'libcusparse_lt-linux-x86_64-%(version)s-archive.tar.xz':
-        'a2f856e78943f5c538bdef1c9edc64a5ed30bf8bb7d5fcb615c684ffe776cc31',
-    'libcusparse_lt-linux-sbsa-%(version)s-archive.tar.xz':
-        '3e420ddbff4eb9ac603f57c7aa8b3d5271112816e244eb55ef9f30c4eb6a04b7',
+    'libcusparse_lt-linux-x86_64-%%(version)s_cuda%s-archive.tar.xz' % local_cudamajver:
+        '483954591766bade877becef126d53908d5fef5d7468b503736af37388669c08',
+    'libcusparse_lt-linux-sbsa-%%(version)s_cuda%s-archive.tar.xz' % local_cudamajver:
+        'b59e2f8ffd154b156b2d74ccd7cad7775385693bec8cb9562596060072c515f2',
 }]
 
-dependencies = [('CUDA', '12.6.0')]
+dependencies = [('CUDA', '12.4.0')]
 
 sanity_check_paths = {
     'files': ['include/cusparseLt.h',
Diff against cuSPARSELt-0.8.0.4-CUDA-12.6.0.eb

easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.8.0.4-CUDA-12.6.0.eb

diff --git a/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.8.0.4-CUDA-12.6.0.eb b/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.8.0.4-CUDA-12.4.0.eb
index 59c5b54d98..e59c12b3bf 100644
--- a/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.8.0.4-CUDA-12.6.0.eb
+++ b/easybuild/easyconfigs/c/cuSPARSELt/cuSPARSELt-0.8.0.4-CUDA-12.4.0.eb
@@ -23,7 +23,7 @@ checksums = [{
         'b59e2f8ffd154b156b2d74ccd7cad7775385693bec8cb9562596060072c515f2',
 }]
 
-dependencies = [('CUDA', '12.6.0')]
+dependencies = [('CUDA', '12.4.0')]
 
 sanity_check_paths = {
     'files': ['include/cusparseLt.h',

Updated software cuTENSOR-2.0.2.5-CUDA-12.4.0.eb

Diff against cuTENSOR-2.3.1.0-CUDA-13.0.0.eb

easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.3.1.0-CUDA-13.0.0.eb

diff --git a/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.3.1.0-CUDA-13.0.0.eb b/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.0.2.5-CUDA-12.4.0.eb
index a8be63328f..dd4624817a 100644
--- a/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.3.1.0-CUDA-13.0.0.eb
+++ b/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.0.2.5-CUDA-12.4.0.eb
@@ -1,7 +1,7 @@
 easyblock = 'Tarball'
 
 name = 'cuTENSOR'
-version = '2.3.1.0'
+version = '2.0.2.5'
 versionsuffix = '-CUDA-%(cudaver)s'
 
 homepage = 'https://developer.nvidia.com/cutensor'
@@ -10,29 +10,31 @@ description = """The cuTENSOR Library is a GPU-accelerated tensor linear algebra
 
 toolchain = SYSTEM
 
-local_arch = {
-    'aarch64': 'sbsa'
-}
 source_urls = [
-    'https://developer.download.nvidia.com/compute/cutensor/redist/libcutensor/linux-%s/' % local_arch.get(ARCH, ARCH)
+    'https://developer.download.nvidia.com/compute/cutensor/redist/libcutensor/linux-%(arch)s/'
 ]
-sources = ['libcutensor-linux-%s-%%(version)s_cuda13-archive.tar.xz' % local_arch.get(ARCH, ARCH)]
+sources = ['libcutensor-linux-%(arch)s-%(version)s-archive.tar.xz']
 
 checksums = [{
-    'libcutensor-linux-sbsa-%(version)s_cuda13-archive.tar.xz':
-        '2e4c24bd1621dac7497ca9edf90bfc5dbdcc38490dafd35821066f96f2934aef',
-    'libcutensor-linux-x86_64-%(version)s_cuda13-archive.tar.xz':
-        '9cb1125f7de01ca319b5c72edeb7169b679b72beacc90354fb18a14056e24372',
+    'libcutensor-linux-sbsa-%(version)s-archive.tar.xz':
+        '5163dd40f11f328e469a6d9b0056c8346f5d59ed538c18d6b954e4ae657c69cc',
+    'libcutensor-linux-x86_64-%(version)s-archive.tar.xz':
+        '0e957ae7b352f599de34b6fa1ba999b0617887f885d7436ac5737d71a6b83baa',
 }]
 
