Current Releases
| Project | DL Package | MVAPICH Library | Userguide |
| HPC-AI | HPC-AI v1.0 PyTorch 2.10.0 DeepSpeed vLLM SGLang MCR-DL MAC-Attention |
MVAPICH-Plus | HPC-AI v1.0 Userguide PyTorch 2.10.0 Userguide DeepSpeed Userguide vLLM Userguide SGLang Userguide MCR-DL Userguide MAC-Attention Userguide |
Older Releases
| Project | DL Package | MVAPICH Library | Userguide |
| ParaInfer-X | ParaInfer-X v1.0 | MVAPICH-Plus | Userguide |
| MPI-Driven DL Training (MPI4DL) | MPI4DL | MVAPICH-Plus |
Userguide |
| MPI-Driven ML Training (MPI4cuML) |
mpi4cuml-0.5 |
MVAPICH-Plus | Userguide |
| Data-Parallel Training with Caffe | OSU-Caffe | MVAPICH-Plus | Userguide |
| High-Performance InfiniBand Support for TensorFlow | RDMA-TensorFlow | MVAPICH-Plus | Userguide |
PyTorch with MVAPICH-Plus
- GitHub for PyTorch 2.10.0 with advanced GPU-Aware MPI backend
- Please email us at hidl-discuss@lists.osu.edu if you encounter any issues. Also let us know if you experience any trouble installing the library on your system.
MCR-DL
- The MCR-DL v0.1 library is distributed under the Apache License.
- MCR-DL (12/22/2023).Total Clones: 669
- GitHub for MCR-DL.
- Please email us at hidl-discuss@lists.osu.edu if you encounter any issues. Also let us know if you experience any trouble installing the library on your system.
ParaInfer-X
- The ParaInfer-X v1.0 library includes Flover, which is distributed under the Apache License.
- ParaInfer-X 1.0 (11/09/2023).Total Clones: 3931
- GitHub for ParaInfer-X.
- Please email us at hidl-discuss@lists.osu.edu if you encounter any issues. Also let us know if you experience any trouble installing the library on your system.
MPI4DL
- The MPI4DL library is distributed under the Apache License.
- MPI4DL 0.6 (11/09/2023).Total Clones: 1047
- GitHub for MPI4DL.
- Please email us at hidl-discuss@lists.osu.edu if you encounter any issues. Also let us know if you experience any trouble installing the library on your system.
MPI4cuML 0.5
- The MPI4cuML library is distributed under the BSD License.
- MPI4cuML 0.5 (11/10/2022).
- Userguide for MPI4cuML 0.5.
- Please email us at hidl-discuss@lists.osu.edu if you encounter any issues. Also let us know if you experience any trouble installing the library on your system.
OSU-Caffe 0.9
- The OSU-Caffe library is distributed under the BSD License.
- OSU-Caffe 0.9 (11/10/2016).
-
As an initial technology preview, we are providing RHEL6 and RHEL7
RPMs. These RPMs contain the OSU-Caffe software on the corresponding
distro. Please note that the RHEL RPMs are
compatible with CentOS as well. For Debian/Ubuntu users, please
follow the instructions in the
install section in the userguide.
RHEL/CENTOS 7 RPMs OFED GNU INTEL MLNX-OFED 3.2 [CUDA 7.5] [CUDA 7.5] RHEL/CENTOS 6 RPMs OFED GNU INTEL MLNX-OFED 3.2 [CUDA 7.5] [CUDA 7.5] -
Please note that all RPMs where built using the following version of the respective compilers:
- gcc (GCC) 4.9.3
- icc (ICC) 15.0.0 20140723
- For INSTALL and usage information please refer to the userguide
- Please email us at mvapich-help@cse.ohio-state.edu if you require a RPM built against a different version of CUDA or OFED. Also let us know if you experience any trouble installing the library on your system.
RDMA-TensorFlow 0.9.1
- The RDMA-TensorFlow library is distributed under the BSD License.
- RDMA-TensorFlow 0.9.1 (07/23/2018).
- Userguide
- Download rdma_tensorflow-0.9.1
-
Please note that the whl package were built using the following:
- gcc (GCC) 4.8.5
- glibc 2.17
- Python 2.7
- CUDA 8.0
- CUDNN 5.0
- Please email us at hidl-discuss@lists.osu.edu if you encounter any issues. Also let us know if you experience any trouble installing the library on your system.
Contact: Dhabaleswar K. (DK) Panda
Dept of Computer Science and Engineering
2001-2026 NBCL. All rights reserved.
Dept of Computer Science and Engineering
2001-2026 NBCL. All rights reserved.
774 Dreese Laboratories
2015 Neil Avenue
Columbus, OH 43210
2015 Neil Avenue
Columbus, OH 43210