GitLab

**Description**
Fix wrong unit of cpu-memory-bw-latency in doc.

Update cuda11.8 image to cuda12.1 based on nvcr23.03 and related versions in the image:
* cuda 11.8 -> 12.1
* nccl 2.15.5 -> 2.17.1
* hpcx: 2.8 -> 2.14
* mlc: 3.9a -> 3.10

Remove unreachable condition when write host list in mpi mode.

Change num_workers to configurable in model benchmark data loader.

**Description**
Collect realtime GPU power when benchmarking.

**Description**
Fix bug to get metric from cmd when error happens(cudnn-function/_time:4)

**Description**
Analyzer: Fix bug in python3.8 due to pandas api change.

**Major Revision**
- force check numeric only in dataframe for analysis
- dataframe.append -> pd.concat
- pd.ExcelWriter.save() -> pd.ExcelWriter.close()

**Description**
This commit fixes wrong `torch.empty_like` usage and missing dtype and
device argument in communication wrappers.

Fix matrix size overflow issue when cast from int to size_t implicitly.

**Description**
Looks `grep cgroup /proc/filesystems` doesn't work for NDv4 whose cgroup
version is v1, but the result of this command got v2 for NDv4. Instead,
checking the file existence to judge the cgroup version.

__Description__

Update TE FP8 model conversion.

__Major Revisions__
* Add 16-byte alignment comment.
* Fix TE layer parameters type.

Support Transformer Engine FP8 in existing PyTorch BERT/GPT2 models by
converting linear/layernorm to TE layers.

**Description**
This PR adds a micro-benchmark of distributed model inference workloads.

**Major Revision**
- Add a new micro-benchmark dist-inference.
- Add corresponding example and unit tests.
- Update configuration files to include this new micro-benchmark.
- Update micro-benchmark README.

---------
Co-authored-by: Peng Cheng <chengpeng5555@outlook.com>

**Description**
Since ubuntu 22.04 will use cgroup V2 and the file structure changed.
Modify the monitor to adapt to cgroup v1 and v2.

Support batch and shape range with multiplication factors in cublaslt
gemm benchmark.

**Description**

- Adding HPL benchmark

---------
Co-authored-by: Ubuntu <azureuser@sbtestvm.jzlku1oskncengjiado35wf1hd.ax.internal.cloudapp.net>
Co-authored-by: Peng Cheng <chengpeng5555@outlook.com>

Fix potential barrier timeout in init_process_group due to race
condition of using the same port. Change to different ports when running
multiple models sequentially in one process.
For example, when running vgg11/13/16/19, will use port 29501~29504
respectively.

**Description**
Support error tolerance in micro-benchmark for CuDNN function


**Major Revision**
- revise micro_base to support running the remaining commands run when
one command failed in the microbenchmark
- make error tolerance as true in cudnn functions

Support FP64/TF32/FP16/BF16 in cublaslt (batch) GEMM.

Bumps [webpack](https://github.com/webpack/webpack) from 5.39.1 to 5.76.1.
- [Release notes](https://github.com/webpack/webpack/releases

)
- [Commits](webpack/webpack@v5.39.1...v5.76.1)

---
updated-dependencies:
- dependency-name: webpack
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>

Pin setuptools version to
[v65.7.0](https://setuptools.pypa.io/en/latest/history.html#v65-7-0) to
avoid breaking changes since v66.0.0.

Limit ansible_runner version to less than 2.3.2 for Python3.6.

Benchmarks: Revision - Support flexible warmup and non-random data initialization in cublas-benchmark  (#479)

**Description**
revise cublas-benchmark for flexible warmup and fill data with fixed
number for perf test to improve the running efficiency.

**Major Revision**
- remove num_in_steps for warmup to support more flexible warmup setting
for users
- Add support to generate input with fixed number for perf test

**Description**
Add suppport to install cpu-only perftest in makefile.
Co-authored-by: Yuting Jiang <yuting.jiang@microsoft.com>
Co-authored-by: Peng Cheng <chengpeng5555@outlook.com>

Free more disk space in GitHub Action VHD.