-
-dependencies = [('CUDA', '13.0.0')]
+local_cudamajver = '12'
+dependencies = [('CUDA', '12.4.0')]
 
 sanity_check_paths = {
     'files': ['include/cutensor.h', 'include/cutensor/types.h',
-              'lib/libcutensor.%s' % (SHLIB_EXT),
-              'lib/libcutensor_static.a'],
+              'lib/%s/libcutensor.%s' % (local_cudamajver, SHLIB_EXT),
+              'lib/%s/libcutensor_static.a' % local_cudamajver],
     'dirs': [],
 }
 
+modextrapaths = {
+    'LD_LIBRARY_PATH': ['lib/%s' % local_cudamajver],
+    'LIBRARY_PATH': ['lib/%s' % local_cudamajver],
+}
+
 moduleclass = 'lib'
Diff against cuTENSOR-2.3.1.0-CUDA-12.9.1.eb

easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.3.1.0-CUDA-12.9.1.eb

diff --git a/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.3.1.0-CUDA-12.9.1.eb b/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.0.2.5-CUDA-12.4.0.eb
index 73ee6c5139..dd4624817a 100644
--- a/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.3.1.0-CUDA-12.9.1.eb
+++ b/easybuild/easyconfigs/c/cuTENSOR/cuTENSOR-2.0.2.5-CUDA-12.4.0.eb
@@ -1,7 +1,7 @@
 easyblock = 'Tarball'
 
 name = 'cuTENSOR'
-version = '2.3.1.0'
+version = '2.0.2.5'
 versionsuffix = '-CUDA-%(cudaver)s'
 
 homepage = 'https://developer.nvidia.com/cutensor'
@@ -10,26 +10,31 @@ description = """The cuTENSOR Library is a GPU-accelerated tensor linear algebra
 
 toolchain = SYSTEM
 
-local_arch = {
-    'aarch64': 'sbsa'
-}
 source_urls = [
-    'https://developer.download.nvidia.com/compute/cutensor/redist/libcutensor/linux-%s/' % local_arch.get(ARCH, ARCH)
+    'https://developer.download.nvidia.com/compute/cutensor/redist/libcutensor/linux-%(arch)s/'
 ]
-sources = ['libcutensor-linux-%s-%%(version)s_cuda12-archive.tar.xz' % local_arch.get(ARCH, ARCH)]
+sources = ['libcutensor-linux-%(arch)s-%(version)s-archive.tar.xz']
 
 checksums = [{
-    'libcutensor-linux-sbsa-%(version)s_cuda12-archive.tar.xz':
-        'f3763cdc7b03ca08e348efb6faa35d461537390ce7d059e279e415b33dad8291',
-    'libcutensor-linux-x86_64-%(version)s_cuda12-archive.tar.xz':
-        'b1d7ad37b24cd66a446ae76ac33bd5125aa58007a604cb64fc9c014a8d685940',
+    'libcutensor-linux-sbsa-%(version)s-archive.tar.xz':
+        '5163dd40f11f328e469a6d9b0056c8346f5d59ed538c18d6b954e4ae657c69cc',
+    'libcutensor-linux-x86_64-%(version)s-archive.tar.xz':
+        '0e957ae7b352f599de34b6fa1ba999b0617887f885d7436ac5737d71a6b83baa',
 }]
-dependencies = [('CUDA', '12.9.1')]
+
+local_cudamajver = '12'
+dependencies = [('CUDA', '12.4.0')]
 
 sanity_check_paths = {
     'files': ['include/cutensor.h', 'include/cutensor/types.h',
-              'lib/libcutensor.%s' % (SHLIB_EXT),
-              'lib/libcutensor_static.a'],
+              'lib/%s/libcutensor.%s' % (local_cudamajver, SHLIB_EXT),
+              'lib/%s/libcutensor_static.a' % local_cudamajver],
     'dirs': [],
 }
+
+modextrapaths = {
+    'LD_LIBRARY_PATH': ['lib/%s' % local_cudamajver],
+    'LIBRARY_PATH': ['lib/%s' % local_cudamajver],
+}
+
 moduleclass = 'lib'

