GitLab

Fix race condition issue when multi ranks rotating the same directory.

Fetch benchmarks results on all nodes, will rsync after each benchmark.
The results directory structure on control node is as follows:

```
outputs/
└── datetime
    ├── nodes
    │   └── node-0
    │       ├── benchmarks
    │       │   ├── benchmark-0
    │       │   │   ├── rank-0
    │       │   │   │   └── results.json
    │       └── sb-exec.log
    ├── sb-run.log
    └── sb.config.yaml
```

* Support custom output directory.
* Update document.

* fix bug for nvidia v100
* hard code the supported dict for different arch.

* replace torch.optim.AdamW with transformers.AdamW.

* Add `sb deploy` command content.
* Fix inline if-expression syntax in playbook.
* Fix quote escape issue in bash command.
* Add custom env in config.
* Update default config for multi GPU benchmarks.
* Update MANIFEST.in to include jinja2 template.
* Require jinja2 minimum version.
* Fix occasional duplicate output in Ansible runner.
* Fix mixed color from Ansible and Python colorlog.
* Update according to comments.
* Change superbench.env from list to dict in config file.

Fix bugs and refine log in single GPU benchmarks:

* Fix none framework issue
* Fix empty parameter bug
* Remove missed mobilenet_v3 models
* Change benchmark registration log to debug level
* Add pid in logging
* Add missing benchmarks in default config
* Fix deprecated logging warn

* Clean up the cache.

* fix return code reset issue

* Support local mode in runner.

* Change default shape of sharding-matmul.

* add cuda flops performance benchmark.

* add python related cudnn microbenchmark

* Save benchmark results to json file.

* add benchmark for cublas test

* format

* revise error handling and test

* add interface to read json file, revise json file path and include .json in packaging

* add random_seed in arguments

* revise preprocess of cublas benchmark

* fix lint error and note error in source code

* update according comments

* revise input arguments from json file to custom str and convert json file to built-in dict list

* restore package config

* fit lint issue

* update platform and comments

* rename files to match source code dir and fix comments error
Co-authored-by: root <root@sb-validation-000001.51z1chmys5fuzfqyo4niepozre.bx.internal.cloudapp.net>

* Support `torch.distributed` mode in runner.
* Support given `proc_num` and `node_num` in `torch.distributed` mode.

* Add integration tests for Ansible playbooks
* Add `gpu_vendor` var to bypass gpu mount

Implement ansible client and runner:
* add ansible client
* add deploy and check_env playbooks

* add kernel launch overhead benchmark.

* use absolute path of input file
* parse registry uri from image
* merge common parts for arguments processing

* Revise MicroBenchmark class to be more flexible.
* use command index not the command as the parameter.
* changes according to discussion.

* Benchmarks: Add Benchmark - Add LSTM model benchmarks.

* Benchmarks: Add Benchmark - Add CNN model benchmarks.

Benchmarks: Code Revision - Fix some issue for BERT benchmark. (#58)

* Benchmarks: Add Benchmark - Add GPT2 model benchmark.

* Benchmarks: Add Feature - Add interface to get all predefine parameters of all benchmarks.

* Benchmarks: Add Benchmark - add computation and communication overlap micro benchmark

* Benchmarks: Add benchmark - fix some format issues and typo

* Benchmarks: Add Benchmark - update according comments and add test

* revise tests

* skip multi gpu test due to no multi gpu
Co-authored-by: v-yujiang <v-yujiang@microsoft.com>

* Superbenchmark: Revise tests - revise benchmark test util to support multi gpu test

* modify test_sharding_matmul.py to match the tests util

* fix missing package in dockerfile
* update benchmark list and parameters
* catch runtime errors
* refine logging info

* add tests for matmul and sharding-matmul benchmarks.

* add decorator for sharding_matmul tests.

* add __init__.py for utils of benchmarks tests.

* disable GPU tests for CPU platform validation.

* fix typo
Co-authored-by: Guoshuai Zhao <guzhao@microsoft.com>
Co-authored-by: Peng Cheng <chengpeng5555@outlook.com>
Co-authored-by: Yifan Xiong <yifan.xiong@microsoft.com>

Co-authored-by: Guoshuai Zhao <guzhao@microsoft.com>
Co-authored-by: Yifan Xiong <yifan.xiong@microsoft.com>

* CLI integration with Executor and Runner

* init runner class with not implemented

* skip unnecessary tests according to env var
* remove useless tests

Add superbench executor class

* add executor class
* update default config to exec benchmarks
* add micro benchmarks and model benchmarks

Benchmarks: Code Revision - Revise BenchmarkRegistry interfaces for integration with executor. (#33)

* revise BenchmarkRegistry interfaces.
* address comments
Co-authored-by: Guoshuai Zhao <guzhao@microsoft.com>

Benchmarks: Add Benchmark - Add Pytorch BERT benchmarks, including bert-base and bert-large.   (#20)

* add pytorch bert benchmarks.

* revise code

* fix issue

* revise code.
Co-authored-by: Guoshuai Zhao <guzhao@microsoft.com>

* move benchmarks registration from registry.py to __init__.py

* revise __init__.
Co-authored-by: Guoshuai Zhao <guzhao@microsoft.com>

* add sample_count argument.

* handle more condidatins.

* address comments.
Co-authored-by: Guoshuai Zhao <guzhao@microsoft.com>