**Description**
Upgrade networkx version to fix installation compatibility issue.

Bumps [@sideway/formula](https://github.com/sideway/formula) from 3.0.0 to 3.0.1.
- [Release notes](https://github.com/sideway/formula/releases

)
- [Commits](hapijs/formula@v3.0.0...v3.0.1)

---
updated-dependencies:
- dependency-name: "@sideway/formula"
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [http-cache-semantics](https://github.com/kornelski/http-cache-semantics) from 4.1.0 to 4.1.1.
- [Release notes](https://github.com/kornelski/http-cache-semantics/releases

)
- [Commits](kornelski/http-cache-semantics@v4.1.0...v4.1.1)

---
updated-dependencies:
- dependency-name: http-cache-semantics
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>

**Description**

- Added stream benchmark
- Added stream unit test
- Added stream example
- Modified docker files to build stream

---------
Co-authored-by: Ubuntu <azureuser@sbtestvm.jzlku1oskncengjiado35wf1hd.ax.internal.cloudapp.net>
Co-authored-by: Peng Cheng <chengpeng5555@outlook.com>
Co-authored-by: Yifan Xiong <xiongyf@yandex.com>

**Description**
Support SuperBench Executor running on Windows.

**Major Revision**
- Lazy import ansible related module

**Description**
The commit(e08b6d3a) installs a rccl
version which is causing "undefined symbol: ncclGetLastError" while
trying to import torch. Revert it to avoid the error.

Bumps [ua-parser-js](https://github.com/faisalman/ua-parser-js) from 0.7.28 to 0.7.33.
- [Release notes](https://github.com/faisalman/ua-parser-js/releases)
- [Changelog](https://github.com/faisalman/ua-parser-js/blob/master/changelog.md

)
- [Commits](faisalman/ua-parser-js@0.7.28...0.7.33)

---
updated-dependencies:
- dependency-name: ua-parser-js
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>

**Description**

Cherry-pick bug fixes from v0.7.0 to main.

**Major Revisions**

* Benchmarks - Fix missing include in FP8 benchmark (#460)
* Fix bug in TE BERT model (#461)
* Doc - Update benchmark doc (#465)
* Bug: Fix bug for incorrect datatype judgement in cublas-function
source code (#464)
* Support `sb deploy` without pulling image (#466)
* Docs - Upgrade version and release note (#467)
Co-authored-by: Russell J. Hewett <russell.j.hewett@gmail.com>
Co-authored-by: Yuting Jiang <yutingjiang@microsoft.com>

**Description**
Fix bug for incorrect datatype judgement in cublas-function source code.

Bumps [json5](https://github.com/json5/json5) from 1.0.1 to 1.0.2.
- [Release notes](https://github.com/json5/json5/releases)
- [Changelog](https://github.com/json5/json5/blob/main/CHANGELOG.md

)
- [Commits](json5/json5@v1.0.1...v1.0.2)

---
updated-dependencies:
- dependency-name: json5
  dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Support traffic patterns under the different devices in NCCL/RCCL test
* change the metrics format if specified the pattern

**Major Revision**

- Add an option for pattern to generate mpi_pattern.txt file if
specified the path.
- In mpi pattern, serial_index and parallel_index will add in each
benchmark as environment variables.

**Minor Revision**
- Fix typo

Support FP8 in PyTorch BERT models:

* add fp8 hybrid/e4m3/e5m2 in precision arguments
* build BERT encoders with `te.TransformerLayer` to repalce
`transformers.BertModel`
* wrap forward steps with fp8 autocast

**Description**
Support the following patterns  in `mpi` mode:
* `k-batch`
* `topo-aware`

Support GEMM benchmark on Hopper GPUs.