Updated software DLPack-1.2-GCCcore-13.2.0.eb

Diff against DLPack-1.2-GCCcore-13.3.0.eb

easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-13.3.0.eb

diff --git a/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-13.3.0.eb b/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-13.2.0.eb
index ab36d9167f..afb81cf369 100644
--- a/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-13.3.0.eb
+++ b/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-13.2.0.eb
@@ -7,7 +7,7 @@ homepage = 'https://dmlc.github.io/dlpack/latest/'
 description = """DLPack is a stable in-memory data structure for an ndarray
 system to interact with a variety of frameworks."""
 
-toolchain = {'name': 'GCCcore', 'version': '13.3.0'}
+toolchain = {'name': 'GCCcore', 'version': '13.2.0'}
 
 github_account = 'dmlc'
 source_urls = [GITHUB_LOWER_SOURCE]
@@ -15,8 +15,8 @@ sources = ['v%(version)s.tar.gz']
 checksums = ['58284a3b004a48450c958a23b30274527ebaf35a061124bbd4193fffa45efbd6']
 
 builddependencies = [
-    ('binutils', '2.42'),
-    ('CMake', '3.31.8'),
+    ('binutils', '2.40'),
+    ('CMake', '3.27.6'),
 ]
 
 sanity_check_paths = {
Diff against DLPack-1.2-GCCcore-14.3.0.eb

easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-14.3.0.eb

diff --git a/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-14.3.0.eb b/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-13.2.0.eb
index 043af0c96e..afb81cf369 100644
--- a/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-14.3.0.eb
+++ b/easybuild/easyconfigs/d/DLPack/DLPack-1.2-GCCcore-13.2.0.eb
@@ -7,7 +7,7 @@ homepage = 'https://dmlc.github.io/dlpack/latest/'
 description = """DLPack is a stable in-memory data structure for an ndarray
 system to interact with a variety of frameworks."""
 
-toolchain = {'name': 'GCCcore', 'version': '14.3.0'}
+toolchain = {'name': 'GCCcore', 'version': '13.2.0'}
 
 github_account = 'dmlc'
 source_urls = [GITHUB_LOWER_SOURCE]
@@ -15,8 +15,8 @@ sources = ['v%(version)s.tar.gz']
 checksums = ['58284a3b004a48450c958a23b30274527ebaf35a061124bbd4193fffa45efbd6']
 
 builddependencies = [
-    ('binutils', '2.44'),
-    ('CMake', '4.0.3'),
+    ('binutils', '2.40'),
+    ('CMake', '3.27.6'),
 ]
 
 sanity_check_paths = {

Updated software pytest-forked-1.6.0-GCCcore-13.2.0.eb

Diff against pytest-forked-1.6.0-GCCcore-14.3.0.eb

easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-14.3.0.eb

diff --git a/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-14.3.0.eb b/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-13.2.0.eb
index 76ac2f7cb2..56cbb558a2 100644
--- a/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-14.3.0.eb
+++ b/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-13.2.0.eb
@@ -6,17 +6,17 @@ version = '1.6.0'
 homepage = 'https://github.com/pytest-dev/pytest-forked'
 description = "Run tests in isolated forked subprocesses."
 
