GitLab

**Description**
Integrate FAMBench into superbench based on docker implementation:
https://github.com/facebookresearch/FAMBench

The script to run all benchmarks is:
https://github.com/facebookresearch/FAMBench/blob/main/benchmarks/run_all.sh

**Description**
Integrate output all nodes diagnosis results.

**Description**
Output results of all nodes in data diagnosis.

**Description**
Add usage for result summary.

**Description**
Integrage result summary and update output format of data diagnosis.

**Major Revision**
- integrage result summary 
- add md and html format for data diagnosis

**Description**
Use config `log_raw_data` to control whether log the raw data into file or not. The default value is `no`. We can set it as `yes` for some particular benchmarks to save the raw data into file, such as NCCL/RCCL test.

Bumps [minimist](https://github.com/substack/minimist) from 1.2.5 to 1.2.6.
- [Release notes](https://github.com/substack/minimist/releases)
- [Commits](https://github.com/substack/minimist/compare/1.2.5...1.2.6

)

---
updated-dependencies:
- dependency-name: minimist
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

**Description**
Add result summary in excel,md,html format.

**Major Revision**
- Add ResultSummary class to support result summary in excel,md,html format.
- Abstract RuleBase class for common-used functions in DataDiagnosis and ResultSummary.

**Description**
Remove fp16 samples type converting time for training cnn and lstm inference.

Add inference config for preview SKUs, including:
* [NC96ads_A100_v4](https://docs.microsoft.com/en-us/azure/virtual-machines/nc-a100-v4-series)
* [NV18ads_A10_v5](https://docs.microsoft.com/en-us/azure/virtual-machines/nva10v5-series)

**Description**
Remove fp16  samples type converting time for cnn and lstm models.

**Description**
Modifications adding GPU-Burn to SuperBench.
- added third party submodule
- modified Makefile to make gpu-burn binary
- added/modified microbenchmarks to add gpu-burn python scripts
- modified default and azure_ndv4 configs to add gpu-burn

**Description**
fix the bug in result writing to files for mpi mode.

**Description**
Add md and html output format for DataDiagnosis.

**Major Revision**
- add md and html support in file_handler
- add interface in DataDiagnosis for md and HTML output

**Minor Revision**
- move excel and json output interface into DataDiagnosis

Fix env file path to absolute path in `docker exec`, in case there're mixed ssh and local connections or different users are used.

**Description**
Abstract RuleBase from DataDiagnosis.

Bumps [url-parse](https://github.com/unshiftio/url-parse) from 1.5.8 to 1.5.10.
- [Release notes](https://github.com/unshiftio/url-parse/releases)
- [Commits](https://github.com/unshiftio/url-parse/compare/1.5.8...1.5.10

)

---
updated-dependencies:
- dependency-name: url-parse
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

**Description**
The BatchNorm operator is not numerically stable in fp16.  PyTorch documentation recommends to keep the BN op in fp32 for fp16 AMP models.  Refer to https://pytorch.org/docs/stable/amp.html#ops-that-can-autocast-to-float32.  Preserving BN in fp32 for superbench more accurately reflects real workloads.

**Description**
Add dockerfile for rocm5.0.1.

Bumps [prismjs](https://github.com/PrismJS/prism) from 1.23.0 to 1.27.0.
- [Release notes](https://github.com/PrismJS/prism/releases)
- [Changelog](https://github.com/PrismJS/prism/blob/master/CHANGELOG.md)
- [Commits](https://github.com/PrismJS/prism/compare/v1.23.0...v1.27.0

)

---
updated-dependencies:
- dependency-name: prismjs
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

**Description**
Add rocm5.0 dockerfile.

**Description**
Fix invalid reference of P2P detection result in gpu_copy.

**Description**
Make gpcnet only for cuda.

**Description**
Fix HIP_ARCHITECTURES is empty issue with cmake>=3.21.0.
Refer to https://github.com/ROCm-Developer-Tools/HIP/pull/2364

Bumps [url-parse](https://github.com/unshiftio/url-parse) from 1.5.1 to 1.5.8.
- [Release notes](https://github.com/unshiftio/url-parse/releases)
- [Commits](https://github.com/unshiftio/url-parse/compare/1.5.1...1.5.8

)

---
updated-dependencies:
- dependency-name: url-parse
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Add T4 configurations for inference.

**Description**
Add multi-rules feature for data diagnosis to support multiple rules' combined check.

**Major Revision**
- revise rule design to support multiple rules combination check
- update related codes and tests

Fix env file path for `docker run`.

Bumps [follow-redirects](https://github.com/follow-redirects/follow-redirects) from 1.14.7 to 1.14.8.
- [Release notes](https://github.com/follow-redirects/follow-redirects/releases)
- [Commits](https://github.com/follow-redirects/follow-redirects/compare/v1.14.7...v1.14.8

)

---
updated-dependencies:
- dependency-name: follow-redirects
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

**Description**
Add support for pytorch>=1.9.0 of init_process_group.

**Major Revision**
- Use PrefixStore(TCPStore) to init_process_group manully for each model run

**Description**
Update rccl-tests submodule to fix divide by zero error.

**Description**
This commit remove NUMA binding for device-to-device tests because NUMA doesn't affect performance, and revise benchmark metrics accordingly.

This commit adds GDR-only nccl-tests for Nvidia machines. Also bump NCCL to v2.10.3-1 to achieve peak performance in this test.

This commit makes data checking in gpu_copy optional, because it will take too long time if message size is large.

**Description**
This commit does the following to optimize result variance in gpu_copy benchmark:
1) Add warmup phase for gpu_copy benchmark to avoid timing instability caused by first-time CUDA kernel launch overhead;
2) Use CUDA events for timing instead of CPU timestamps;
3) Make data checking an option that is not preferred to be enabled in performance test;
4) Enlarge message size in performance benchmark.

Fix typo in document.

Support T4 and A10 in GEMM benchmark.

Support customized env for all modes in configuration.

Fix bug of GPU scan logic in bidirectional tests.

**Description**
Please write a brief description and link the related issue if have.

**Major Revision**
- Sync (do allreduce max) the E2E training results among all workers.
- Avoid using ':0' in metric name if there has only one rank having output.