Commits · b592a7c793ff8b8235681030d51ae196b91d0f5a · tsoc / superbenchmark

21 Oct, 2021 1 commit

Benchmarks: Build Pipeline - Add gpcnet as git submodule and building logic (#228) · b592a7c7

Yuting Jiang authored Oct 21, 2021

**Description**
Add gpcnet as git submodule and building logic.

**Major Revision**
- add gpcnet as a submodule
- add build logic in third_party/Makefile

b592a7c7

02 Sep, 2021 1 commit

Dockerfile - Fix ulimit nofile in Docker images (#183) · 4e431f11

Yifan Xiong authored Sep 02, 2021

__Description__

Resolve "too many open files" issue when runnning NCCL/RCCL on multiple nodes using Docker images, increase nofile number in limits.conf.

4e431f11

01 Sep, 2021 2 commits
- Dockerfile: Add Package - Install openmpi for ROCm images (#181) · 115cd2e6
  guoshzhao authored Sep 01, 2021
```
**Description**
Install openmpi-4.0.0 for ROCm images.
```
  115cd2e6
- Benchmarks: Docker Benchmarks - Setup Docker-in-Docker environment (#180) · 7d947757
  guoshzhao authored Sep 01, 2021
```
**Description**
Setup docker environment in docker container.

**Major Revision**
- Install docker client for cuda and rocm images.
- Mount /var/run/docker.sock from host
```
  7d947757
31 Aug, 2021 1 commit
- Dockerfile: Add dockerfile - Add rocm 4.0 and 4.2 dockerfile with pytorch1.7.0 (#164) · a7f508e4
  guoshzhao authored Aug 31, 2021
```
**Description**
Add dockerfile `rocm4.0-pytorch1.7.0.dockerfile` and `rocm4.2-pytorch1.7.0.dockerfile` for `rocm` platform.
```
  a7f508e4