-toolchain = {'name': 'GCCcore', 'version': '14.3.0'}
+toolchain = {'name': 'GCCcore', 'version': '13.2.0'}
 
 sources = [SOURCE_TAR_GZ]
 checksums = ['4dafd46a9a600f65d822b8f605133ecf5b3e1941ebb3588e943b4e3eb71a5a3f']
 
 builddependencies = [
-    ('binutils', '2.44'),
+    ('binutils', '2.40'),
 ]
 dependencies = [
-    ('Python', '3.13.5'),
-    ('Python-bundle-PyPI', '2025.07'),
+    ('Python', '3.11.5'),
+    ('Python-bundle-PyPI', '2023.10'),
 ]
 
 moduleclass = 'tools'
Diff against pytest-forked-1.6.0-GCCcore-14.2.0.eb

easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-14.2.0.eb

diff --git a/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-14.2.0.eb b/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-13.2.0.eb
index d2dd0aa28f..56cbb558a2 100644
--- a/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-14.2.0.eb
+++ b/easybuild/easyconfigs/p/pytest-forked/pytest-forked-1.6.0-GCCcore-13.2.0.eb
@@ -6,17 +6,17 @@ version = '1.6.0'
 homepage = 'https://github.com/pytest-dev/pytest-forked'
 description = "Run tests in isolated forked subprocesses."
 
-toolchain = {'name': 'GCCcore', 'version': '14.2.0'}
+toolchain = {'name': 'GCCcore', 'version': '13.2.0'}
 
 sources = [SOURCE_TAR_GZ]
 checksums = ['4dafd46a9a600f65d822b8f605133ecf5b3e1941ebb3588e943b4e3eb71a5a3f']
 
 builddependencies = [
-    ('binutils', '2.42'),
+    ('binutils', '2.40'),
 ]
 dependencies = [
-    ('Python', '3.13.1'),
-    ('Python-bundle-PyPI', '2025.04'),
+    ('Python', '3.11.5'),
+    ('Python-bundle-PyPI', '2023.10'),
 ]
 
 moduleclass = 'tools'

@Flamefire
Copy link
Copy Markdown
Contributor Author

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#4082
FAILED
Build succeeded for 7 out of 8 (total: 2 hours 51 mins 23 secs) (8 easyconfigs in total)
c21 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 580.65.06, Python 3.9.21
See https://gist.github.com/Flamefire/b2fed2e5ac995647c342289c3fd53902 for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#4082
FAILED
Build succeeded for 7 out of 8 (total: 2 hours 48 mins 59 secs) (8 easyconfigs in total)
i8020 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 7352 24-Core Processor (zen2), 8 x NVIDIA NVIDIA A100-SXM4-40GB, 580.65.06, Python 3.9.21
See https://gist.github.com/Flamefire/4bb47c31d9cda9e7722bfee315f12d68 for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#4082
FAILED
Build succeeded for 7 out of 8 (total: 3 hours 5 mins 2 secs) (8 easyconfigs in total)
c50 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 580.65.06, Python 3.9.21
See https://gist.github.com/Flamefire/4714c5d88e8369017631ad88764a1e51 for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#4082
SUCCESS
Build succeeded for 8 out of 8 (total: 3 hours 20 mins 6 secs) (8 easyconfigs in total)
c56 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 580.65.06, Python 3.9.21
See https://gist.github.com/Flamefire/5b4bfcd802f31a4147f1541205a0dc81 for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#4082
FAILED
Build succeeded for 7 out of 8 (total: 3 hours 10 mins 51 secs) (8 easyconfigs in total)
i8023 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 7352 24-Core Processor (zen2), 8 x NVIDIA NVIDIA A100-SXM4-40GB, 580.65.06, Python 3.9.21
See https://gist.github.com/Flamefire/880b8da6a6af11fcfcbdd1d1af6df7ee for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#4082
FAILED
Build succeeded for 7 out of 8 (total: 4 hours 44 mins 41 secs) (8 easyconfigs in total)
c116 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 580.65.06, Python 3.9.21
See https://gist.github.com/Flamefire/1b5d0fbe0e64948afbdb4826e9397929 for a full test report.

@Flamefire
Copy link
Copy Markdown
Contributor Author

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#4082
FAILED
Build succeeded for 7 out of 8 (total: 5 hours 27 mins 28 secs) (8 easyconfigs in total)
i8035 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 7352 24-Core Processor (zen2), Python 3.9.21
See https://gist.github.com/Flamefire/80cd1b950125d9d42cd0522e05bb115d for a full test report.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant